graphics card — PSU for GTX 970 + i7 4790
Ask Question
Asked
Modified
7 years, 3 months ago
Viewed
6k times
For my new PC I’m currently looking at getting an Intel i7 4790 (non «k» version) and a GTX 970 as the main components for the system. It will also be running an SSD, HDD, Network card, and 32GB of RAM. I’ve highlighted the two main components of the system, because, as it stands, the combined TDP of both the CPU and GPU is 684W (Gigabyte on the GPU side). I understand that TDP doesn’t necessarily relate to the actual wattage draw of the component, however a page reviewing several GPUs (including the 970, here) details the 970s TDP at only 145W, which would make the combined TDP of the two main components at 229W — so:
a) Why do manufacturers state the TDP so much higher than what reviews state?
b) I was looking at getting a 750W 80+ Gold Modular PSU from Corsair, but after looking into this further, I contemplated getting a 450W PSU instead. Would I get away with doing so?
- graphics-card
- power-supply
5
Go with the 750 watt power supply. It will be able to handle your system. I had a 500 watt power supply before and had issues with the computer shutting down due to high load from the graphics card when I upgraded to an i7 CPU and 970GTX graphics card.
A) The TDP that comes from the manufacturer is measured at peak performance with a higher than normal ambient temperature. That means when they measure the TDP, the graphics card has probably been running at or close to 100% for a few minutes. This is hardly ever the case when gaming and you might see your graphics card average 40-60% usage during your gaming. This will obviously vary depending on what games you play. Same works for the CPU. The TDP is measured at 100% usage. If your CPU hits 100% usage then something is wrong.
B) Yes, a 750w power supply would be overkill. A 600-660w power supply would be enough. That being said I would still get the 750w power supply because it’s a corsair (great brand) and because it will help you future proof your machine.
450W is plenty for this combination if you don’t plan to add another GPU or a large RAID-Array in the future. I have recently built a very similar setup with a i7-4790k, a GTX-970, 32 GB DDR-3 2400 RAM, a m2 SSD and 6 fans on a ASUS Z97-Deluxe Board. The CPU is overclocked to 4x 4.5 GHz (~1.25V) and I run it with a 450W 80+ Gold PSU (the cute little SFX device from Silverstone). I measured the system with a power meter and it draws up to 430W from the wall socket when I run Prime64 and Furmark simultaneously. With full CPU load alone it needs approximately 280W. I think a non overclocked CPU might need at least 70W less. You should calculate with at least 10W per mechanical hard disk or optical drive and at least 5W for every extra SSD or fan.
The power rating of a PSU refers to a combined power scenario of all voltages it can deliver. This may not be the combination that your system needs. A typical system with current components consumes almost all it’s power from the 12V supply, so you should look for a PSU that can deliver a sufficiently high output at 12V alone already. I.e. my PSU is rated at 36A on 12V, so it will be fine at least up to 480W when measured at the wall socket (36A*12V*0.90). The 0.90 factor accounts for the 90% efficiency at full load.
Some PSUs still use a multi-rail 12V design which can be problematic as it drastically limits the 12V supply to individual components, and I would recommend to look for a single rail PSU instead. Single rail PSUs have only one 12V power rating, i.e. 36A, while multi-rail PSUs are usually rated as 18A+18A or 2x18A or similar. I actually tried to set up this board and the CPU with a older 750W quad-rail PSU (4x14A) at first and I didn’t get it stable with overclocking while it was no problem when I connected it to the 450W single rail PSU.
I would recommend not to use a much more powerful than needed PSU if you already know that you will not even need that much power in the future. Not only are lower rated PSUs typically less expensive but they are actually more efficient at typical idle loads since the 80+ rating doesn’t require a minimum efficiency below 20% load and thus the efficiency far below 20% is usually much worse than what you would expect with the given 80+ rating even for high quality PSUs. This means that a comparatively cheap 450W 80+ Bronze PSU may well draw less power from the wall socket than a much more expensive 750W Gold PSU whenever the computer is not really busy (i.e. when you only use the browser or watch a movie which should translate to 50..80W for your hardware).
2
Please take into consideration that a power supply never delivers his main number in watts. It will be always a lower when 100% use. A 80 GOLD certified or something similar, guarantees that the power supply will deliver 80% of his number as a final wats. So If you go for a 80 Gold 450W power supply, you can safely rely on a maximum of 360W of pure power cranking out the power supply.
Since 100% use of both GPU and CPU is very very VERY unlikely, I would say you can safely run both in a 450w. Hoever I do have my PC with a 4790K and no GPU, but when I had one it was a 960GTX and I had no issues.
PCpartpicker.com let you build your system and it will give you a reference watts usage comsumption for better measuring.
