|
Skill Ripjaws V 32GB 3600Mhz
It does this by pushing the clock speed first to the boost clock speed and then to the base speed. Kind of like GPU Boost 3.0, but better. In a nutshell, the cooler you keep the card and core, the longer you can run your dynamic max boost clock speeds.
At the top right is a radio button that opens the MSI Overclocking Scanner. The GUI is not as visually appealing as the EVGA version, but it is just as functional.
0 clock speed of 2010MHz, or 270MHz over the factory boost clock speed of 1740MHz. That’s a roughly 16% bump in clock speed.MSI GeForce RTX 2070 Super Gaming X Trio
3.
Testing & Final thoughts1. Introduction & Packaging2. Visual inspection3. Testing Final & thoughts
Testing
The following system was used to test the graphics card:
– Processor: Intel i9 9900K @ 5 GHz
– Motherboard: ASRock Z390 Phantom Gaming X
– RAM: Ballistix Tactical Tracer 16GB DDR4 @ 2666MHz
– SSD: Samsung 840 EVO 120GB
– HDD: WD 1 TB Blue
– PSU: Seasonic Snow Silent 750W
– CPU Cooler: CoolerMater MasterLiquid 240
– Case: Phanteks Enthoo Luxe White
– Thermal compound: Arctic Cooling MX-4
To test the graphics card, the NVIDIA GeForce 441.34 driver was used, being the latest NVIDIA driver available at the time of writing the review.
All the games used to test the graphics cards are set to their highest quality settings, unless otherwise stated, with the resolution 1920 x 1080.
Both the games and benchmarks used were run three times.
The software used to measure the temperature and frames per second (FPS) are the following:
– AIDA64 Extreme 6.10.5200
– GPU-Z v2.26.0
– MSI Afterburner 4.6.0
– Fraps
The overclocking has been done using the default fan settings. Each graphic card overclocks differently due to different batches of memory chips used as well as the variation in the manufacturing process of the graphics core. This means that each graphics card will overclock differently in both frequencies obtained and voltage required.
In the case of the MSI GeForce RTX 2070 Super Gaming X Trio, a maximum overclock of 2002MHz Base Clock and 2004MHz Memory Clock was achieved.
The test results are as follows:
Grand Theft Auto V (Low 0.1%)
The Witcher 3: Wild Hunt (Low 0.1%)
Metro Exodus (Low 0.
1%)
Shadow Of The Tomb Raider (Low 0.1%)
Temperature
The Idle tests consists of the system left at the Desktop for 30 minutes, the power consumption is measured during the test.
Full Load: Represents the power consumption of the graphics card while playing a modern video game, in this case the benchmark of Tomb Raider 2013 with the following settings: 1920 x 1080, Ultra Preset, TressFX OFF, 4xAA.
Peak: Represents the power consumption of the graphics card during a synthetic benchmark, in this case FurMark 1.15.1.
The noise level was measured at three different distances from the fans (10cm, 20cm, 30cm.) the noise was measured with a margin of error of ± 5 db. In the case of the MSI GeForce RTX 2070 Super Gaming X Trio, the three fans will turn off when the graphic card has a light load.
This means, that during internet browsing, movie rendering or other light tasks, both fans will not spin to eliminate the noise completely.
Final thoughts
The MSI GeForce RTX 2070 Super Gaming X Trio is available for 515 US dollars, a whole 15 USD more than the Founder’s Edition RTX 2070 Super. For this price difference, the added features include a triple fan cooling system, plenty of RGB integration and a metallic backplate.
The design of this graphics card is similar with other RTX Gaming X models. It is also easy to match with any system configuration, thanks to the colors used for the backplate and the fan shroud. In addition, the RGB elements add more customizing options for the graphics card.
The performance out of the box is very good; however, the overclocking is limited by the power limit. The RTX 2070 Super Gaming X Trio has the same 215W power limit as the Founder’s Edition RTX 2070 Super. Nevertheless, the reviewed graphics card reached a manual overclock of 2002MHz Base Clock and 2004MHz Memory Clock.
The cooling system is effective and also silent, using three fans that have different sizes to accommodate the Nvidia NVLink connector. In addition, these fans will stop when the graphics card has a temperature lower than 60 °C. The heatsink uses six heatpipes to transfer the heat and plenty of thermal pads for cooling the Micro memory chips and the VRM components.
In conclusion, the MSI GeForce RTX 2070 Super Gaming X Trio is a very good graphics card, with a good gaming performance and silent cooling system. The price of 515 USD is on the same level as other RTX 2070 Super models, and is well worth it, when considering all the features offered. In addition, two games are also included with this graphics card, Control and Wolfenstein: Youngblood, which add more value to the overall package.
3. Testing & Final thoughts1. Introduction & Packaging2.
Visual inspection3. Testing Final & thoughts
How to overclock your CPU and GPU
Learning how to overclock your CPU and GPU is one of the easiest ways to improve your gaming PC’s performance. While overclocking used to be something only liquid-nitrogen equipped techies did, the practice is more user-friendly than ever in 2022, thanks to recent advancements in software and technology.
For those looking to make the most out of their silicon, knowing how to overclock is one of the most valuable things you can learn. Yes, you could always perform a CPU upgrade to boost fps, but why do that when you can increase performance for free instead?
With component prices continuing to rise, the argument for overclocking the best gaming CPU and best graphics card is stronger than ever right now, and there’s very little reason not to. Just follow the steps below to get yourself started, and after a bit of work (and maths) you’ll be wondering why you didn’t learn how to overclock sooner.
Before you get started, it’s worth bearing in mind that while overclocking is far less dangerous than it used to be – with fail-safes built into modern silicon – there are still risks involved. Running any hardware outside of officially rated parameters will likely invalidate your warranty, and there’s still a chance you could brick your components if you push them too hard. Suffice to say, you have been warned.
Now the official warning is out of the way with let’s get down to business. The business of making your gaming PC even more awesome, that is, starting with how to overclock your CPU.
How to overclock the CPU
1. Download CPU overclocking software
When it comes to overclocking the CPU, it’s a slightly different bag of chips to boosting the performance of your GPU. Though the basic principles remain the same when learning both how to overclock the CPU and the GPU: be patient and take it steady.
There’s one other thing to remember and that’s to pick your battles. Not every processor can be overclocked these days. You need to make sure your CPU has an unlocked multiplier to be able to get serious performance boosts. For Intel, that means picking a K-series chip, and for AMD any of its latest Ryzen CPUs or older Black Edition parts. You can hit the Windows and Pause/Break keys together on your keyboard to find out what processor you have in your rig, if you’re not sure.
Check out our guide to the best CPUs for gaming for our take on the top processors around right now. They just happen to also be the ones that are most comfortable with having the clocks tweaked, too.
You’ll also need some other software to help you out, but once more, it’s all available for free. First, you’ll need more monitoring software, and we’d suggest picking out CPU-Z to keep an eye on the real-time CPU clock speed and multipliers, and also Real Temp to keep track of the temperatures inside the processor package.
The simplest CPU benchmarking tool is Cinebench.
It’s a multi-threaded processor-based rendering test which will stress your silicon and spit out an index score at the end to help gauge relative performance. Finally, download Prime95 to stress test your CPU once you’re happy with the final clock speed.
AMD’s Ryzen CPUs even have their own Ryzen Master software specifically designed for use with the chips. Though you will have greater access to the different settings of your AM4 board going through the BIOS rather than the more basic software.
When you come to testing, it’s worth remembering that using the Ryzen Master application takes up around 10% of your CPU’s resources, so always shut it down if you’re benchmarking how well your overclock has taken.
But, personally, we always find it just as easy to do everything from within the old-school confines of your motherboard’s BIOS. And also, because it’s mostly keyboard-based, you can pretend you’re a hacker out of a mid-’90s TV show or clueless Hollywood blockbuster.
2. Update your BIOS and drivers
Once again, you need to make sure your PC is fully up to date. That doesn’t just mean Windows Update and your drivers – you also want to make sure your motherboard BIOS is the latest firmware. To do this, you’ll need to find out the exact make and model of your current motherboard, as well as what BIOS version you’re currently running. Handily, the CPU-Z app you’ve just downloaded will help if you don’t know off the top of your head.
Boot up CPU-Z and click the Mainboard tab across the top. The manufacturer and model of your board will be listed in the window that pops up, as well as what BIOS version is currently installed. Armed with this knowledge, you can head off to the manufacturer’s website, search for your board, and download the latest firmware BIOS update. We’ve got a much more in-depth guide on how to perfrom a BIOS update too!
It’s not worth messing around here – you need to find the EXACT match between BIOS and board.
Simply downloading one that looks about right won’t be enough. Chances are the BIOS will reject it. But, at worst, it could brick your motherboard completely. Beware. Once downloaded, unzip the BIOS update to a USB stick, and reboot with your pendrive in place.
Hitting either F2, F8, F12 or Del (why can’t all these manufacturers stick to a standard hotkey?) when the PC POST screen appears (before the Windows loading screen loads up) will take you into your BIOS screen. From there you’ll likely have to go into the Tools tab. What BIOS you see will differ on a manufacturer-to-manufacturer basis, but they’re all broadly similar. You should find a BIOS update utility that will take your downloaded firmware update and guide you through the process.
3. Load the default motherboard settings
Once your BIOS is updated you’ll want to reboot back into it from fresh and find the option to load the optimised settings. This is essentially a factory reset for your motherboard, but it’ll be tied to the latest firmware you’ve just installed, so it’s worth doing before you start anything.
Check the boot settings afterwards to make sure it hasn’t changed which of your storage drives the PC boots from.
You may also find your BIOS has a handy auto-overclocking function – many of the modern boards will come with such features. They can, however, be a bit hit and miss, but it’s worth giving it a go first to see what your motherboard can do on its own. If nothing else, it’ll give you a better platform to continue overclocking manually.
4. Change your clock multiplier
Your processor’s clock speed is worked out by multiplying the base clock (BCLK) by the CPU multiplier. The BCLK will likely be set at 100 MHz by default, so for a Core i5 6600K, for example, the multiplier will be set at 35 out of the box producing a final clock speed of 3.5 GHz.
Boosting your CPU’s clock speed, then, is simply a matter of stepping the processor’s clock speed multiplier up by one, testing whether the machine still boots, and then running a quick Cinebench test to check stability.
You will need to set the CPU multiplier/ratio setting to Manual/Sync All Cores and enter the desired number. You should then see a calculation of the final clock speed somewhere on the page. Then save the settings and restart your machine, booting into Windows.
Once on your Windows desktop, start up CPU-Z and Real Temp, then run the CPU test in Cinebench to make sure your processor is operating stably. Keep an eye on CPU-Z to check whether the chip is running at the expected speed and check the processor’s temperature in Real Temp to make sure it’s not running insanely hot.
Then, if everything works as it should, reboot into the BIOS and rinse and repeat until either the machine refuses to boot (why, hello Mrs. Bluescreen of Death, how nice to see you…) or it falls over partway through your stability benchmark.
When that happens, simply reboot into your BIOS again to change the settings. Chances are the POST screen will lock when you restart, with a message saying ‘overclocking has failed’, immediately giving you the option to jump into the BIOS screen.
From here, knock the multiplier down by one, save, and reboot testing for stability in Windows.
If you get into a horrible boot cycle where your machine freezes before you can get into your BIOS, don’t panic – all is not lost. You may need to delve into your machine and get a look at your motherboard, though. This is because you probably have to hard reset your motherboard’s BIOS. The method can differ from manufacturer to manufacturer – Google is your friend here. It will either be a little button on the back panel of your board (if you’re lucky), a button on the motherboard inside your case (if you’re not quite so lucky), or a jumper switch you need to short on the PCB itself (if you smashed a mirror over a black cat).
5. Check for stability
When you’ve reached the limits of what boosting the multiplier can get you then it’s time to make sure your machine is still 100% stable. Prime95 is a great stress-testing tool for CPUs – if your overclock can survive ten minutes running Prime95, chances are your new clock speed is pretty damn solid.
Boot up both CPU-Z (to check the clock speed stays constant) and Prime95, selecting the small FFTs when the torture test screen pops up.
Run the torture test for at least ten minutes, and then, when you’re happy with the chip’s stability, you’ll need to click the Test tab in the Prime95 window and stop the test manually. Just shutting the window down won’t necessarily do the job.
Now you should have a rock-solid CPU overclock, hopefully at least boosting the minimum frame rates in-game, and potentially freeing up a little extra performance from your graphics cards as a whole.
6. There must be more to it than that….
Of course there is. If you want to go even further with your expensive silicon then there’s a rabbit hole to jump down. Get up to your elbow in CPU overclocking and you can start messing around with the voltages and base clock settings, but that way lies greater long-term risk to your hardware, and a lot more intricate fiddling in various BIOS settings, too.
Besides, fiddling with voltages also increases the heat output and may only get you a little higher clock speed. You can also increase the BCLK of some processors but doing that is more prone to failure.
You can improve the cooling of your CPU and your PC as a whole, which may help net you a higher final clock speed. If you’re running the same stock cooler that came boxed up with your processor then your chip is likely to be getting more toasty than it needs. Picking up the best gaming motherboard to pair with a solid cooler can help too.
How to overclock your GPU
One-click overclocking
If you really want to take any of the tinkering (and most of the fun) out of overclocking your Nvidia GPU, you can instead opt for Nvidia’s Scanner functionality. This handy tool, found within various overclocking apps including EVGA Precision and MSI Afterburner, sets and validates your GPU overclock for you – leaving only memory clocks down to the user.
1. Download GPU overclocking software
If you want to get your feet wet, however, the first step is to make sure your PC is completely up to date.
That ought to either flag up any gremlins lurking in your system or help get rid of them before they become an issue. Check Windows Update to get your OS up to date and ensure you have the latest drivers for your graphics card, whether it’s an AMD or Nvidia GPU.
Next, grab some overclocking software. My personal choice for perking up your graphics card is MSI’s Afterburner software. It uses the classic Rivatuner backend (like Asus’s GPU Tweak and EVGA’s Precision apps), but comes with an easy-to-use interface and a handy on-screen display to keep track of things in-game. It’s a free download, takes only a few minutes to install, and can be used on any GPU, not just those from MSI.
As well as the software for enacting any overclock you want to achieve you should get yourself some benchmarking applications too. Have a look in your Steam library; you will find some of the games you already own have built-in benchmarks – everyone’s got a copy of GTA 5, right? But, while games are great for giving you a tangible before and after shot of your PC’s performance, they’re a bit awkward to use when you’re trying to find the limits of your hardware.
We suggest downloading the Unigine Heaven benchmark. It’s a GPU-intensive 3D rendering test that will stress your card with a continuous loop you can run in a window on your desktop.
2. Benchmark your performance
You need to get some baseline performance numbers for how your rig currently performs. Benchmarking is a rather dry, tedious process, but it’ll allow you to feel warm and fuzzy inside after you’ve nailed your extra GPU performance, because you can see in black and white what all that overclocking effort has afforded you.
Grab one of your favourite games with a built-in benchmark – we use the likes of GTA 5, the Total War games, Shadow of the Tomb Raider, Hitman, and Far Cry 5 – and then run the test (usually found in the graphics options) at your current settings, noting down the final average and minimum frame rate scores. It’s also worth benchmarking the Heaven test at your monitor’s native resolution too (hit F9 and sit back). You could also download the free version of 3DMark and run the basic FireStrike test to get a reliable index score.
3. Adjust the memory frequencies
We can’t stress this enough – be patient. That’s the key to getting the most out of your GPU with the basic overclocking we’re doing here. You’re very unlikely to do any damage to your graphics card while you’re testing its limits, but if you go in heavy-handed, pushing the clocks up as high as the sliders allow, the minimal risks are definitely increased.
First, boot up Afterburner and make sure the little Windows logo isn’t lit up – clicking that button ensures the current boosted settings are applied when the system boots up. You don’t want that happening when you’re trying to find the limits, only when you’re sure the system is stable.
Next, boot up Heaven in a window, one that’s small enough so you can view and access the Afterburner control panel while Heaven is looping through its floating medieval village scenery. On a native 1080p screen, run it at 1280 x 720, and then, with a 1440p or 4K panel, run the test at 1920 x 1080.
Now you’re ready to start tweaking and the best place to start is with the memory clocks. Start shifting the memory slider to the right in 5-10 MHz increments, hitting the tick button to apply the boost each time. After each step, check the looping Heaven run for visual artefacts appearing on-screen. Memory glitches will manifest themselves as either solid blocks, blobs of colour, or pin-sharp stars. Keep pushing the memory slider up in small increments until you start to see those telltale signs of failing memory… or until the card or system crashes.
Once you get to that stage, dial back the clock speed by either one or two steps, apply the new memory clock, and leave Heaven running for a longer period to make sure it remains stable at the new speed. If you see more memory artifacts, drop down another 5-10 MHz increment, and test again.
Now that you’ve found the memory limits, note down the stable clock speed offset, and reset the card to its default settings.
4. Overclock the GPU
This is where the real performance-changing work goes down and also where you’re going to start generating the most heat. Before you start upping the GPU clock speed, push the power limit slider up to max – allowing the graphics card to deliver more power to the silicon if it can – and also push up the temperature limit a little higher to account for the increased heat generation.
Using the same process we used when overclocking the video memory, push the GPU clock speed up by 5-10 MHz increments, checking the game window for artefacts as you apply each step change.
Processor artefacts manifest themselves in different ways to memory issues. Keep an eye out for pixel-sized dots of different colours appearing around the screen, random shards of colour, or bright and coloured full-screen flashes. As soon as you see anything like that it’s time to dial back the GPU clock speed in the same way you did with the memory to find the stable boosted clock speed.
5. Test both overclocks at once
Now you want to see how everything looks with both the GPU and memory clocks set to the peak levels you discovered individually. Don’t be surprised if, when putting both sliders up to where they operated stably earlier, you now get lots of artefacts or crashes. If that happens (and it probably will), take note of whether the glitches you see relate to either GPU or VRAM, and knock back the relevant slider by one 5 MHz notch.
If you get a system crash, on the other hand, simply reboot and set both memory and processor speeds back one step, and test again. When you feel confident about your settings it’s time to stress test.
6. Test and test again
Close down Heaven in windowed form and restart it fullscreen at your native resolution and leave it looping for a good ten minutes to make sure it remains fully stable at your new clocks. When you are positive it is stable, hit F9 to benchmark, and see what performance increase your efforts have afforded you.
Now replicate your earlier gaming and 3DMark tests to see what else you’ve gained.
Finally, hit that little Windows logo on Afterburner and light it up so that every time you boot you’re running at your newly discovered, overclocked settings.
7. Upgrading your GPU cooler
Not all graphics cards are created equal and neither are the chips inside them. You cannot be guaranteed a particular overclock (or indeed any overclock at all) as every GPU will have a different limit because of vagaries in the manufacturing process. We’ve also heard of issues with later GTX 1070 cards, using Micron instead of Samsung memory, that limits their memory overclocking performance compared with the original memory modules of the Founders Editions.
You might get lucky and find your GPU will boost like a beast. Similarly, you might not be able to get much extra performance at all. But as certain accidental horse-meat suppliers are wont to say: every little helps.
There are other things you can do to help, however.
If your card is a reference version, using either a standard Nvidia or AMD cooler design, you can buy relatively cheap third-party coolers to replace them and potentially deliver higher overclocking performance. Corsair even offers adapters so you can attach one of their closed-loop CPU water coolers to your graphics card, though that is probably going too far if you’re just trying to eke out another few months of gaming from your ageing GPU. Otherwise, just make sure your case has good airflow, with a healthy amount of intake and exhaust fans.
8. How to use AMD WattMan
Using Afterburner, or other Rivatuner-based applications, isn’t the only option. With AMD Radeon cards you can simply use the WattMan tool included with the Radeon Settings driver software.
It’s a dedicated option for modern AMD cards and allows you to do per-application overclocking for the best PC games. Anything before the 400-series cards you won’t have full access to everything that WattMan can offer, though you will still be able to play with some sliders.
The overclocking process is similar to the one we detailed earlier – patiently pushing up clock speeds while checking Heaven in a separate window – but there are a few, slightly different steps. First, you need to raise the minimum and target fan speeds, then switch from auto to manual and increase the temperature targets. Again, though, you’ll need to push the power limit to maximum before you tweak the clocks.
For the memory, you need to raise ‘State 1’ in small increments and hit apply to check stability while the GPU slider needs to be raised in 0.5% steps instead of the 5-10 MHz jumps we used in Afterburner.
Once more, it’s all about patience and vigilance.
MSI GeForce RTX 2070 Armor — Review 2018
MSI GeForce RTX 2070 Armor — Review 2018 — [SITE_NAME]
MSI’s GeForce RTX 2070 Armor graphics card has great cooling, plus overclocking headroom to spare. It’s solid for 1440p play, but it performs much like the outgoing GTX 1080, so step up to an RTX 2080 if you want a generational performance gain.
4.0 Excellent
-
Pros
- Quiet, effective cooling.
- Ample overclocking headroom.
- Excellent 1080p and 1440p gaming performance.
-
Cons
- Priced higher than GTX 1070 it replaces.
- Not powerful enough for maxed-out 4K gaming in every game.
The GeForce RTX 20-series launch in September 2018 was dominated by coverage of the GeForce RTX 2080 and flagship GeForce RTX 2080 Ti graphics processors (GPUs), the two most powerful in the lineup. The enthusiast-level RTX 2070, slated for a later release, largely went ignored—until now. The RTX 2070 card under review here is the MSI GeForce RTX 2070 Armor, a customized version of the reference board that Nvidia outlines for its partner companies. Its $499 list price is the least you can expect to spend for an RTX 2070-based graphics card, as that’s Nvidia’s recommended starting price for this GPU family.
Plenty of other RTX 2070 cards cost more; as a case in point, Nvidia sells its own factory-overclocked RTX 2070 Founders Edition for $599. But this MSI board indeed may be the dark horse of the RTX cards we have tested so far: It brings the idea of «value,» for the first time, into the RTX discussion.
GeForce RTX 2070 Essentials
The RTX 2070 is the intended successor of the circa-2016 GeForce GTX 1070. Our testing showed the RTX 2070 offered compelling performance improvements with an economic catch: It’s priced $100 higher than that outgoing card. The RTX 2070’s pricing instead lines up with the GeForce GTX 1080, a precarious situation for the newer card since the two cards showed equivalent performance in most of our tests. This isn’t a problem if you’re upgrading from much older hardware or are buying a graphics card for the first time, but it’s a tougher pill to swallow if you already have one of the better 10-series «Pascal»-based cards.
First off, I’ll highly recommend reading the first half of our GeForce RTX 2080 Founders Edition review for the complete scoop on the new Nvidia «Turing» architecture behind the RTX 20 series.
This article will focus on the differences between the RTX 2070 and the GTX 1070, as well as on comparisons to the pricier RTX 2080.
View All 9 Photos in Gallery
First, here are the comparable specifications between the RTX 2070 and the GTX 1070…
|
GeForce RTX 2070 Reference Card («Turing») |
GeForce GTX 1070 Founders Edition («Pascal») |
|
|
Nvidia CUDA Cores |
2,304 |
1,920 |
|
Base Clock |
1,410MHz |
1,506MHz |
|
Boost Clock |
1,620MHz |
1,683MHz |
|
Memory Speed |
14Gbps |
8Gbps |
|
Standard Memory Config |
8GB GDDR6 |
8GB GDDR5 |
|
Memory Interface Width |
256-bit |
256-bit |
|
Memory Bandwidth |
448GB/s |
256GB/s |
|
Graphics Card Power |
175 watts |
150 watts |
|
Number of Transistors |
10. |
7.2 billion |
|
Die Size |
445mm2 |
314mm2 |
|
Manufacturing Process |
12nm FinFET |
16nm |
|
Power Connectors |
One eight-pin |
One eight-pin |
|
Recommended Price |
$499 |
$399 |
8 billionThe 20 percent increase in the CUDA core count on the RTX 2070 hints at its rough performance advantage. However, the real-world gain from that specification alone is likely to be greater, seeing as, according to Nvidia, the CUDA cores on Turing cards boast up to a 50 percent per-core performance increase. Moreover, the far greater memory bandwidth from the RTX 2070’s high-tech GDDR6 memory should help when gaming at higher resolutions. The RTX 2070 has a mild increase in its power rating but can still be powered by a single eight-pin connector.
Its die size is also 42 percent larger, a good indicator that Nvidia’s yields from its 12nm FinFET process are good.
These are just the comparable specifications of the RTX 2070 and the GTX 1070. The RTX 2070 adds two new hardware specifications not before seen in a consumer-level graphics card: Tensor cores and ray tracing (RT) cores. The former power a key feature of the RTX 20 series, Deep Learning Super Sampling (DLSS). Briefly, this AI-driven technology is designed to increase the performance of edge smoothing in games while improving visual quality. Meanwhile, the RT cores enable ray tracing for enhanced visual effects. Ray tracing isn’t a new idea (Pixar has been using it for years in its 3D-rendered movies), but it’s so far gone unused in video games because of its substantial performance overhead. The RT cores are designed to hardware-accelerate ray tracing to make it usable in real-time.
Both DLSS and ray tracing have thus far gone untested in the real world. They don’t «just work» in all games; the game needs to be designed to use them.
Nvidia has a list of upcoming games with support, but they’re still awaiting release as I type this. Therefore, the value proposition of the GeForce RTX 20 series remains partially hedged in faith.
GeForce RTX 2070 Versus RTX 2080
Leaving Pascal behind, here are the RTX 2070 and the RTX 2080 reference cards side-by-side…
|
RTX 2070 Reference Card |
RTX 2080 Reference Card |
|
|
GPU Model |
Nvidia TU106 |
Nvidia TU104 |
|
Shading Multiprocessors (SMs) |
36 |
46 |
|
CUDA Cores |
2,304 |
2,944 |
|
Tensor Cores |
288 |
368 |
|
RT Cores |
36 |
46 |
|
Giga-Rays Per Second Rating |
6 |
8 |
|
RTX-OPS |
42 |
60 |
|
Base Clock |
1,410MHz |
1,515MHz |
|
Boost Clock |
1,620MHz |
1,710MHz |
|
Memory Speed |
14Gbps |
14Gbps |
|
Standard Memory Config |
8GB GDDR6 |
8GB GDDR6 |
|
Memory Interface Width |
256-bit |
256-bit |
|
Memory Bandwidth |
448GB/s |
448GB/s |
|
Graphics Card Power |
175 watts |
225 watts |
|
Number of Transistors |
10. |
13.6 billion |
|
Die Size |
445mm2 |
545mm2 |
|
Power Connectors |
One eight-pin |
One six-pin, one eight-pin |
|
Recommended Price |
$499 |
$699 |
8 billionThe number of Shading Multiprocessors (SMs) is the major differentiator between the cards. Each Turing SM contains 64 CUDA cores, eight Tensor cores, and one RT core processing engine. Look to the RTX-OPS number for organic performance comparisons between RTX 20-series cards, Nvidia’s published metric for measuring performance workloads. RTX-OPS is a combination of Tensor core, CUDA core, RT core, and INT32 operations in a single rendered frame.
The catch with RTX-OPS is that it’s a combination of metrics. If a game doesn’t utilize ray tracing, for example, then the performance no longer scales according to RTX-OPS.
Assuming a game does make use of all components of RTX-OPS, the RTX 2080 could have a 43 percent performance advantage over the RTX 2070. That corresponds with the roughly 40 percent price increase for the RTX 2080. As I write this, though, RTX 2080 cards are mostly parked around the $799 mark, not $699, whereas cards built around the GeForce RTX 2070 GPU are readily found at Nvidia’s recommended $499 price.
MSI RTX 2070 Armor Walk-Around
The MSI RTX 2070 Armor is bigger than a traditional two-slot graphics card in every dimension, measuring 12.1 inches long, 6.1 inches tall, and 2 inches thick. Compare that to MSI’s blower-style RTX 2070 Aero 8G, which is 10.5 inches long, 4.5 inches tall, and 1.6 inches thick.
The top view of the card is dominated by its twin-fan cooling design. The fans use ball bearings for quiet operation, which I found to be the case in my testing. Through the fans you can see the large aluminum-finned heatsink and five large copper heatpipes.
The shroud around the fans is plastic, but I wasn’t necessarily looking for metal there.
The brushed-metal backplate on the flip side of the card is a passive heatsink. Thermal pads on its underside transfer heat from the GDDR6 memory modules.
The MSI logo on the side of the card is backlit with RGB LEDs. MSI’s Mystic Light software allows you to change the colors and patterns, or simply turn it off. This is the only lighting zone on this card. It doesn’t do enough for the looks to make it a major selling point, in my view. For real flair and lighting gusto, the pricier MSI RTX 2070 Gaming graphics cards offer more comprehensive lighting.
Also, up here are the six-pin and eight-pin power connectors. Nvidia specifies the RTX 2070 reference card with one eight-pin connector. The supplementary six-pin connector on this card is there to provide extra power for overclocking. Absent on every RTX 2070 is an NVLink connector for SLI support. Only the GeForce RTX 2080 and RTX 2080 Ti offer SLI support within the RTX 20 series, and just two-card SLI at that.
The connectors on the backplate match those of the RTX 2080. They comprise three DisplayPort 1.4a video-out connectors, a single HDMI 2.0b video-out connector, and one VirtualLink (USB Type-C) connector for next-generation virtual-reality headsets. The latter have yet to come to market.
So, the big question: How did it perform? PC Labs ran numerous DirectX 11- and DirectX 12-based synthetic benchmarks and games on the RTX 2070 Armor. This will provide an illuminating comparison to current-generation graphics cards.
The spotlight will be on how the RTX 2070 compares to the GTX 1070 and GTX 1070 Ti, but equally so on how it stacks up to the similarly priced GTX 1080.
Our test rig is equipped with an Intel Core i7-8700K processor, 16GB of G.Skill DDR4 memory, a solid-state boot drive, and an Aorus Z370 Gaming 7 motherboard. Note for a few tests we used our backup test rig because of technical issues. It was equipped with an Intel Core i7-7700K processor, 16GB of Corsair RAM, an MSI Gaming Z270 M5 motherboard, and a solid-state boot drive.
Its performance for the benchmarks we run should be very close to those of the main rig.