I would say no one who doesnt use SLI or a VERY beffy GPU will ever need more than 450 — 500. And wattage gets lower and lower each day.
Do take into consideration that Power Supplies are designed to deliver their main number in watts. If you put a 600w supply into your system and you are using 100w most of the time, you are wasting a ton of electricity because power s. don’t know how to manage loads, they are either working or not working. The only thing that changes are the electricity requirements of your system components if they are under load or not.
further analysis:
https://www.youtube.com/watch?v=LiWThqgFfI4
1
Gigabyte G1 Gaming GeForce GTX 970 4GB Review (Page 12 of 13)
Page 12 — Power Usage, Temperature, Noise
Using a wall power measurement device, I noted down the total system power consumption with the Gigabyte G1 Gaming GeForce GTX 970 4GB installed. Do keep in mind this includes every component of the system — including power supply efficiency loss — and not just the graphics card only. Also, the load conditions cannot be directly compared against this base value, since the entire system is under load, and as such the CPU will also contribute to the increased power usage to some extent. With that in mind, the Gigabyte G1 Gaming GeForce GTX 970 4GB’s power consumption reached a maximum of 314W load from 74W idle — a difference of 240W. Gigabyte likes to promote their lower RDS(on) MOSFETs, quality solid capacitors, 2oz copper PCB, and low power loss ferrite core design. In the past, I have seen a difference in power consumption as much as 10W against reference cards with regular components. The above results were obtained when running 3DMark’s Fire Strike Extreme test, and your graphics card is not likely to see any higher loads, especially under normal usage. To be honest, you are not going to get anywhere near this under intense gaming sessions, so really — this is just for interest’s sake, haha.
For the purpose of this review, I left the fan on default settings, so its speed is allowed to vary accordingly with temperature. I also wanted to see what the card is capable of doing inside my low airflow chassis configuration. Most people should get better results in real life than our hot running test bench environment. With that in mind, I left the stock paste intact for testing before taking it apart for the photo session on Page 3. Its thermal interface material was applied properly from the factory, which we have seen during disassembly. Even under our intense Furmark load tests, the Gigabyte G1 Gaming GeForce GTX 970 4GB peaked out at only 58c; a figure that is simply amazing, thanks to the 600W WindForce 3X cooler and the GM204’s low TDP. It is important to point out this is the worst case scenario — you will not hit this temperature under normal gaming sessions. Most cases should have better airflow than my configuration anyway.
As far as noise is concerned, while this is very subjective, I am quite a picky person on noise, and the loudest component in my entire system is probably my Noctua ultra low noise fans — and they are not loud at all. In my opinion, there is no objective measurement of noise, as measuring sound pressure level is often impractical, because human ears are more sensitive to some frequencies than others. On a scale from 0-10 where 0 is silent and 10 is the loudest, I would rate the Gigabyte G1 Gaming GeForce GTX 970 4GB to be at 3.0/10 at 34% fan speed. 34% is the default fan speed during idle, and this is also the lowest you can set it to. It does not produce a significant amount of noise until it goes above 50%, but it will be nice if it can be tuned to around 20-25% to appeal to silent PC enthusiasts. The fan will kick up to at full load, unless set otherwise. On a scale from 0-10, where 0 is silent and 10 is the loudest, I would rate the Gigabyte G1 Gaming GeForce GTX 970 4GB to be at 7.0/10 at 100% fan speed. Seriously, it is pretty loud at that speed — but at least the fan runs smoothly. If you want your card to be appropriate in a quiet PC configuration, keep the fan running at minimum. Once you bump it past that magic mark, you are going to hear it quite clearly. That said, as with all Gigabyte WindForce cards I have used in the past, when you are not gaming, it is pretty darn quiet. The G1 Gaming GTX 970 is a little louder than the Gigabyte GeForce GTX 760 2GB at idle, but I am just being picky here.