The next section goes through each benchmark step by step. Skip down to the conclusion if you’re looking for the TLDR version.
Synthetic Benchmarks
3DMark Fire Strike Ultra
Synthetic benchmarks can be good predictors of real-world gaming performance. Futuremark’s circa-2013 Fire Strike Ultra is still a go-to for 4K-based gaming. We’re looking only at the graphics subscore, not the overall score.
The MSI RTX 2070 Armor held just a 20 percent lead over the GTX 1070 Founders Edition. This was the smallest difference I saw between the two across any benchmark we ran. The RTX 2080 Founders Edition wasn’t quite one-third faster, while the RTX 2070 maintained a roughly even keel with the GTX 1080 Founders Edition.
3DMark Time Spy and Time Spy Extreme
The Time Spy set is Futuremark’s DirectX 12-enabled benchmark for predicting the performance of DirectX 12-enabled games.
It uses major features of the API, including asynchronous compute, explicit multi-adapter, and multi-threading.
Conversely to 3DMark Fire Strike Ultra, the Time Spy and Time Spy Extreme benchmarks showed the largest advantages for the MSI RTX 2070 over the GTX 1070 Founders Edition across our benchmarks: 53 percent. The RTX 2070 also scored a notable 16 percent higher than the GTX 1080 Founders Edition. These results indicate the RTX 20 series is more proficient at making use of the newer DirectX 12 API than the GTX 10-series cards.
Unigine Superposition
Our last synthetic benchmark is Unigine’s 2017 release, Superposition. This benchmark does incorporate ray tracing, but it’s done in software, not hardware, and thus doesn’t utilize the RT cores of the RTX 20 series. (Note that we used the backup test rig for the RTX 2070 results in this test.)
The MSI RTX 2070 Armor did particularly well against the GTX 1070 Founders Edition, showing a 31 percent performance advantage at the 8K Optimized setting.
The performance advantage progressively decreased as the resolution increased, however; I saw the opposite trend with most of our tests.
Real-World Gaming
The following benchmarks are games that you can play. The charts themselves will list the settings we used (typically, the highest in-game presets and, if available, DirectX 12).
One game notably absent from our performance results is Shadow of the Tomb Raider. Despite using the latest available driver for the MSI RTX 2070 Armor, the game consistently crashed prior to the load screen on both of our test rigs. Strangely enough, we didn’t experience this behavior with the RTX 2080-based cards we reviewed.
Rise of the Tomb Raider
The 2015 predecessor to Shadow of the Tomb Raider is still a great benchmark.
The MSI RTX 2070 Armor’s lead over the GTX 1070 Founders Edition gradually increased with screen resolution, topping 36 percent at 4K. It’s likely the fast GDDR6 memory on the RTX 2070 is partially responsible for that pattern.
Meanwhile, the MSI RTX 2070 maintained pace with the GTX 1080 Founders Edition, and it was left behind by the RTX 2080 Founders Edition by 18 to 26 percent.
Although the MSI RTX 2070 produces a playable average frame rate in this game’s benchmark at a 4K resolution, I found it could be a choppy experience at times when I played the game myself. The RTX 2070 isn’t quite powerful enough to make the game consistently playable at 4K at Very High settings. I dialed back some of the in-game shadow and shading experience to smooth it out. For a no-compromise 4K gaming experience, consider spending the extra for an RTX 2080. It’s not as if games are going to get less demanding in the future.
Far Cry 5 and Far Cry Primal
The fourth and fifth installments in the Far Cry series are based on DirectX 11, but still demanding. We’re looping the benchmark charts together since they benchmark similarly. (Note we used the backup test rig for the RTX 2070 results in Far Cry Primal.)
The 1080p screen resolution results are likely CPU-limited.
The MSI RTX 2070 Armor leads the GTX 1070 Founders Edition in Far Cry 5 by 41 percent at 4K, and it also beat out the GTX 1080 Founders Edition by a few percentage points. As I found with Rise of the Tomb Raider, Far Cry 5 was also a somewhat choppy experience at 4K with the Ultra visual-quality settings. Dialing back a few of the settings is necessary if you want to achieve an average closer to 60 frames per second (fps).
Final Fantasy XV
We’ll take a respite from fps-based benchmarks for Final Fantasy XV. (Note we used the backup test rig for the RTX 2070 results in this test, too.)
The MSI RTX 2070 Armor saw some of its best gains over the GTX 1070 Founders Edition in this points-based benchmark. In turn, the RTX 2080 Founders Edition showed some of its most minimal gains over the MSI RTX 2070 Armor.
World of Tanks Encore
This is another non-fps-based benchmark that’s available as a free download. It’s not super-demanding, but it is still a reliable test reflecting massively multiplayer battle games.
Chalk up another win for the MSI RTX 2070 Armor; it held a 29 percent advantage at 4K over the GTX 1070 Founders Edition. The RTX 2080 Founders Edition was 26 percent ahead of the RTX 2070 at 4K, a result that’s in line with previous benchmarks, and that correlates with its 28 percent increase in CUDA core count over the RTX 2070.
Tom Clancy: The Division
A 2016 release that remains tough to handle, here’s our final DirectX 12-specific game test.
The MSI RTX 2070 Armor’s performance relative to the GTX 1070 Founders Edition jumps from 25 percent at 1440p to 40 percent at 4K. This is another indication that the higher memory bandwidth of the RTX 2070 is beneficial at higher resolutions.
The 4K numbers from the RTX 2070, however, are too low for consistent in-game playability. You’ll need to vary a few of the in-game settings to get a smoother experience. Alternatively, compromise on a 1440p resolution, at which the RTX 2070 produces buttery-smooth frame rates even with the visual quality settings cranked.
Overclocking Performance
I used MSI’s Afterburner software to overclock the MSI RTX 2070 Armor. The software’s built-in Nvidia Scanner API tests the GPU for its ideal voltage/frequency curve. The several scans I ran were consistent, and I ended up with this…
The base clock went from 1,410MHz to 1,574MHz, and the boost clock went from 1,620MHz to 1,784MHz, an increase of 164MHz on both. Those are heftier relative gains than I achieved with RTX 2080-based cards. The results I received aren’t guaranteed; every GPU will be different, so your mileage may vary.
Contrast the overclocked numbers to those of the MSI RTX 2070 Armor 8G OC card, which has an overclocked 1,740MHz boost clock from the factory. It goes for $549 at retail, or $50 more than the non-overclocked RTX 2070 Armor card I’m testing. That premium hardly seems worthwhile given the card is otherwise identical, and I surpassed its factory overclock with a few mouse clicks.
The overclocking made a substantial difference in benchmarks and games.
Note the numbers in the following chart aren’t all comparable to the numbers in the benchmarks section since I used the backup test rig based on the Intel Core i7-7700K.
|
MSI RTX 2070 Armor (Stock) |
MSI RTX 2070 Armor (OC via Nvidia Scanner) |
Overclock vs Stock |
|
|
3DMark Fire Strike Ultra (Graphics Score) |
5,095 |
5,562 |
+9% |
|
3DMark Time Spy (Graphics Score) |
8,471 |
9,211 |
+9% |
|
Rise of the Tomb Raider (Very High, DX12, 1,920 by 1,080) |
136fps |
145fps |
+7% |
|
Rise of the Tomb Raider (Very High, DX12, 2,560 by 1,440) |
93fps |
99fps |
+6% |
|
Rise of the Tomb Raider (Very High, DX12, 3,840 by 2,160) |
49fps |
52fps |
+6% |
|
Far Cry 5 (Ultra, 1,920 by 1,080) |
113fps |
117fps |
+4% |
|
Far Cry 5 (Ultra, 2,560 by 1,440) |
87fps |
92fps |
+6% |
|
Far Cry 5 (Ultra, 3,840 by 2,160) |
46fps |
49fps |
+7% |
|
AVERAGE |
+7% |
The synthetic 3DMark benchmarks showed the maximum gains from the overclocking: 9 percent.
The built-in benchmarks in Rise of the Tomb Raider and Far Cry 5 showed 6 to 7 percent gains at 4K. Those are respectable results for automatic overclocking. A couple of extra fps at 4K makes an appreciable difference when you’re averaging under 60fps.
Thermals and Cooling
I’ll be analyzing how the MSI RTX 2070 Armor controls its GPU core temperature, memory clock, and boost clock while under stress. I used 20 loops of the 3DMark Fire Strike Ultra stability test to accomplish this, which takes about 10 minutes to run.
Here’s the card performing at its factory clocks. I excluded the memory clock, as it maintained its 1,750MHz rating throughout the test…
The core clock averaged 1,722MHz, or considerably higher than its rated 1,620MHz boost clock. The GPU increases the boost clock automatically if power and thermal conditions permit. There was some fluctuation in the core clock, which is normal, but it stayed in a tight range; the standard deviation was 15MHz.
The most important takeaway is that it stayed consistent over time.
The average GPU core temperature of 62 degrees C suggests this card has plenty of thermal headroom. I did the testing in a cool room (65 degrees F), but I didn’t hear the fans in the RTX 2070 Armor spinning without putting my ear right next to the card. I have no reason to believe this card would have thermal problems in a warmer environment.
Now let’s see how the RTX 2070 Armor did in its overclocked state. I again left out the memory clock, which continued to maintain its factory frequency (it wasn’t overclocked).
The overclock I achieved with MSI Afterburner officially set the boost clock to 1,784MHz, but here it ran at a much higher average of 1,921MHz. The core clock still stayed in a controlled range, with the standard deviation increasing by less than 1MHz. There was no change in the average GPU temperature of 62 degrees C. The fans were likely spinning faster, as the card must have been consuming more power, but I didn’t hear them.
There’s no doubt that the RTX 2070 Armor has plenty of thermal room for further overclocking.
Super-Solid Gamer, But a Worthy Upgrade?
The MSI GeForce RTX 2070 Armor put up a fine fight in our testing. Its cooling system was both quiet and effective, and it showed plenty of overclocking headroom. As a matter of fact, given the relative ease of overclocking, we’d think twice before spending extra for a factory-overclocked RTX 2070 card. It’s also priced dead-on at Nvidia’s recommended $499 price, the minimum you can expect to fork out for an RTX 2070.
The RTX 2070 is well-suited for high-refresh 1080p and 1440p gaming. It’s certainly capable of 4K gaming, although we had to lower the in-game settings in some games to get an average closer to 60fps. We’d recommend making the jump to an RTX 2080 if 4K is your aim.
The value proposition for the RTX 2070 depends on what card you’re coming from. The RTX 2070 is a sensible purchase if you’re upgrading from an outdated platform, as it wouldn’t make sense to buy the older but similar-performing GeForce GTX 1080 at the same $499 price.
That reasoning also applies if you’re buying a serious graphics card for the first time. Moreover, the RTX 2070 can be viewed as a sensible upgrade from the GTX 1070 if you want to move up to the next performance tier: It was 35 percent faster, on average, across our testing.
The pricing for the RTX 2070, though, is what turns Nvidia’s home run into a two-bagger or a triple. The generational per-dollar performance increase that Nvidia spoiled us with in the Pascal cards and that we have thus come to expect just isn’t there this time. It would have been, had the RTX 2070 been priced to match the GTX 1070, but the RTX 2070 is $100 more. Nonetheless, it’s still the best gaming GPU you can buy for $499. Its yet-untested DLSS and ray tracing features can only add to its appeal. We’re greatly looking forward to testing them out, just as we are the opportunity to welcome AMD’s next-generation offerings into the ring.
MSI GeForce RTX 2070 Armor
excellent
The RTX 2070 is the intended successor of the circa-2016 GeForce GTX 1070.
Our testing showed the RTX 2070 offered compelling performance improvements with an economic catch: It’s priced $100 higher than that outgoing card. The RTX 2070’s pricing instead lines up with the GeForce GTX 1080, a precarious situation for the newer card since the two cards showed equivalent performance in most of our tests. This isn’t a problem if you’re upgrading from much older hardware or are buying a graphics card for the first time, but it’s a tougher pill to swallow if you already have one of the better 10-series \»Pascal\»-based cards.
\n
First off, I’ll highly recommend reading the first half of our GeForce RTX 2080 Founders Edition review for the complete scoop on the new Nvidia \»Turing\» architecture behind the RTX 20 series. This article will focus on the differences between the RTX 2070 and the GTX 1070, as well as on comparisons to the pricier RTX 2080.
\n\n
\nView All 9 Photos in Gallery\n\n
\n\n
First, here are the comparable specifications between the RTX 2070 and the GTX 1070.
..
\n\n
| \n\n | \n
GeForce RTX 2070 Reference Card (\»Turing\») \n |
\n
GeForce GTX 1070 Founders Edition (\»Pascal\») \n |
| \n
Nvidia CUDA Cores \n |
\n
2,304 \n |
\n
1,920 \n |
| \n
Base Clock \n |
\n
1,410MHz \n |
\n
1,506MHz \n |
| \n
Boost Clock \n |
\n
1,620MHz \n |
\n
1,683MHz \n |
| \n
Memory Speed \n |
\n
14Gbps \n |
\n
8Gbps \n |
| \n
Standard Memory Config \n |
\n
8GB GDDR6 \n |
\n
8GB GDDR5 \n |
| \n
Memory Interface Width \n |
\n
256-bit \n |
\n
256-bit \n |
| \n
Memory Bandwidth \n |
\n
448GB/s \n |
\n
256GB/s \n |
| \n
Graphics Card Power \n |
\n
175 watts \n |
\n
150 watts \n |
| \n
Number of Transistors \n |
\n
10. \n |
\n
7.2 billion \n |
| \n
Die Size \n |
\n
445mm2 \n |
\n
314mm2 \n |
| \n
Manufacturing Process \n |
\n
12nm FinFET \n |
\n
16nm \n |
| \n
Power Connectors \n |
\n
One eight-pin \n |
\n
One eight-pin \n |
| \n
Recommended Price \n |
\n
$499 \n |
\n
$399 \n |
8 billion\n\n
The 20 percent increase in the CUDA core count on the RTX 2070 hints at its rough performance advantage. However, the real-world gain from that specification alone is likely to be greater, seeing as, according to Nvidia, the CUDA cores on Turing cards boast up to a 50 percent per-core performance increase.
Moreover, the far greater memory bandwidth from the RTX 2070’s high-tech GDDR6 memory should help when gaming at higher resolutions. The RTX 2070 has a mild increase in its power rating but can still be powered by a single eight-pin connector. Its die size is also 42 percent larger, a good indicator that Nvidia’s yields from its 12nm FinFET process are good.
\n
These are just the comparable specifications of the RTX 2070 and the GTX 1070. The RTX 2070 adds two new hardware specifications not before seen in a consumer-level graphics card: Tensor cores and ray tracing (RT) cores. The former power a key feature of the RTX 20 series, Deep Learning Super Sampling (DLSS). Briefly, this AI-driven technology is designed to increase the performance of edge smoothing in games while improving visual quality. Meanwhile, the RT cores enable ray tracing for enhanced visual effects. Ray tracing isn’t a new idea (Pixar has been using it for years in its 3D-rendered movies), but it’s so far gone unused in video games because of its substantial performance overhead.
The RT cores are designed to hardware-accelerate ray tracing to make it usable in real-time.
\n
Both DLSS and ray tracing have thus far gone untested in the real world. They don’t \»just work\» in all games; the game needs to be designed to use them. Nvidia has a list of upcoming games with support, but they’re still awaiting release as I type this. Therefore, the value proposition of the GeForce RTX 20 series remains partially hedged in faith.
\n
\n\n
GeForce RTX 2070 Versus RTX 2080
\n
Leaving Pascal behind, here are the RTX 2070 and the RTX 2080 reference cards side-by-side…
\n\n
| \n\n | \n
RTX 2070 Reference Card \n |
\n
RTX 2080 Reference Card \n |
| \n
GPU Model \n |
\n
Nvidia TU106 \n |
\n
Nvidia TU104 \n |
| \n
Shading Multiprocessors (SMs) \n |
\n
36 \n |
\n
46 \n |
| \n
CUDA Cores \n |
\n
2,304 \n |
\n
2,944 \n |
| \n
Tensor Cores \n |
\n
288 \n |
\n
368 \n |
| \n
RT Cores \n |
\n
36 \n |
\n
46 \n |
| \n
Giga-Rays Per Second Rating \n |
\n
6 \n |
\n
8 \n |
| \n
RTX-OPS \n |
\n
42 \n |
\n
60 \n |
| \n
Base Clock \n |
\n
1,410MHz \n |
\n
1,515MHz \n |
| \n
Boost Clock \n |
\n
1,620MHz \n |
\n
1,710MHz \n |
| \n
Memory Speed \n |
\n
14Gbps \n |
\n
14Gbps \n |
| \n
Standard Memory Config \n |
\n
8GB GDDR6 \n |
\n
8GB GDDR6 \n |
| \n
Memory Interface Width \n |
\n
256-bit \n |
\n
256-bit \n |
| \n
Memory Bandwidth \n |
\n
448GB/s \n |
\n
448GB/s \n |
| \n
Graphics Card Power \n |
\n
175 watts \n |
\n
225 watts \n |
| \n
Number of Transistors \n |
\n
10. \n |
\n
13.6 billion \n |
| \n
Die Size \n |
\n
445mm2 \n |
\n
545mm2 \n |
| \n
Power Connectors \n |
\n
One eight-pin \n |
\n
One six-pin, one eight-pin \n |
| \n
Recommended Price \n |
\n
$499 \n |
\n
$699 \n |
8 billion\n\n
The number of Shading Multiprocessors (SMs) is the major differentiator between the cards. Each Turing SM contains 64 CUDA cores, eight Tensor cores, and one RT core processing engine. Look to the RTX-OPS number for organic performance comparisons between RTX 20-series cards, Nvidia’s published metric for measuring performance workloads. RTX-OPS is a combination of Tensor core, CUDA core, RT core, and INT32 operations in a single rendered frame.
\n
The catch with RTX-OPS is that it’s a combination of metrics. If a game doesn’t utilize ray tracing, for example, then the performance no longer scales according to RTX-OPS. Assuming a game does make use of all components of RTX-OPS, the RTX 2080 could have a 43 percent performance advantage over the RTX 2070. That corresponds with the roughly 40 percent price increase for the RTX 2080. As I write this, though, RTX 2080 cards are mostly parked around the $799 mark, not $699, whereas cards built around the GeForce RTX 2070 GPU are readily found at Nvidia’s recommended $499 price.
\n
\n\n
MSI RTX 2070 Armor Walk-Around
\n
The MSI RTX 2070 Armor is bigger than a traditional two-slot graphics card in every dimension, measuring 12.1 inches long, 6.1 inches tall, and 2 inches thick. Compare that to MSI’s blower-style RTX 2070 Aero 8G, which is 10.5 inches long, 4.5 inches tall, and 1.6 inches thick.
\n\n
The top view of the card is dominated by its twin-fan cooling design. The fans use ball bearings for quiet operation, which I found to be the case in my testing. Through the fans you can see the large aluminum-finned heatsink and five large copper heatpipes. The shroud around the fans is plastic, but I wasn’t necessarily looking for metal there.
\n\n
The brushed-metal backplate on the flip side of the card is a passive heatsink. Thermal pads on its underside transfer heat from the GDDR6 memory modules.
\n\n
The MSI logo on the side of the card is backlit with RGB LEDs. MSI’s Mystic Light software allows you to change the colors and patterns, or simply turn it off. This is the only lighting zone on this card. It doesn’t do enough for the looks to make it a major selling point, in my view. For real flair and lighting gusto, the pricier MSI RTX 2070 Gaming graphics cards offer more comprehensive lighting.
\n\n
Also, up here are the six-pin and eight-pin power connectors. Nvidia specifies the RTX 2070 reference card with one eight-pin connector. The supplementary six-pin connector on this card is there to provide extra power for overclocking. Absent on every RTX 2070 is an NVLink connector for SLI support. Only the GeForce RTX 2080 and RTX 2080 Ti offer SLI support within the RTX 20 series, and just two-card SLI at that.
\n\n
The connectors on the backplate match those of the RTX 2080. They comprise three DisplayPort 1.4a video-out connectors, a single HDMI 2.0b video-out connector, and one VirtualLink (USB Type-C) connector for next-generation virtual-reality headsets. The latter have yet to come to market.
\n
So, the big question: How did it perform? PC Labs ran numerous DirectX 11- and DirectX 12-based synthetic benchmarks and games on the RTX 2070 Armor. This will provide an illuminating comparison to current-generation graphics cards.
\n
The spotlight will be on how the RTX 2070 compares to the GTX 1070 and GTX 1070 Ti, but equally so on how it stacks up to the similarly priced GTX 1080.
\n\n
Our test rig is equipped with an Intel Core i7-8700K processor, 16GB of G. Skill DDR4 memory, a solid-state boot drive, and an Aorus Z370 Gaming 7 motherboard. Note for a few tests we used our backup test rig because of technical issues. It was equipped with an Intel Core i7-7700K processor, 16GB of Corsair RAM, an MSI Gaming Z270 M5 motherboard, and a solid-state boot drive. Its performance for the benchmarks we run should be very close to those of the main rig.
\n
The next section goes through each benchmark step by step. Skip down to the conclusion if you’re looking for the TLDR version.
\n
\n\n
Synthetic Benchmarks
\n
3DMark Fire Strike Ultra
\n
Synthetic benchmarks can be good predictors of real-world gaming performance. Futuremark’s circa-2013 Fire Strike Ultra is still a go-to for 4K-based gaming. We’re looking only at the graphics subscore, not the overall score.
\n\n
The MSI RTX 2070 Armor held just a 20 percent lead over the GTX 1070 Founders Edition. This was the smallest difference I saw between the two across any benchmark we ran. The RTX 2080 Founders Edition wasn’t quite one-third faster, while the RTX 2070 maintained a roughly even keel with the GTX 1080 Founders Edition.
\n
3DMark Time Spy and Time Spy Extreme
\n
The Time Spy set is Futuremark’s DirectX 12-enabled benchmark for predicting the performance of DirectX 12-enabled games. It uses major features of the API, including asynchronous compute, explicit multi-adapter, and multi-threading.
\n\n
Conversely to 3DMark Fire Strike Ultra, the Time Spy and Time Spy Extreme benchmarks showed the largest advantages for the MSI RTX 2070 over the GTX 1070 Founders Edition across our benchmarks: 53 percent. The RTX 2070 also scored a notable 16 percent higher than the GTX 1080 Founders Edition. These results indicate the RTX 20 series is more proficient at making use of the newer DirectX 12 API than the GTX 10-series cards.
\n
Unigine Superposition
\n
Our last synthetic benchmark is Unigine’s 2017 release, Superposition. This benchmark does incorporate ray tracing, but it’s done in software, not hardware, and thus doesn’t utilize the RT cores of the RTX 20 series. (Note that we used the backup test rig for the RTX 2070 results in this test.)
\n\n
The MSI RTX 2070 Armor did particularly well against the GTX 1070 Founders Edition, showing a 31 percent performance advantage at the 8K Optimized setting. The performance advantage progressively decreased as the resolution increased, however; I saw the opposite trend with most of our tests.
\n
\n\n
Real-World Gaming
\n
The following benchmarks are games that you can play. The charts themselves will list the settings we used (typically, the highest in-game presets and, if available, DirectX 12).
\n
One game notably absent from our performance results is Shadow of the Tomb Raider. Despite using the latest available driver for the MSI RTX 2070 Armor, the game consistently crashed prior to the load screen on both of our test rigs. Strangely enough, we didn’t experience this behavior with the RTX 2080-based cards we reviewed.
\n
Rise of the Tomb Raider
\n
The 2015 predecessor to Shadow of the Tomb Raider is still a great benchmark.
\n\n
The MSI RTX 2070 Armor’s lead over the GTX 1070 Founders Edition gradually increased with screen resolution, topping 36 percent at 4K. It’s likely the fast GDDR6 memory on the RTX 2070 is partially responsible for that pattern. Meanwhile, the MSI RTX 2070 maintained pace with the GTX 1080 Founders Edition, and it was left behind by the RTX 2080 Founders Edition by 18 to 26 percent.
\n
Although the MSI RTX 2070 produces a playable average frame rate in this game’s benchmark at a 4K resolution, I found it could be a choppy experience at times when I played the game myself. The RTX 2070 isn’t quite powerful enough to make the game consistently playable at 4K at Very High settings. I dialed back some of the in-game shadow and shading experience to smooth it out. For a no-compromise 4K gaming experience, consider spending the extra for an RTX 2080. It’s not as if games are going to get less demanding in the future.
\n
Far Cry 5 and Far Cry Primal
\n
The fourth and fifth installments in the Far Cry series are based on DirectX 11, but still demanding. We’re looping the benchmark charts together since they benchmark similarly. (Note we used the backup test rig for the RTX 2070 results in Far Cry Primal.)
\n\n\n
The 1080p screen resolution results are likely CPU-limited. The MSI RTX 2070 Armor leads the GTX 1070 Founders Edition in Far Cry 5 by 41 percent at 4K, and it also beat out the GTX 1080 Founders Edition by a few percentage points. As I found with Rise of the Tomb Raider, Far Cry 5 was also a somewhat choppy experience at 4K with the Ultra visual-quality settings. Dialing back a few of the settings is necessary if you want to achieve an average closer to 60 frames per second (fps).
\n
Final Fantasy XV
\n
We’ll take a respite from fps-based benchmarks for Final Fantasy XV. (Note we used the backup test rig for the RTX 2070 results in this test, too.)
\n\n
The MSI RTX 2070 Armor saw some of its best gains over the GTX 1070 Founders Edition in this points-based benchmark. In turn, the RTX 2080 Founders Edition showed some of its most minimal gains over the MSI RTX 2070 Armor.
\n
World of Tanks Encore
\n
This is another non-fps-based benchmark that’s available as a free download. It’s not super-demanding, but it is still a reliable test reflecting massively multiplayer battle games.
\n\n
Chalk up another win for the MSI RTX 2070 Armor; it held a 29 percent advantage at 4K over the GTX 1070 Founders Edition. The RTX 2080 Founders Edition was 26 percent ahead of the RTX 2070 at 4K, a result that’s in line with previous benchmarks, and that correlates with its 28 percent increase in CUDA core count over the RTX 2070.
\n
Tom Clancy: The Division
\n
A 2016 release that remains tough to handle, here’s our final DirectX 12-specific game test.
\n\n
The MSI RTX 2070 Armor’s performance relative to the GTX 1070 Founders Edition jumps from 25 percent at 1440p to 40 percent at 4K. This is another indication that the higher memory bandwidth of the RTX 2070 is beneficial at higher resolutions.
\n
The 4K numbers from the RTX 2070, however, are too low for consistent in-game playability. You’ll need to vary a few of the in-game settings to get a smoother experience. Alternatively, compromise on a 1440p resolution, at which the RTX 2070 produces buttery-smooth frame rates even with the visual quality settings cranked.
\n
\n\n
Overclocking Performance
\n
I used MSI’s Afterburner software to overclock the MSI RTX 2070 Armor. The software’s built-in Nvidia Scanner API tests the GPU for its ideal voltage/frequency curve. The several scans I ran were consistent, and I ended up with this…
\n\n
The base clock went from 1,410MHz to 1,574MHz, and the boost clock went from 1,620MHz to 1,784MHz, an increase of 164MHz on both. Those are heftier relative gains than I achieved with RTX 2080-based cards. The results I received aren’t guaranteed; every GPU will be different, so your mileage may vary.
\n
Contrast the overclocked numbers to those of the MSI RTX 2070 Armor 8G OC card, which has an overclocked 1,740MHz boost clock from the factory. It goes for $549 at retail, or $50 more than the non-overclocked RTX 2070 Armor card I’m testing. That premium hardly seems worthwhile given the card is otherwise identical, and I surpassed its factory overclock with a few mouse clicks.
\n
The overclocking made a substantial difference in benchmarks and games. Note the numbers in the following chart aren’t all comparable to the numbers in the benchmarks section since I used the backup test rig based on the Intel Core i7-7700K.
\n\n
| \n\n | \n
MSI RTX 2070 Armor (Stock) \n |
\n
MSI RTX 2070 Armor (OC via Nvidia Scanner) \n |
\n
Overclock vs Stock \n |
| \n
3DMark Fire Strike Ultra (Graphics Score) \n |
\n
5,095 \n |
\n
5,562 \n |
\n
+9% \n |
| \n
3DMark Time Spy (Graphics Score) \n |
\n
8,471 \n |
\n
9,211 \n |
\n
+9% \n |
| \n
Rise of the Tomb Raider (Very High, DX12, 1,920 by 1,080) \n |
\n
136fps \n |
\n
145fps \n |
\n
+7% \n |
| \n
Rise of the Tomb Raider (Very High, DX12, 2,560 by 1,440) \n |
\n
93fps \n |
\n
99fps \n |
\n
+6% \n |
| \n
Rise of the Tomb Raider (Very High, DX12, 3,840 by 2,160) \n |
\n
49fps \n |
\n
52fps \n |
\n
+6% \n |
| \n
Far Cry 5 (Ultra, 1,920 by 1,080) \n |
\n
113fps \n |
\n
117fps \n |
\n
+4% \n |
| \n
Far Cry 5 (Ultra, 2,560 by 1,440) \n |
\n
87fps \n |
\n
92fps \n |
\n
+6% \n |
| \n
Far Cry 5 (Ultra, 3,840 by 2,160) \n |
\n
46fps \n |
\n
49fps \n |
\n
+7% \n |
| \n
AVERAGE \n |
\n\n | \n\n | \n
+7% \n |
\n\n
The synthetic 3DMark benchmarks showed the maximum gains from the overclocking: 9 percent. The built-in benchmarks in Rise of the Tomb Raider and Far Cry 5 showed 6 to 7 percent gains at 4K. Those are respectable results for automatic overclocking. A couple of extra fps at 4K makes an appreciable difference when you’re averaging under 60fps.
\n
\n\n
Thermals and Cooling
\n
I’ll be analyzing how the MSI RTX 2070 Armor controls its GPU core temperature, memory clock, and boost clock while under stress. I used 20 loops of the 3DMark Fire Strike Ultra stability test to accomplish this, which takes about 10 minutes to run.