Page Index
1. Introduction, Packaging, Specifications
2. NVIDIA GeForce GTX 970 Architecture
3. A Closer Look, Installation, Test System
4. Benchmark: 3DMark
5. Benchmark: Battlefield 4
6. Benchmark: BioShock Infinite
7. Benchmark: Crysis 3
8. Benchmark: GRID 2
9. Benchmark: Metro: Last Light
10. Benchmark: Thief
11. Benchmark: Unigine: Heaven 4.0
12. Power Usage, Temperature, Noise
13. Overclocking and Conclusion
NVIDIA GeForce GTX 970
→
Iron →
Video card →
Graphics Card Performance and Specifications → NVIDIA GeForce GTX 970
General Information | |
Developer | NVIDIA |
Released | 2014 |
Video card category | Desktop |
Video card type | Discrete |
Interface | PCIe 3. 0 |
Maximum resolution | 4096 x 2160 |
GPU | |
Architecture | Maxwell 2.0 |
Core | GM204 |
Number of chips 10 Process technology | 28 nm |
Number of transistors | 5200 million |
Area | 398 mm² |
Core clock | 1050 — 1178 MHz |
Universal shader units | 1664 |
56 | |
TMUs | 104 |
Pixel fillrate | 58.8 GPixel/s |
Texture fillrate (texel fillrate) | 109.0 GTexel/s |
Video memory | |
Type | GDDR5 |
Volume | 4096 Mb |
Frequency | 7012 MHz |
Bus width | 256 bit |
7 Max. power consumption (TDP) | 145 W |
Max. allowable temperature | 98° C |
Min. power supply requirements | 500 W |
Auxiliary power connectors | 6-pin + 6-pin supported APIs and technologies |
DirectX | 12 |
OpenGL | 4.5 |
OpenCL | 1.2 |
Shader Model | 5.0 |
SLI / CrossFireX | SLI |
Other technologies | • NVIDIA G-Sync • NVIDIA Adaptive V-Sync • NVIDIA GameStream • NVIDIA GPU Boost • HDCP • NVIDIA CUDA • NVIDIA PhysX • NVIDIA Surround • NVIDIA 3D Vision • FXAA • TXAA • MFAA • NVIDIA VXGI • NVIDIA DSR |
Compare
GeForce GTX 9 70
with other graphics card
( ~ 700 models )
Game performance
GeForce GTX 970
Video card rating
( + specifications )
Processor comparison service
( ~ 3000 models)
Processor rating
(+ specs)
This site uses cookies to ensure you get the best experience. If you continue to use this site, we will assume that you agree to this.
OK
More details
Compare AMD Radeon RX 5500 and NVIDIA GeForce GTX 970
Comparative analysis of AMD Radeon RX 5500 and NVIDIA GeForce GTX 970 video cards by all known characteristics in the categories: General information, Specifications, Compatibility, dimensions, requirements, Memory, Technology support, Video outputs and ports, API support.
Analysis of video card performance by benchmarks: PassMark — G3D Mark, PassMark — G2D Mark, Geekbench — OpenCL, GFXBench 4.0 — Car Chase Offscreen (Frames), GFXBench 4.0 — Manhattan (Frames), GFXBench 4.0 — T-Rex (Frames), GFXBench 4.0 — Car Chase Offscreen (Fps), GFXBench 4.0 — Manhattan (Fps), GFXBench 4.0 — T-Rex (Fps), 3DMark Fire Strike — Graphics Score, CompuBench 1.5 Desktop — Face Detection (mPixels/s), CompuBench 1.5 Desktop — Ocean Surface Simulation (Frames/s), CompuBench 1.5 Desktop — T-Rex (Frames/s), CompuBench 1. 5 Desktop — Video Composition (Frames/s), CompuBench 1.5 Desktop — Bitcoin Mining (mHash/s).
AMD Radeon RX 5500
versus
NVIDIA GeForce GTX 970
Benefits
Reasons to choose AMD Radeon RX 5500
- Newer graphics card, release date difference 5 year(s) 0 month(s) 903 02
- approx. 59% faster core clock: 1670 MHz vs 1050 MHz
- Boost core clock 57% faster: 1845 MHz vs 1178 MHz
- Texturing speed 49% faster: 162.36 GT/s vs 109vs 11499
- About 1% more performance in GFXBench 4.0 — Manhattan (Fps) benchmark: 3717 vs 3698
- About 32% more performance in 3DMark Fire Strike — Graphics Score: 4813 vs 3637
- About 1% less power consumption: 148 Watt vs 150 Watt
- About 9% more PassMark — G3D Mark performance: 9640 vs 8851 9 0302
903 23
Date release | 7 Oct 2019 vs 19 September 2014 |
Core frequency | 1670 MHz vs 1050 MHz |
Boost core clock | 1845 MHz vs 1178 MHz |