\n
Here’s the card performing at its factory clocks. I excluded the memory clock, as it maintained its 1,750MHz rating throughout the test…
\n\n
The core clock averaged 1,722MHz, or considerably higher than its rated 1,620MHz boost clock. The GPU increases the boost clock automatically if power and thermal conditions permit. There was some fluctuation in the core clock, which is normal, but it stayed in a tight range; the standard deviation was 15MHz. The most important takeaway is that it stayed consistent over time.
\n
The average GPU core temperature of 62 degrees C suggests this card has plenty of thermal headroom. I did the testing in a cool room (65 degrees F), but I didn’t hear the fans in the RTX 2070 Armor spinning without putting my ear right next to the card. I have no reason to believe this card would have thermal problems in a warmer environment.
\n
Now let’s see how the RTX 2070 Armor did in its overclocked state. I again left out the memory clock, which continued to maintain its factory frequency (it wasn’t overclocked).
\n\n
The overclock I achieved with MSI Afterburner officially set the boost clock to 1,784MHz, but here it ran at a much higher average of 1,921MHz. The core clock still stayed in a controlled range, with the standard deviation increasing by less than 1MHz. There was no change in the average GPU temperature of 62 degrees C. The fans were likely spinning faster, as the card must have been consuming more power, but I didn’t hear them. There’s no doubt that the RTX 2070 Armor has plenty of thermal room for further overclocking.
\n
\n\n
Super-Solid Gamer, But a Worthy Upgrade?
\n
The MSI GeForce RTX 2070 Armor put up a fine fight in our testing. Its cooling system was both quiet and effective, and it showed plenty of overclocking headroom. As a matter of fact, given the relative ease of overclocking, we’d think twice before spending extra for a factory-overclocked RTX 2070 card. It’s also priced dead-on at Nvidia’s recommended $499 price, the minimum you can expect to fork out for an RTX 2070.
\n\n
The RTX 2070 is well-suited for high-refresh 1080p and 1440p gaming. It’s certainly capable of 4K gaming, although we had to lower the in-game settings in some games to get an average closer to 60fps. We’d recommend making the jump to an RTX 2080 if 4K is your aim.
\n
The value proposition for the RTX 2070 depends on what card you’re coming from. The RTX 2070 is a sensible purchase if you’re upgrading from an outdated platform, as it wouldn’t make sense to buy the older but similar-performing GeForce GTX 1080 at the same $499 price. That reasoning also applies if you’re buying a serious graphics card for the first time. Moreover, the RTX 2070 can be viewed as a sensible upgrade from the GTX 1070 if you want to move up to the next performance tier: It was 35 percent faster, on average, across our testing.
\n
The pricing for the RTX 2070, though, is what turns Nvidia’s home run into a two-bagger or a triple. The generational per-dollar performance increase that Nvidia spoiled us with in the Pascal cards and that we have thus come to expect just isn’t there this time. It would have been, had the RTX 2070 been priced to match the GTX 1070, but the RTX 2070 is $100 more. Nonetheless, it’s still the best gaming GPU you can buy for $499. Its yet-untested DLSS and ray tracing features can only add to its appeal. We’re greatly looking forward to testing them out, just as we are the opportunity to welcome AMD’s next-generation offerings into the ring.
\n\n
\n
MSI GeForce RTX 2070 Armor
\n
\n\n
excellent
\n
\n
\n
\n
\n\n
\n
\n\n\n\n
\n
\n
\n
«}}
The RTX 2070 SUPER vs.
the GTX 1080 Ti Overclocking Showdown
The RTX 2070 SUPER vs. the GTX 1080 Ti Overclocking Showdown with 40 Games
This overclocking showdown is a follow-up to the RTX 2070 SUPER Founders Edition (FE) review where we are now focusing on its overclocked performance versus the overclocked GTX 1080 Ti Founders Edition. Today, we have optimized and maxed-out our RTX 2070 SUPER and GTX 1080 Ti overclocks with all performance options set to their highest limits to get the maximum performance from each card.
The RTX 2070 SUPER Overclocking
In our original review of the RTX 2070 SUPER last week, we did not have time to overclock it. At stock it is able to peak around 1975MHz but drops down to about 1900MHz under gaming load after warm-up.
After much testing, we settled on a +140MHz offset to the core which allows it to reach 2025MHz when cold, but it averages above 2000MHz under full load after warm-up in gaming. Temperatures never hit 80C with the stock fan profile even when fully overclocked but they average just above 75C.
The RTX 2070 SUPER is quiet with the stock fan profile at stock settings, but it gets a bit more noticeable when it is fully overclocked. Our memory overclock is good and we achieved a +900MHz offset to the GDDR6. We found that +950MHz was also possible, but its performance stopped increasing over 900MHz; and at +1000MHz it began artifacting.
The GTX 1080 Ti Founders Edition Overclocking
We overclocked the GTX 1080 Ti in our original OC review versus the overclocked Titan Xp. “We were able to add 150 MHz to the core and a 500 MHz offset to the memory … with the result that we reached a steady 1999 MHz core and 6000 MHz (12000 MHz) memory clocks.”
Time and electromigration have not been particularly kind to our GTX 1080 Ti Founders Edition. It has been our gaming card for well over two years and it can no longer achieve the same overclock it once held. Our current stable overclock allows for +130 MHz offset to the core and +350 MHz offset to be added to the memory.
This evaluation should give a rough idea of how well these Turing and Pascal GPUs each scale with overclocking and how they compare with each other. We are pitting a moderately overclocked GTX 1080 Ti FE against an overclocked RTX 2070 SUPER FE to see how well they each scale in games, and to compare relative performance gains made from overclocking each card. There are faster GTX 1080 Tis which overclock better, and there will also be better overclocking RTX 2070 SUPER AIB editions.
Let’s check the performance results of 40 games with our games at stock and overclocked speeds using the Founders Editions of the RTX 2070 SUPER versus the GTX 1080 Ti after the Testing configuration.
Testing Platform
We test 40 games and 3 synthetic benchmarks at 1920×1080 and at 2560×1440. In addition, Hitman 2 and Total War: Warhammer II are benchmarked on DX11 and on DX12. Our latest game is Anno 1800.
Our testing platform is a recent install of Windows 10 64-bit Home Edition v1903, and we are using an i7-8700K which turbos all 6 cores to 4.8GHz, an EVGA Z370 FTW motherboard, and 16GB of XTREEM DDR4 at 3866MHz. The games, settings, and hardware are identical except for the cards being compared.
Before we run our overclocked benchmarks, let’s check out the test configuration.
Test Configuration – Hardware
- Intel Core i7-8700K (HyperThreading and Turbo boost are on to 4.8GHz for all cores; Coffee Lake DX11 CPU graphics).
- EVGA Z370 FTW motherboard (Intel Z370 chipset, latest BIOS, PCIe 3.0/3.1 specification, CrossFire/SLI 8x+8x), supplied by EVGA
- T-FORCE XTREEM 16GB DDR4 (2x8GB, dual channel at 3866MHz), supplied by Team Group
- GTX 1080 Ti 11GB Founders Edition, at stock FE clocks and overclocked, on loan from NVIDIA
- RTX 2060 SUPER 8GB Founders Edition, at stock FE clocks and overclocked, on loan from NVIDIA.
- 2 x 480 GB Team Group SSD; One for Radeon and one for GeForce
- 1.92 TB San Disk enterprise class SSD
- 2 TB Micron 1100 SSD
- Seasonic 850W Gold Focus power supply unit
- EVGA CLC 280mm CPU water cooler, supplied by EVGA
- EVGA NuAudio soundcard, supplied by EVGA
- Edifier 1280T powered speakers
- EVGA DG-77, mid-tower case supplied by EVGA
- Monoprice Crystal Pro 4K
Test Configuration – Software
- GeForce 431.36 WHQL drivers are used for the GTX 1080 Ti and 431.16 (press pre-launch functionally identical) drivers for the RTX 2070 SUPER.
- VSync is forced off.
- AA enabled as noted in games; all in-game settings are specified
- Gaming results show average frame rates in bold including minimum frame rates shown on the chart next to the averages in a smaller italics font. OCAT uses .1ms frametime measurements for the minimums also shown in italics.
- Highest quality sound (stereo) used in all games.
- Windows 10 64-bit Home edition latest version (v1903) and fully updated. All DX11 titles were run under DX11 render paths. DX12 titles are generally run under the DX12 render path unless performance is lower than with DX11. Three games use the Vulkan API.
- Latest DirectX
- All 40 games are updated to their latest versions at time of publication.
- MSI Afterburner, latest beta.
- OCAT, latest version
- Fraps, latest version
- Heaven 4.0 Unigine benchmark
40 PC Game benchmark suite & 3 synthetic tests
Synthetic
- Firestrike – Basic & Extreme
- Time Spy DX12
- Superposition
DX11 Games
- Grand Theft Auto V
- The Witcher 3
- Fallout 4
- Rainbow Six Siege
- Overwatch
- For Honor
- Ghost Recon Wildlands
- Mass Effect: Andromeda
- Prey
- ARK: Survival Evolved
- Project CARS 2
- Middle Earth: Shadow of War
- Total War: Warhammer II
- Destiny 2
- Star Wars: Battlefront II
- Monster Hunter: World
- Kingdom Come: Deliverance
- Final Fantasy XV
- Far Cry 5
- Conan: Exiles
- F1 2018
- Assassin’s Creed: Odyssey
- Call of Duty: Black Ops 4
- Hitman 2
- Just Cause 4
- Resident Evil 2
- Devil May Cry 5
DX12 Games
- Tom Clancy’s The Division
- Gears of War 4
- Civilization VI
- Sniper Elite 4
- Forza 7
- Total War: Warhammer II
- Shadow of the Tomb Raider
- Battlefield V
- Hitman 2
- Metro Exodus
- Tom Clancy’s The Division 2
- Anno 1800
Vulkan Games
- DOOM
- Wolfenstein: The New Colossus
- Strange Brigade
NVIDIA Control Panel settings
Here are the NVIDIA Control Panel settings that match AMD’s settings.
We used the latest beta of Afterburner to set both GeForce’s highest Power and Temperature targets. By setting the Power Limits and Temperature limits to maximum, they do not throttle, but they can reach and maintain its individual maximum clocks. This is particularly beneficial for almost all higher power cards.
Let’s head to the performance results of our overclocking showdown between the GTX 1080 Ti FE and the RTX 2070 SUPER FE.
Performance summary charts
Here are the performance results of 40 games and 3 synthetic tests comparing the stock and overclocked GTX 1080 TI FE 11GB versus the stock and overclocked RTX 2070 SUPER 8GB Founders Edition. The highest settings are always chosen and the settings are listed on the charts. The benches were run at 1920×1080 and at 2560×1440.
Most gaming results show average framerates in bold text, and higher is better. Minimum framerates are next to the averages in italics and in a slightly smaller font. Several games benched with OCAT show average framerates but the .1 minimums are expressed by frametimes in ms where lower numbers are better.
The first two columns are devoted to the RTX 2070 SUPER – first stock, then overclocked. The third and fourth columns represent the GTX 1080 Ti, overclocked then stock (so the overclocked RTX 2070 SUPER results are right next to the overclocked GTX 1080 Ti results). Wins between stock results are given in yellow text and wins between overclocked results are given in gold. If there is a performance tie, both sets of numbers are colored. As always, open each chart in a separate tab for the best viewing.
We see that both cards each gain performance from overclocking. As NVIDIA claims, the RTX 2070 SUPER is overall a bit faster than the GTX 1080 Ti but there results are somewhat inconsistent. The Pascal card appears to be stronger in our older titles whereas Turing is faster in many of the newer games and especially in DX12 games.
Let’s check out our conclusion.
Conclusion
This has been quite an interesting exploration for us in evaluating the stock and overclocked RTX 2070 SUPER FE versus the stock and overclocked GTX 1080 Ti. The RTX 2070 SUPER FE overall is a bit faster than the GTX 1080 Ti using our 40-game benchmark suite at stock. Overclocked the new Turing SUPER pulls away a bit further thanks to the Pascal card being unable to reach the OC it once managed, probably due to aging and electromigration.
Of course, this overclocking showdown is a warm-up for our RX 5700 XT Anniversary Edition review. We are looking forward to posting a series of reviews starting this weekend that will compare its stock and overclocked performance to the GTX 1080 Ti, the RTX 2070 SUPER, the RTX 2060 SUPER, and to the liquid cooled Vega 64.
Stay tuned. In the meantime, if you have any comments or questions, feel free to post them in the comments section below.
Happy Gaming!
CPU for the Nvidia GeForce RTX 2070 For 2021 [Our Reviews and Comparisons]
Table of Contents
When selecting a dedicated graphics card for high-end performance, it is vital to also consider what will be doing all the other work in your machine: the CPU. Having a powerful CPU allows the other, non-graphics code in a graphically-intensive application or game to be processed smoothly and not bottleneck the GPU.2
Think of this review as a dating service for electronics: Hot dedicated GPU (has its own RAM!) looking for temporally-stable, standard-clocked CPU for mutual fun. Prefer motherboards with a wide and snappy data bus. Serious inquiries only. Text me at #shaders.noprobs.
CPU performance is critical for optimal GPU function and there are several considerations that must be addressed if you want to achieve the best results. Among these are motherboard design and cooling.
Building your own machine has come a long way from the days of Heathkit computers. Most electronics components retailers and wholesalers have complete DIY kit computer packages where all the children in the kindergarten are guaranteed to play well together.
Going it on your own, however, usually requires a bit of research to ensure that the motherboard has a compatible bus width and clock speed for your CPU of interest and that you can build in a cooling system and the right power supply to drive all the components, including the GPU. Custom builds are really a topic for a different time, so today we are going to focus on which CPUs play nicely with the GeForce RTX 2070 by Nvidia.
PRODUCTFEATURESLATEST PRICE
| 1. AMD Ryzen 7 3800XT |
|
Check Price |
| 2. Ryzen 5 3600XT |
|
Check Price |
| 3. Intel Core i9-9900K |
|
Check Price |
| 4. Intel Core i7-10700F |
|
Check Price |
| 5. Intel Core i7-9700KF |
|
Check Price |
| 6. AMD Ryzen 7 2700X |
|
Check Price |
| 7. Ryzen 5 2600X |
|
Check Price |
| 8. EVGA 01G-P3-1312-LR |
|
Check Price |
| 9. |
|
Check Price |
| 10. Intel Core i5-10500 |
|
Check Price |
The GeForce RTX 2070 is the first GPU in a new approach to graphics processing, performing real-time ray tracing (42 ray-tracing operations per second), and utilizing an artificial intelligence system (Turing architecture) to interpolate graphics parameters. Graphics generation is surprisingly realistic and completely lacks the cartoonish image feel of earlier graphics cards.
The on-board RAM, which is an 8 GB GDDR6 operating at 1410 MHz (a maximum boost of 1620 MHz) and processing 14 Gbps, supports a total of 2304 CUDA (compute unified device architecture) cores producing a peak of 6 Giga rays/s.
As its default factory setting, the RTX 211070 is overclocked and is stabilized by a 6-phase power supply to provide all the power demands of the board. A pair of on-board 13-blade, dual-axis fans that drive airflow through a vapor chamber play some heat exchange tricks to maintain low operating temperatures, even under heavy load.
All this electronic goodness would be a moot point if you paired the RTX 2070 with a lemon of a CPU. In for a penny, in for a pound: if you are going to invest in a high-end GPU, you need a comparable CPU to handle all the other stuff.
You wouldn’t buy a 5000 square foot home and not have at least a part-time cleaning person, right? The same goes here for having a proper CPU.
Our list here will nit-pick the details and cast some aspersions on these processors to achieve some balance in our review, but keep in mind that there is not an ugly dog in the bunch. Even our last pick in the list (Intel Core i5-9600K) is quite a capable device and will definitely produce pleasing performance results.
You May Also Like: CPUs for Nvidia RTX2060
Below is our list of processors that work well with the GeForce RTX 2070 by Nvidia. They are listed in benchmarked order based on overall system performance in tests conducted by PassMark.
The processors vary in price per computing power and you may find processors lower down on the list more affordable and just as effective. Here are our top 10 for the year 2020.
1. AMD Ryzen 7 3800XT
Product Highlights
The Ryzen 7 3800XT by AMD is the latest release in their Ryzen 3000 line and features 7 nm Zen 2 architecture and excellent performance for in an 8-core chip. The 3800XT clock speed is 3.9 GHz with boosts to 4.7 GHz and also has an L2 (4 MB) and an L3 cache (32 MB).
However, the chipset does not include a CPU cooler (AMD says you do not need it for most applications). Intel has been selling their i7 chips for some time also without coolers
It is also an unlocked chipset, so overclocking with a maximum 1% increase in performance is an option. If you want to overclock and need a cooler, we recommended using either an AMD Wraith Prism or MSI MAG CoreLiquid 240R for heavy-duty gaming.
The 3800XT is marketed as a competitive gaming and video streaming chipset and is guaranteed to remain cool with standard heat exchange systems. The gaming rating by CPU Agent is 95%.
Features
-
- Architecture: 7 nm Zen 2 (generation 4)
- Clock: 3.9 GHz (burst to 4.7 GHz)
- Cores: 8
- Cache: L2 (4 MB), L3 (32 MB)
What We Like About AMD Ryzen 7 3800XT
The performance of the 3800XT is exemplary in gaming tests and runs on a par with Intel’s Core i9-10900K. This is a slight improvement over the earlier 3800X chipset by a 4% boost in performance and 6% boosts in single-core processing speed.
Considering that the 3800XT is basically the same price as the 3800X, the increase in performance has been well-received. The PassMark benchmark score is 23876 (current top of the commercial field is the AMD Ryzen ThreadRipper 3990X scored at 80748, which is a high-end chipset for graphics workstations).
What We Don’t Like About AMD Ryzen 7 3800XT
There are not too many bad things to say about eh 3800XT, other than its marginal performance improvement over the 3800X. The difference in price between the two processors is negligible (around $50 difference).
Under gaming load, the 3800XT ran somewhat warm at 78.5°C without a cooler. The chipset also uses a lot of power, consuming about 120-124 Watts.
Regardless of all the computer power, the 3800XT can still present a bottleneck to GPUs like the RTX 2070 if the display is run at 1080p under ultra-graphics settings. The average bottleneck loss is around 13.5% under those conditions.
PROS
- Affordable compared with other chips of similar performance
- Great for gaming
- Efficient multithreading
- New technology that will remain relevant for some time
CONS
- Uses a lot of power
- Can bottleneck GPUs on ultra-settings
2.
Ryzen 5 3600XT
Product Highlights
The AMD Ryzen 5 3600XT is considered a mid-range CPU that features 6-cores with a base clock speed of 3.8 GHz and burst speed to 4.5 GHz. The 3600XT is another AMD offering in their Matisse Refresh 7 nm architecture series.
Overclocking is an option with the 3600XT, but stable speed increases will only be about 1%. One of the advantages of 3600XT is its price point, which is very affordable for a chipset of this capability.
Newer games are moving toward multithreading and this is where Ryzen architecture really shines, despite falling short of Intel processors in single-core performance. Boost mode tends to offset these shortcomings, and the AMD excels at handling multithreading code.
The gaming rating by CPU Agent is 94%.
Features
-
- Architecture: 7 nm Zen 2 (generation 4)
- Clock: 3.8 GHz (burst to 4.5 GHz)
- Cores: 6
- Cache: L2 (4 MB), L3 (32 MB)
What We Like About Intel Core i9-9900K
The i9-9900K with 8 cores is a multithreading powerhouse that really excels in terms of video editing and gaming prowess. Bottlenecking with 1080p graphics on ultra-settings was only on average 6.5%, which was quite surprising to us.
Running newer games and streaming straight to Twitch can be handled with no problems by the i9-9900K. The PassMark benchmark score is 18885.
What We Don’t Like About Intel Core i9-9900K
Despite its multithreading capabilities and ability to efficiently process code and drive fast refresh rates, freeing up the GPU and providing low bottleneck problems, what you get will be limited to factory settings as overclocking does not provide significant boosts in performance.
This is pretty much true for all the top-end processors we will review here. Basically, they are nearly maxed out straight from the box.
Power consumption is an issue with the i9-9900K, particularly under turbo boost mode which consumes 210 Watts of power. That is an important factor in your build, as you will need a power supply to cover it.
PROS
- Adept at multithreading
- Tops for gaming with low bottleneck score
- Calculation speed for large spreadsheets is impressive
- High boost speed (watch out for power requirements!)
CONS
- Pricey
- Thermal controls are not the newest
- Despite being an i9, architecture is already dated
4.
Intel Core i7-10700F
Product Highlights
The Core i7-10700F by Intel is a newly released, high-end processor that sports 8 cores built on 14 nm Comet Lake architecture (what is it with Intel and lakes, anyway?). The base clock speed for the i7-10700F is a hearty 2.9 GHz with burst speeds of 4.8 GHz.
Out of the box, there is basically no, we repeat, no overclocking potential in this chip as it already is cranking at maximum. Although we mentioned that overclocking is limited in the chips reviewed above, we feel the need to take a moment to explain why going forward, overclocking may well become passé for higher-end chipsets.
Both Intel and AMD have made great strides in boosting computing power as well as a lot of progress in temperature control for their newer chips, as many are sold without accompanying CPU coolers. The need for coolers is also moot since the chips are not really capable of being overclocked.
Overclocking in the past was mainly because chips were sold locked and underclocked to control heat, so newer chipsets are now more optimal for maximum performance based on their architecture.
The gaming rating by CPU Agent is 96%.
Features
-
- Architecture: 14 nm (generation 10)
- Clock: 2.9 GHz (burst to 4.8 GHz)
- Cores: 8
- Cache: L2 (2 MB), L3 (16 MB)
What We Like About Intel Core i7-10700F
Multithread processing is a developed talent for the i7-10700F and will give this chip some legs for future use. The in-built cooling feature for the i7-10700F is a standard pancake-style cooling plate.
As a multimedia editing platform, the i7-10700F has teeth and will easily perform these tasks, provided it is mated to a nice motherboard like a 4470 or other similar design. Gaming performance is excellent at 96% rated by CPU Agent.
The bottleneck score with the RTX 2070 averaged about 11.3% at 1080p with ultra-graphics settings, which is still very good. The PassMark benchmark score is 16043.
What We Don’t Like About Intel Core i7-10700F
There is no form factor for the i7-10700F that can accommodate integrated Intel Graphics, but that is really true for most of the chips on this list. Despite this apparent shortcoming (which is really not one), the cost production factor of the i7-10700F is less than comparable Intel offerings, such as the Core i7-9700.
Although a new Intel offering, the architecture is still 14 nm where many new chips are being produced in the 10 nm architecture range.
That being said, the chip is not much faster than many i5 processors offered by Intel, which are just as capable for a lower price.
PROS
- Great for productivity
- Great for gaming
- Efficient (only uses 65 Watts)
- Pricing per thread of performance is low
CONS
- The base speed for all-core performance is low
- Uses a PCIe 3.0 interface, which is dated
5. Intel Core i7-9700KF
Product Highlights
The Core i7-9700KF is a recently released CPU by Intel under their Coffee Lake Refresh line of 14 nm architectures. It has an 8-core processor running at 3.6 GHz with burst speeds up to 4.9 GHz.
The chip is fully functional for multithread processing for productivity and also is quite robust in the gaming arena. The i7-9700KF has no form factor for integrated graphics, which saves a little on production costs.
As far as game performance, the i7-9700KF is rated by CPU Agent at 97%. Bottlenecking with the RTX 2070 at 1080p using ultra graphics settings averaged about 10.2%, which is rather good.
Overclocking gains are maximum with a 2% increased performance, higher than other chips we have reviewed.
Features
-
- Architecture: 14 nm (generation 9)
- Clock: 3.6 GHz (burst to 4.9 GHz)
- Cores: 8
- Cache: L2 (2 MB), L3 (12 MB)
What We Like About Intel Core i7-9700KF
The Core i7-9700KF is an excellent choice for productivity and for video editing and computer graphics design, provided it is mated with an appropriate motherboard to make the best use of all its features. The lack of support for integrated graphics allows Intel to offer this chip at a lower price than comparably-performing Intel chips.
The bottleneck score with the RTX 2070 averaged only 10.2% at 1080p with ultra-graphics settings, which is a decent score. The PassMark benchmark score is 14674.
For these reasons, the i7-9700 series dominates the mainstream CPU market.
What We Don’t Like About Intel Core i7-9700KF
The i7-9700KF is price-competitive with other similar chips, but we feel it is overpriced for generation 9, 14 nm architecture. Power use is rather high at processor idle, coming in at 95 Watts.
PROS
- A gamers CPU
- Excellent base speed using all cores
- Multitask performance is better than competitors
- Robust boost speed
CONS
- Pricing is a bit high for this chip
- Thermals are dated an no cooler is included
- Uses a PCIe 3.0 bus
6. AMD Ryzen 7 2700X
Product Highlights
The Ryzen 7 2700X by AMD is an 8-core, 12 nm architecture based on AMD’s Pinnacle Ridge design (we guess AMD likes ridges). The processor clock speed at idle is 3.7 GHz and boosts to 4.3 GHz, all while consuming 105 Watts of power.
This is a hallmark of many AMD designs in that they are known to be more energy-efficient and produce less heat under load. Overclocking potential is much greater than other chips we examined with the 2700X capable of a full 2% boost in performance.
Bottlenecking at 1080p under ultra-graphics settings with the RTX 2070 averaged 22.5%, leading to a gaming score of 88% by CPU Agent. PassMark benchmark score was 17558.
Features
-
- Architecture: 12 nm Zen+ (generation 2)
- Clock: 3.7 GHz (burst to 4.3 GHz)
- Cores: 8
- Cache: L2 (4 MB), L3 (16 MB)
What We Like About AMD Ryzen 7 2700X
The 2700X has high productivity ratings (83% by CPU Agent) due to multithreading abilities and has the potential for overclocking, yielding a 2% boost in performance. Although AMD chips tend to fall short on single-core performance compared with Intel, many newer game titles are making better use of multithreaded CPUs.
What We Don’t Like About AMD Ryzen 7 2700X
For some game titles that rely on single-core processing, the 2700X underperforms compared with comparably-priced Intel offerings, as many users have noted.
PROS
- Affordable for the performance
- Great for gaming
- All-core base speed is excellent
- Better at multitasking compared with similar competitor chips
CONS
- Eats a lot of power
- Thermals are dated
7. Ryzen 5 2600X
Product Highlights
The AMD Ryzen 5 2600X is the little brother of the newer 2700X and has a similar architecture, based on Pinnacle Ridge 12 nm design with 6 cores instead of 8. The clock speed is 3.6 GHz with burst capability of 4.2 GHz. Its gaming rating by CPU Agent is 88% with an RTX 2070 bottleneck of 24. 9% at 1080p ultra graphics settings.
Features
-
- Architecture: 12 nm Zen+ (generation 2)
- Clock: 3.6 GHz (burst to 4.2 GHz)
- Cores: 6
- Cache: L2 (3 MB), L3 (16 MB)
What We Like About Ryzen 5 2600X
The 2600X is a very capable chip that can be picked up for a good price (under $200). Overclocking will yield a 2% increase in overall performance.
It uses less power than its older brother at 95 Watts and is also cool-running. PassMark benchmark is 14076.
What We Don’t Like About Ryzen 5 2600X
Not as highly multithreaded as other processors, relying mainly on one core.
PROS
- Very affordable
- Good for gaming
- Most sales packages contain a CPU cooler
CONS
- The thermals are becoming a dated design
- This is somewhat old technology (but still works)
8.
Intel Core i7-8700K
Product Highlights
The Intel Core i7-8700K is a robust mid-grade processor with 6 cores and 14 nm Coffee Lake architecture. Clock speed at idle is 3.7 GHz and burst speeds are at 4.7 GHz.
Power usage is high at 95 Watts. Bottlenecking with the RTX 2070 at 1080p and ultra-graphics settings were rather good for a mid-grade processor (12.5%), which gave a gaming score of 96%
See Related Topic: RTX Laptops
Features
-
- Architecture: 14 nm Coffee Lake (generation 8)
- Clock: 3.7 GHz (burst to 4.7 GHz)
- Cores: 6
- Cache: L2 (1.5 MB), L3 (16 MB)
What We Like About Intel Core i7-8700K
Overclock potential for the 8700K is 4%. PassMark benchmark was 13964.
What We Don’t Like About Intel Core i7-8700K
For an older technology chip, despite its excellent performance, the 8700K is overpriced.
PROS
- A gamer’s CPU
- Adept at multithreading
- Great burst speed
CONS
- Technology is getting dated
- More expensive than necessary
9. Intel Core i5-9600KF
Product Highlights
The Core i5-9600KF by Intel is a mid-grade processor with 6 cores and 14 nm architecture. Base clock speed is 3.7 GHz with bursts to 4.6 GHz.
Gaming score was excellent at 93% and RTX 2070 bottlenecking was 19.1% at 1080p and ultra-graphics settings.
Features
- Architecture: 14 nm Coffee Lake Refresh (generation 9)
- Clock: 3.7 GHz (burst to 4.6 GHz)
- Cores: 6
- Cache: L2 (1.5 MB), L3 (9 MB)
What We Like About Intel Core i5-9600KF
The 9600KF has marginal speed improvements over its predecessors in the 9600 line. It is a good multithreading processor, a solid performer, and very reasonably priced (you can find some for close to $100 if you look hard enough).
The overclocking potential is 8.5%. PassMark benchmark is 10928.
What We Don’t Like About Intel Core i5-9600KF
The standard package does not include a CPU cooler.
PROS
- Strong gaming CPU
- Vigorous boost speed
CONS
- Dated thermals technology
- No CPU cooler included
10. Intel Core i5-10500
Product Highlights
The Core i5-10500 is on CPU on our list with integrated graphics that will have to be disabled for the RTX 2070 to run. The i5-10500 has 6 cores and a 3.1 GHz clock speed with 4.5 GHz burst.
Features
-
- Architecture: 14 nm Comet Lake (generation 10)
- Clock: 3.1 GHz (burst to 4.5 GHz)
- Cores: 6
- Cache: L2 (1.5 MB), L3 (12 MB)
What We Like About Intel Core i5-10500
The gaming score is 93% with a bottleneck on the RTX 2070 of 16. 9% at 1080p and ultra-graphics settings. PassMark benchmark is 13380.
What We Don’t Like About Intel Core i5-10500
There is no overclocking potential for this chip.