Texturing speed | 162. 36 GT/s vs 109 billion/sec |
PassMark — G2D Mark | 784 vs 779 |
Geekbench — OpenCL | 41198 vs 24105 |
GFXBench 4.0 — Car Chase Offscreen (Frames) | 12069 vs 11499 |
GFXBench 4.0 — Manhattan (Frames) | 3717 vs 3698 |
GFXBench 4.0 — T-Rex (Frames) | 3351 vs 3340 |
GFXBench 4.0 — Car Chase Offscreen (Fps) | 12069 vs 11499 |
GFXBench 4.0 — Manhattan (Fps) | 3717 vs 3698 |
GFXBench 4. 0 — T-Rex (Fps) | 3351 vs 3340 |
3DMark Fire Strike — Graphics Score | 4813 vs 3637 |
Reasons to choose NVIDIA GeForce GTX 970
Energy consumption (TDP) | 148 Watt vs 150 Watt |
PassMark — G3D Mark | 9640 vs 8851 |
Benchmark comparison
GPU 1: AMD Radeon RX 5500
GPU 2: NVIDIA GeForce GTX 970
PassMark — G3D Mark |
|
|||
PassMark — G2D Mark |
|
|||
Geekbench — OpenCL |
|
|||
GFXBench 4. 0 — Car Chase Offscreen (Frames) |
|
|||
GFXBench 4.0 — Manhattan (Frames) |
|
|||
GFXBench 4.0 — T-Rex (Frames) |
|
|||
GFXBench 4.0 — Car Chase Offscreen (Fps) |
|
|||
GFXBench 4. 0 — Manhattan (Fps) |
|
|||
GFXBench 4.0 — T-Rex (Fps) |
|
|||
3DMark Fire Strike — Graphics Score |
|
Name | AMD Radeon RX 5500 | NVIDIA GeForce GTX 970 |
---|---|---|
PassMark — G3D Mark | 8851 | 9640 |
PassMark — G2D Mark | 784 | 779 |
Geekbench — OpenCL | 41198 | 24105 |
GFXBench 4. 0 — Car Chase Offscreen (Frames) | 12069 | 11499 |
GFXBench 4.0 — Manhattan (Frames) | 3717 | 3698 |
GFXBench 4.0 — T-Rex (Frames) | 3351 | 3340 |
GFXBench 4.0 — Car Chase Offscreen (Fps) | 12069 | 11499 |
GFXBench 4.0 — Manhattan (Fps) | 3717 | 3698 |
GFXBench 4.0 — T-Rex (Fps) | 3351 | 3340 |
3DMark Fire Strike — Graphics Score | 4813 | 3637 |
CompuBench 1. 5 Desktop — Face Detection (mPixels/s) | 105.107 | |
CompuBench 1.5 Desktop — Ocean Surface Simulation (Frames/s) | 1225.96 | |
CompuBench 1.5 Desktop — T-Rex (Frames/s) | 8.737 | |
CompuBench 1.5 Desktop — Video Composition (Frames/s) | 35.714 | |
CompuBench 1.5 Desktop — Bitcoin Mining (mHash/s) | 489.635 |
Feature comparison
AMD Radeon RX 5500 | NVIDIA GeForce GTX 970 | |
---|---|---|
Architecture | RDNA | Maxwell 2. 0 |
Issue date | Oct 7, 2019 | 19 September 2014 |
Place in the ranking | 223 | 327 |
Type | Desktop | Desktop |
Codename | GM204 | |
Price at first issue date | $329 | |
Price now | $407.76 | |
Price/performance ratio (0-100) | 28.59 | |
Boost core clock | 1845 MHz | 1178MHz |
Number of Compute | 22 | |
Core frequency | 1670 MHz | 1050MHz |
Peak Half Precision (FP16) Performance | 10. 4 TFLOPs | |
Peak Single Precision (FP32) Performance | 5.2 TFLOPs | |
Pixel fill rate | 59GP/s | |
Stream Processors | 1408 | |
Texturing speed | 162.36 GT/s | 109 billion / sec |
Power consumption (TDP) | 150 Watt | 148 Watt |
Number of transistors | 6400 million | 5,200 million |
Number of CUDA conveyors | 1664 | |
Floating point performance | 3. 920 gflops | |
Process | 28nm | |
Maximum temperature | 98 °C | |
Number of shaders | 1664 | |
Recommended power supply | 550 Watt | 500 Watt |
Additional power connectors | 1 x 8-pin | 2x 6-pins |
Tire | PCI Express 3.0 | |
Height | 4.376″ (11. 1 cm) | |
Interface | PCIe 3.0 x16 | |
Length | 10.5″ (26.7cm) | |
SLI support | 4x | |
Maximum memory size | 4GB | 4GB |
Memory bandwidth | 224 GB/s | 224GB/s |
Memory bus width | 128bit | 256 Bit |
Memory type | GDDR6 | GDDR5 |
Memory frequency | 7. 0 GB/s | |
Shared memory | 0 | |
4K h364 Decode | ||
4K h364 Encode | ||
h365/HEVC Decode | ||
h365/HEVC Encode | ||
HDMI 4K Support | ||
3D Vision | ||
Adaptive Vertical Sync | ||
CUDA | ||
GameStream | ||
GameWorks | ||
GeForce Experience | ||
GeForce ShadowPlay | ||
GPU Boost | ||
SLI | ||
Surround | ||
Audio input for HDMI | Internal | |
Video connectors | 1x DVI, 1x HDMI, 3x DisplayPort, Dual Link DVI-I, HDMI 2.
|