PROS
- Very affordable
- Has integrated graphics if you want to switch over to save power
- Great boost speed
- Great for gaming and has a CPU cooler included
CONS
- No overclocking potential
Final Verdict
Although we do like AMD chips, our main pick would be the Intel Core i5-9600KF for performance and a reasonable price. Bottlenecking is down near 10% for this chip. A good alternative is the Intel Core i7-8700K with similar performance, but will usually tend to sell at a higher price.
Overclocking MSI GeForce RTX 2070 Armor 8G Graphics Card
Contents
As most of you know, with most graphics cards you can apply a simple series of tricks to slightly increase overall performance. Generally, you can adjust core clock speeds and voltages. Overclocking the MSI GeForce RTX 2070 Armor 8G graphics card is the topic of today.
Overclocking the GeForce RTX 2080: Dry Ice Cooling
By increasing the frequency of the graphics card memory and GPU , we can force the video card to increase the calculation cycles per second . It sounds complicated, but it really can be done in less than a few minutes. I always recommend to novice users and beginners not to increase the frequency higher than by 5% on the core and video memory.
Example: If your card is running at 600 MHz (which is quite common these days), I suggest you don’t increase the frequency above 30 to 50 MHz.
How to overclock MSI GeForce RTX 2070 Armor 8G
More advanced users often raise the frequency a little higher. Usually, when your 3D graphics start showing artifacts like white dots («snow»), you should back off by 25 MHz and leave it at that.
Usually when you overclock the map will start showing artifacts or empty polygons. And it can even freeze. Carefully find this limit, and then retreat at least 20 MHz from the moment the artifact was discovered.
Look carefully and observe carefully. I don’t really know why you need to overclock a proven card anyway, but we’ll look into it anyway. In general … you always overclock at your own peril and risk! So be careful.
We have prepared a new beta version (download) with AfterBurner that you can use to overclock your graphics card. Can be overclocked manually as well as with a new curve based on the OC Scanner. You will see that most cards will stop at about the same levels due to all kinds of hardware protection in them.
As a result, we got the following overclocking parameters:
- Temperature 88 Degrees C
- Power Limiter: 114%0078
- Mem frequency + 1000 MHz
- Voltage +100%
- Default fan RPM
Above, the relative performance difference between the reference card, cards with the corresponding default frequencies, with results in percent.
On the far right, where you can see «Average Difference %», this is the average of the four games tested. When overclocking manually, we got 10%. What can be seen from the reference performance with this setting.
Overclocking a GeForce RTX 2070 with OC Scanner
NVIDIA has introduced a new overclocking and tuning method, starting with RTX (and only). Basically, now you can automatically overclock the processor. By clicking the OC scanner button, you can test the software algorithm. And also check the intervals and voltage frequency curves.
The scanner will try to determine a stable overclock, which is represented by the curve, and this curve can be applied to your setup.
In (beta 4.6.0) Afterburner press Control+F and you will see the OC scanner button. Pressing the scan button may take 10 to 20 minutes of testing. This will define a new stable acceleration curve. And as soon as everything is completed in Afterburner, click apply, and everything will be done.
Oddly enough, everything is pretty fast, just like with manual tuning. And the OC scanner curve is much friendlier:
- Set the power limiter to MAX
- Find a stable memory frequency (+1000 in our case)
- Leave voltage control as default
- Click OC Scanner (takes 10 minutes)
- When done click Apply in AB.
- Run the test and see the results.
- Enjoy almost perfect automatic settings.
This is a softer way to tune the than manual tuning. The algorithm is defined by NVIDIA. With my added memory and power limiter settings. My tip , Find a stable memory overclock, turn on the power limiter and try to scan. In the end, you may be surprised. However, the results vary.
That’s all, thank you for your attention.
Views Today: 3,904
Related Posts0037
MSI GeForce RTX 2070 Armor 8G review. Big test GeForce RTX 2070
-
Thursday, 28 March 2019
-
- 1
- 2
- 3
- 4
- 5
(6 votes)
-
Video cards
-
Author
Alexander
The TU106 graphics processor has become the main one for two video cards — GeForce RTX 2070 and GeForce RTX 2060. We will get acquainted with the older model in this review using the MSI GeForce RTX 2070 Armor 8G as an example. Let’s compare the performance with older solutions and cards of the previous generation, and evaluate the overclocking capabilities.
The structure of the GPU and the features of the new architecture are described in a separate review. So just to recap, the TU106 has 2304 CUDA cores, 288 tensor cores, and 36 RT cores, plus 144 texture units and 64 render units. In terms of the number of functional blocks, the GeForce RTX 2070 is inferior to the old GeForce GTX 1080 video adapter, but this is offset by architectural improvements. The GeForce RTX 2070 is also equipped with fast GDDR6 memory with an effective frequency of 14000 MHz, which runs on a 256-bit bus. The declared frequencies for the processor are 1410/1620 (Base/Boost) in case of simple versions and 1410/1710 for NVIDIA’s Founders Edition models.
The considered version from MSI is one of the simple video cards with the minimum frequencies for the series. At the same time, we decided to evaluate the capabilities of the Founders Edition by additionally conducting tests with increased frequencies. Also, there will be mandatory tests during overclocking to maximum frequencies.
MSI GeForce RTX 2070 Armor 8 G
Comes in medium size MSI Armor box. Against the background of other MSI video cards, they stand out with a large set of polygraphy, there is even a small comic book and cardboard cup holders in the kit.
The video card is large, its length is more than 30 cm. The cooling system takes up more than two slots, which has become the standard for many older GeForce RTX.
Housing in black and white. Its aggressive swept-back shape and large fans with a special curve of the blades are reminiscent of the models of the Gaming line.
The back is completely covered with a metal backplate. The logo in the shape of a dragon is applied to the surface.
There are two power connectors in the corner: 6-pin and 8-pin.
There is an illuminated window on the white edge. Color and lighting effects can be changed through the Mystic Light app.
The video card has five ports for displaying images: three DisplayPort, HDMI and VirtualLink (USB Type-C).
The cooling system looks very solid. Unlike MSI Gaming, Armor does not have a large plate at the base, direct contact with heat pipes is implemented. They are well fitted to each other to minimize possible gap between them. But the side tubes are only partially in contact with the GPU die.
A plastic casing with two fans is screwed onto the radiator from the outside. The radiator itself consists of two sections of thin aluminum plates pierced by five copper tubes.
Heat sink base removes heat from memory chips and some other components.
The board is covered with a radiator plate that cools the elements of the supply circuit. There are thermal pads on the chokes and capacitors of the GPU power subsystem that rest against the main heatsink. But there is little benefit from contact with thin ribs.
Large fans up to 100 mm diameter (95 mm impeller) are used. They are made using TorX technology — the special shape of the blades improves the power of the air flow, increasing the efficiency of the entire cooling system. Zero Frozr technology ensures absolute silence when idle by stopping the fans at low temperatures.
The backplate is also part of the cooling system. This metal plate draws heat away from hot areas under the graphics chip and memory. Contact is made through elastic pads. The rest of the plate surface is covered with a protective film.
The massive PCB clearly uses PCBs from MSI’s older graphics cards.
The GPU power system has 8 phases, but 6 are soldered. The two-phase memory power system is located on the left side of the board.
The fuses appeared in the power circuit, which were not in the MSI GeForce GTX 10.
Full marking of the processor TU106-400-A1. The board has 8 Micron memory chips (8OA77 D9WCW).
Operating frequencies correspond to standard specifications: this is an average Boost Clock of 1620 MHz and an effective memory frequency of 14000 MHz.
EVGA Precision X1 utility was used to monitor graphics card parameters. Testing was carried out on an open stand at a room temperature of 21 °C. Under these conditions, the temperature of the GPU did not exceed 69 °C. Boost frequencies varied greatly, for example, in the Final Fantasy XV benchmark, the core frequency range was at the level of 1650-1710 MHz.
The highest frequencies were observed in Strange Brigade, here the operating range was 1720-1800 MHz. Peak Boost values reached 1845-1860 MHz.
The fans were running at 1400-1550 rpm with minimal noise. Inside the case in more severe temperature conditions, heating and noise will be higher, but the video card will not go beyond comfortable performance.
During overclocking, the core frequency was increased by 170 MHz, which, with an increase in the power limit, ensured an increase in the average frequencies to about 1900 MHz at peak values up to 2070 MHz. The memory was able to operate at an effective frequency of 16242 MHz. In heavy tests (Final Fantasy XV 4K), Boost drops to 1860 MHz can be observed. But in most games, the frequencies exceeded the level of 1900 MHz.
The final core frequencies are not very high, although on the older GeForce RTX 2080 we also received average frequencies below 2 GHz under heavy load. You can hope for better overclocking and more stable Boost frequencies if you use the top versions of the GeForce RTX 2070 with an increased power limit.
Our copy of the MSI GeForce RTX 2070 will replace the Founders Edition with a corresponding increase in core frequencies.
The technical data of the participants are given in the table below.
Characteristics of test participants
|
GeForce RTX 2080 Founders Edition |
GeForce GTX 1080 Ti |
GeForce RTX 2070 Founders Edition |
GeForce RTX 2070 |
GeForce GTX 1080 |
Radeon RX Vega 64 |
|
|
Architecture |
Turing |
Pascal |
Turing |
Turing |
Pascal |
Vega |
|
Core |
TU104 |
GP102 |
TU106 |
TU106 |
GP104 |
Vega 10 |
|
Number of transistors, million pieces |
13600 |
12000 |
10800 |
10800 |
7200 |
12500 |
|
Technical process, nm |
12 |
16 |
12 |
12 |
16 |
14 |
|
Core area, sq. |
545 |
471 |
445 |
445 |
314 |
486 |
|
Number of CUDA stream processors |
2944 |
3584 |
2304 |
2304 |
2560 |
4096 |
|
Number of tensor cores |
368 |
— |
288 |
288 |
— |
— |
|
Number of cores RT |
46 |
— |
36 |
36 |
— |
— |
|
Number of texture units ROP |
184 |
224 |
144 |
144 |
160 |
256 |
|
Number of render units |
64 |
88 |
64 |
64 |
64 |
64 |
|
Core clock (Base/Boost), MHz |
1515-1800 |
1480-1582 |
1410-1710 |
1410-1620 |
1607-1733 |
1274-1546 |
|
Memory bus, bit |
256 |
352 |
256 |
256 |
256 |
2048 |
|
Memory type |
GDDR6 |
GDDR5X |
GDDR6 |
GDDR6 |
GDDR5 |
HBM2 |
|
Effective memory frequency, MHz |
14000 |
11016 |
14000 |
14000 |
10000 |
1890 |
|
Memory, GB |
8 |
11 |
8 |
8 |
8 |
8 |
|
Interface |
PCI-E 3. |
PCI-E 3.0 |
PCI-E 3.0 |
PCI-E 3.0 |
PCI-E 3.0 |
PCI-E 3.0 |
|
Power TDP, W |
225 |
250 |
185 |
175 |
180 |
295 |
The official data for Boost are indicated, which characterize the average frequency level. The graphs with the results show the full range of GPU operating frequencies — from the base value to the peak Boost.
Test stand
- Processor: Intel Core i7-6950X @4.2GHz
- motherboard: MSI X99S MPOWER
- memory: DDR4 HyperX Predator HX432C16PB3K4/32 (4×8 GB), 3200 MHz
- system drive: SSD Kingston A400 SA400S37/240G
- optional drive: Hitachi HDS721010CLA332, 1 TB
- power supply: Seasonic SS-750KM
- operating system: Windows 10 Ultimate x64
- GeForce driver: NVIDIA GeForce 416.
34/417.35
- Radeon Driver: AMD Adrenalin Edition 18.12.1.1
Tested at 2560×1440 and 3840×2160. Synthetic benchmarks come first, games are listed in alphabetical order. Only 4K mode is enabled in CPU-intensive games. On the left side of the diagrams, the minimum frame rate is indicated, on the right, the average fps. The full methodology tested is described here.
More in this category:
« Review and testing GeForce GTX 1050 3GB
Review and testing MSI GeForce RTX 2070 Gaming Z 8G »
More in this category:
« Review and testing GeForce GTX 1050 3GB
Review and testing MSI GeForce RTX 2070 Gaming Z 8G »
MSI GeForce RTX 2070 Super GAMING X Trio Video Card Review: Triple the Power
Real competition in the discrete graphics market benefits both developers and potential owners of new graphics cards. The first helps to stay in good shape and move forward, the second — to spend their money with maximum return. A good example of this is the improved Super versions of the popular models of the GeForce RTX line, which appeared partly as a response to the release of new AMD products. One of the most interesting updates of this wave was the GeForce RTX 2070 Super graphics card. An extraordinary model of the line came to our review — MSI GeForce RTX 2070 Super Gaming X Trio. What is such a large version with a massive cooling system capable of?
To further accelerate the GeForce RTX 2070, a GPU replacement was indispensable. In the standard version for this series, a GPU (TU106-400A) with all unlocked blocks was already used, and the chip itself worked at fairly high frequencies. Therefore, the appearance of TU104 as part of the Super-modification was quite logical.
Course
Development in Python
Introduce popular language programming to earn $1300 via rіk
REGISTER!
| GeForce RTX 2080 | GeForce RTX 2070 Super | GeForce RTX 2070 | |
| Crystal name | TU104-400A | TU104-410 | TU106-400A |
| Graphic architecture | Turing | Turing | Turing |
| Manufacturing process | 12 nm | 12 nm | 12 nm |
| Crystal area | 545 mm² | 545 mm² | 445 mm² |
| Number of transistors, billion | 13. |
13.6 billion | 10.8 billion |
| GPU clock speed | 1515/1710 MHz | 1605/1770 MHz | 1410/1620 MHz |
| Stream Processors | 2944 | 2560 | 2304 |
| Texture blocks | 184 | 160 | 144 |
| Rasterizers | 64 | 64 | 64 |
| RT cores | 46 | 40 | 36 |
| Tensor cores | 368 | 320 | 288 |
| Memory (type, capacity) | GDDR6 8 GB | GDDR6 8 GB | GDDR6 8 GB |
| Memory bus | 256 bit | 256 bit | 256 bit |
| Effective memory clock speed | 14000 MHz | 14000 MHz | 14000 MHz |
| Memory bandwidth | 448 GB/s | 448 GB/s | 448 GB/s |
| Power consumption, W | 215 W | 215 W | 175 W |
As a result, the number of CUDA computers increased from 2304 to 2560 (+11. 1%), and texture units — from 144 to 160. Also, the number of hardware RT computers increased proportionally (from 36 to 40) and tensor cores — from 288 to 320. The memory subsystem has not changed. For the video card, the same 8 GB of high-speed GDDR6 is used, operating at an effective 14,000 MHz. A 256-bit bus in this embodiment provides a total throughput of 448 GB / s.
Looking at the specification table, we can say that the GeForce RTX 2070 Super is more likely a somewhat lighter version of the GeForce RTX 2080 than an accelerated GeForce RTX 2070. It is only related to the latter by the recommended price of $ 500 for the American market and nominal positioning.
Recall that the GeForce RTX 2070 Super, as well as the enhanced junior modification, the GeForce RTX 2060 Super, were introduced on July 9th. Just two days after the official launch of the Radeon RX 5700/5700 XT. It’s probably just a coincidence. Not really. Such an update to the GeForce RTX line obviously somewhat blurred the appearance of new products from AMD, which initially planned to compete with the classic GeForce RTX 2070 and RTX 2060. With the advent of Super-versions of adapters, the main opponent had to hastily adjust the price tags for their products. But, that’s a completely different story.
MSI GeForce RTX 2070 Super GAMING X Trio
For the launch of the GeForce RTX 2070 Super, MSI introduced almost a dozen different modifications of this series of adapters. The manufacturer offers similar versions in the starting lines VENTUS, ARMOR, improved GAMING, as well as three-fan GAMING Trio, which were previously used only for top-end adapters with NVIDIA chips.
It is in this design that the MSI GeForce RTX 2070 Super GAMING X Trio that came to our review was made. Of course the video card is factory overclocked. The graphics processor received a frequency formula of 1605/1800 MHz, instead of the recommended 1605/1770 MHz. GDDR6 memory runs at a specified 14,000 MHz.
In a similar design, the manufacturer also offers a modification of the MSI GeForce RTX 2070 Super GAMING Z Trio, which, in addition to the GPU (1605/1815 MHz), also has accelerated local memory (15,500 MHz), which is now extremely rare.
Delivery set
The video card is offered in a fairly large box, the dimensions are rather typical for a motherboard. On the packaging are large-scale photos of the device and the model name.
The adapter comes with a set of documentation, a CD with drivers and software, and an additional metal bracket to support and securely fix the video card in the PC case. Such elements are no longer a novelty for «heavyweights», they can be useful especially during the transportation of the system unit.
Design and layout
The large size of the video card is the first thing you notice after unpacking the device. Yes, and one and a half kilograms of weight hint at the fact that we are dealing with an uncommon graphics adapter.
Equipped with a massive Tri-Frozr cooler. The overall radiator block consists of two stacked cassettes of aluminum plates with an original wavy profile shape, which additionally increases the dispersion area and makes it possible to use the air flow generated by the fans more efficiently.
To speed up the transfer of heat, the blocks are pierced by seven nickel-plated heat pipes, one of which has a diameter of 8 mm, the rest — 6 mm. For uniform heat removal from the entire surface of the GPU chip, the design of the heat sink includes a nickel-plated copper distribution plate.
The height of the plates is also increased (22-28 mm), which automatically makes the cooling system bulky, making it impossible to get by with two slots for installing a video card. The manufacturer does not even try to downplay the size of the device, introducing specific terms like «2.7-slot» or something similar. To install a video card, you definitely need three free slots.
In the profile of the plates there are specific deflectors that concentrate the air flow in the areas where the heat pipes are located.
As expected MSI GeForce RTX 2070 Super GAMING X Trio uses original PCB with enhanced power supply. The layout of the stabilizer suggests a 10-phase circuit (8 + 2) with spectacular DrMOS assemblies. The MP2888A controller from Monolithic Power Systems controls the VRM settings for the GPU, and the uP9 controls the two-phase voltage regulator on the memory chips.512P.
Power assemblies and chokes are in contact with the main radiator block through thermal pads. Additional cooling is also provided for memory chips. A separate aluminum plate covers the GDDR6 chips, and is also attached to the interface panel with screws, forming an additional stiffening rib.
On the reverse side, the printed circuit board is covered with an aluminum plate with versatile decorative polishing. The protective element is also used to optimize the temperature regime. Thermal pads in the GPU area help speed up heat dissipation from the hot zone on the PCB. Of the interesting options, one can also note additional decorative holes on the plane of the panel to improve ventilation.
Three axial fans are used to blow the radiator block. Two of them have a diameter of 100 mm and one more — 90 mm. The fans have double ball bearings and are designed according to the Torx FAN 3.0 design, which has a special shape of the blade profile with a curve and specific notches. According to the developers, such improvements increase the static pressure without increasing the fan speed.
There are a lot of small nuances in the overall design of the cooler, which form the general impression of the developers’ attention to details. One example is an additional clamp for cables that connect fans and lights.
Two 8-pin connectors are provided for connecting the auxiliary power supply. In this configuration, the adapter can receive up to 375 watts of power, given the capabilities of the PCI-E x16 slot. Recall that the thermal package of the GeForce RTX 2070 Super is 215 W, and in the basic configuration, video cards can easily get by with a bunch of 6 + 8 connectors, but for tuned versions, the 8 + 8 option is quite expected.
However, despite a certain factory overclocking of the MSI GeForce RTX 2070 Super GAMING X Trio, the recommendations for the power supply for a system with this video card remain the same as for PCs with a reference GeForce RTX 2070 Super — from 650 W.
The NVLink connector on the top edge reminds you of the support for SLI-links with two video cards, with minimal synchronization costs. In the case of devices based on chips with the Turing architecture, NVLink connectors were previously only equipped with the GeForce RTX 2080/2080 Ti, now this option is available starting with the GeForce RTX 2070 Super.
MSI GeForce RTX 2070 Super GAMING X Trio has additional RGB lighting. And here it is in abundance. The logo of the game series with a recognizable dragon silhouette is highlighted on the top panel. Matted translucent segments with a set of LEDs are fixed on the upper and lower edges of the front panel. Sources provide a pleasant soft glow, which, if necessary, can be very intense.
The backlight is controlled by Mystic Light, which is part of the proprietary MSI Dragon Center complex. Two dozen different modes are available here, it is possible to synchronize the backlight with that for other components of the platform.
The interface panel has four video outputs: three full-sized DisplayPort 1.4 and HDMI 2.0b. The manufacturer did not include in the connectors a specific USB Type-C port with a VirtualLink channel for connecting virtual reality helmets using a single cable that transmits both the video signal and power. If the prospect of connecting such headsets is not of great importance to you, then the absence of such a connector is a small loss.
Part of the interface plate is perforated to vent heated air outside the system unit. Axial-type fans are used for the video card, and the radiator plates are transversely oriented, so there is no large outward flow here, rather, we are talking about additional ventilation holes.
The MSI GeForce RTX 2070 Super GAMING X Trio has impressive overall dimensions. We have already mentioned the three-slot layout, but it is also important to note that the video card is 328 mm long. In this case, before buying an adapter, you should definitely make sure that there is free space in the case for such a device. If you use a case with the current spacious layout, then no questions will arise. Owners of older models with bays for 5.25-inch devices and non-removable hard drive cages should be more careful.
A metal bracket is also supplied with the video card for a reason. The MSI GeForce RTX 2070 Super GAMING X Trio weighs in at just over 1.5kg, so some extra insurance won’t hurt. Here we can also recall the benefits of the metal plating of PCI Express x16 slots.
Heat and Noise Level
ZeroFrozr technology is implemented for the video card, which means that the fans stop when the load is minimal and the GPU heats up. This technique is often used by manufacturers for video cards with a large radiator block, but the MSI GeForce RXT 2070 Super GAMING X Trio was able to surprise here.
The fact is that in the passive silent mode of cooling, the temperature of the GPU on the open bench increased to only 37–38C at 24C indoors. Recall that we are talking about the TU104, and not some ultra-economical entry-level chip. The indicators are very good, the complex design of the cooler with a sweeping radiator is bearing fruit.
Under gaming load, the temperature of the GPU increased to 65C. At the same time, the fan speed did not exceed 1200 rpm. At such speeds, the operation of the cooling system is barely audible against the background of other components of the platform. For example, in our case, with a starting 36 dBA after increasing the speed to the set one, the noise level increased to only 37.5 dBA. The values were obtained using a UNI-T UT353 household sound level meter, placed half a meter from the video card. In office conditions, such a difference is difficult to catch against the background of the fans of the test power supply and the processor cooler.
It remains only to repeat, noting the minimum level of noise emitted. Let’s assume the MSI GeForce RTX 2070 Super GAMING X Trio is one of the quietest GeForce RTX 2070 Super on the market. Perhaps even the quietest.
Returning to the received telemetric indicators, we note that the GPU in games was automatically accelerated to 2010 MHz. This is the peak recorded value, typical frequencies for the chip were ~1930–1970 MHz.
Overclocking
MSI GeForce RTX 2070 Super GAMING X Trio initially has a factory overclocked GPU, but it was also interesting to determine the additional frequency potential of the adapter.
The video card allows you to increase the Power Limit parameter to 111%. Using this opportunity, the base value of the GPU frequency was increased from 1605 MHz to 1700 MHz. At the same time, the average acceleration value increased to 1895 MHz. Good, but not record-breaking. The success of a particular instance of the GPU is still important. Yes, and I would like to be able to increase the Power Limit parameter to 15-20%. However, additional 10-20-30 MHz will be important only for competitive stages, in practice, this difference has almost no effect on performance.
Recall that the actual GPU frequency changes dynamically depending on the nature of the load. GPU Boost automatically speeds up the GPU when the temperature and power consumption of the graphics card is acceptable. Therefore, the actual frequency of the chip can be noticeably higher than the declared average value.
After additional overclocking, the chip accelerated up to 2070 MHz. The TU104 processor often takes the bar at 2000-2050 MHz, and then — as you’re lucky.
Memory chips were also further accelerated. In this case, Micron chips worked stably at an effective 16,000 MHz, instead of the standard 14,000 MHz. The 14.3% increase in memory bandwidth will certainly be useful for heavy graphics modes.
After additional overclocking, the GPU temperature increased at the peak to 67C, while the fan speed grew minimally — up to ~1250 rpm. The level of emitted noise has not changed much. The sound level meter registered ~38 dBA.
Performance 1920×1080
We studied the performance of the GeForce RTX 2070 Super in sufficient detail during the first acquaintance with the adapter of the new line. Super-update significantly expanded the capabilities of the devices in this series. Replacing the GPU from TU106 to TU104 with an increased number of functional blocks and increased operating frequencies affected the number of fps.
Video cards in this class already provide very comfortable performance in Full HD mode with maximum quality settings. Moreover, the GeForce RTX 2070 Super is well suited for owners of gaming monitors with a resolution of 1920×1080 and a screen refresh rate of 144/165 Hz.
Even with maximum graphics quality in heavy projects, the video card can often provide an average of more than 75 frames / s, and in slightly less resource-intensive games — more than 100 frames / s. In this case, the effect of «144 Hz monitor» can already be felt not only in relatively lightweight competitive games.
Comparing the performance with that of other test participants, we can note a strong advantage over the Radeon RX 5700 XT. The difference in the number of frames varies from a vanishingly small 5% to quite noticeable 20-30%. In fairness, it should be noted that initially these are not quite direct competitors. Even the stated recommended price of the model with the Navi 10 chip is $100 lower than the RTX 2070 Super — $400 vs. $500. However, there are still no more productive devices in the AMD range. Radeon VII can be considered as such with a stretch, but the active life cycle of this model is already behind.
At the same time, against the background of the GeForce RTX 2060 Super, which is equal in price, the RX 5700 XT looks quite good. However, now we are still more interested in the capabilities of the GeForce RTX 2070 Super, so we increase the image resolution and evaluate the results.
Performance 2560×1440
Until the advent of improved Super-versions of video cards, the GeForce RTX 2070 could be considered the starting model for games in 1440p mode, but for such conditions it was still very desirable to strengthen the position. As the resolution increases, the load on the GPU increases significantly, and the amount of local memory required by the graphics adapter also increases.
The GeForce RTX 2070 Super generally handles games very well at 2560×1440. The GeForce RTX 2080 has a 5-10% advantage here, but it’s worth considering that in our case we are using a somewhat accelerated model of this series — the Founders Edition.
In many games, even at maximum graphics quality, you can get an average of more than 75 fps. In simpler projects, and 100 fps is not the limit. However, there are a number of games in which, when using ultra settings, video cards like the GeForce RTX 2070 Super barely manage to get the average 60 frames / s. If you want to improve performance, you will either have to look towards the twice as expensive GeForce RTX 2080 Ti, or reduce requests by playing with graphics settings.
After overclocking the MSI GeForce RTX 2070 Super GAMING X Trio, the performance of the graphics card increased by 5-8%. Such acceleration has already allowed the video card to get close to the performance of the GeForce RTX 2080 Founders Edition, which means that it will surely bypass the classic RTX 2080 with the recommended frequency formula. There is an additional increase in fps here, but still it does not fundamentally change the situation.
As an experiment, for the most difficult projects, we used regular presets and profiles with high graphics quality settings, instead of the maximum. After such simple adjustments, the number of frames / s in most cases increased by 15-30%. If you already have a 1440p gaming monitor with a refresh rate of 144/165 Hz, then for some games this acceleration may not seem enough. But such issues are resolved on an individual basis.
For example, in the same Borderlands 3 effect is «volumetric fog» that is unreasonably resource-intensive. If this setting alone is changed from high quality to medium quality, performance will immediately increase from 73/63 to 88/74 fps with minimal visual loss.
Metro Exodus is also quite heavy at the settings with maximum graphics quality, and it’s worth making allowances for the complex scene of the built-in benchmark, with drawdowns that are generally uncharacteristic for a regular game. If you want more here in 1440p mode, then switching to a profile with medium quality will already allow you to get 101/51 fps.
The long-awaited PC version of Red Dead Redemption 2 proved to be quite flexible in terms of graphics settings. Of course, if you basically managed to launch the game and get to the appropriate options. In our case, there were also some starting rough edges, but the presence of a built-in benchmark became an argument for finding solutions to the problematic launch of the game. In Ultra mode, the novelty is very resource-intensive.
For those who want to get a stable 60+ fps at quality settings with a high level of detail, NVIDIA recommends using the GeForce RTX 2060 for 1080p and the GeForce RTX 2070 Super for 1440p. For the same performance in 4K, you will need a GeForce RTX 2080 Ti, but in this case you will have to play around with the options. Fortunately, there are many options available here, so choosing a balanced mode will certainly work. It is important not to limit yourself to the available presets. Advanced guides for the corresponding optimizations are already available on the network.
Rockstar Games is actively working on fixing numerous starting inconsistencies that have seriously spoiled the first impression of the game. For its part, NVIDIA is also trying to improve the situation. The company initially offered a Hotfix version of the driver (GeForce 441.34) to eliminate possible lags for owners of 4/6-core processors when using the Vulkan API. These fixes are also included in the current WHQL version of GeForce 441.41.
Owners of video cards with AMD chips having problems with Red Dead Redemption 2 should also update to the latest current driver version — Adrenalin 19.11.3.
Ray tracing
Game developers are actively trying to work with hybrid rendering, using ray tracing for various tasks. The method is very promising, and for those who get used to the DXR ecosystem earlier, it will be easier to further optimize their projects. So far, NVIDIA has to develop the platform alone. However, AMD also plans to implement hardware-based tracing acceleration in future RDNA 2 graphics chips. Moreover, next-generation consoles should also support ray tracing. The launch of both types of products is expected as early as next year.
So far, only owners of GeForce RTX video cards can appreciate the nuances of hybrid rendering. There are still not as many game projects with tracing as we would like, but they appear. In some games, the effects are better felt, in some weaker. Their implementation requires significant computing resources, therefore, certain losses cannot be dispensed with here.
Many people are probably concerned about performance when tracing is enabled. Here you can often achieve good results with additional settings. Somewhere to refuse the maximum quality and intensity of tracing, somewhere to connect DLSS.