1080 vs titan x: Nvidia GeForce GTX 1080 vs Nvidia GeForce GTX Titan X: What is the difference?

59

max 153

Average: 49.5

max 153

Average: 49.5

SPECviewperf 12 test score — specvp12 creo-01

59

max 153

Average: 52.5

max 153

Average: 52.5

SPECviewperf 12 test score — specvp12 catia-04

103

max 189

Average: 91. 5

max 189

Average: 91.5

SPECviewperf 12 evaluation — Catia

103

max 189

Average: 88.6

max 189

Average: 88.6

SPECviewperf 12 test score — specvp12 3dsmax-05

145

max 316

Average: 189.5

max 316

Average: 189.5

SPECviewperf 12 test score — 3ds Max

143

max 269

Average: 169.8

max 269

Average: 169.8

GeForce GTX 1080 Ti video card preview

Traditionally, TITAN X series video cards have higher performance compared to GTX series devices. However, this rule is not typical for the new GeForce GTX 1080 Ti: this video card practically leveled the capabilities of the ultimate high-end TITAN X models with devices of a more modest, household GTX series.

Graphics capabilities of new products from NVIDIA

The heart of the GeForce GTX 1080 Ti is a slightly simplified version of the GP102 GPU, which is also used in devices of the GeForce TITAN X line. It retains the same number of CUDA units and texture units (3584 and 224, respectively) , but the number of used rasterization units is reduced to 88 (in the TITAN X model, their number is 96).

Another difference between the novelty and the premium model is the use of a non-standard memory bus width. It is 352 bits (in TITAN X this figure is 384 bits). The amount of pre-installed RAM has also been reduced from 12 to 11 GB, but at the same time, the GDDR5X frequency has been increased from 10,008 MHz to 11,008 MHz. And given the increase in the operating frequency of both the GPU and the installed memory chips, the new line of video cards will not be inferior in performance to the higher-end GeForce TITAN X model. And the significantly lower cost of the GTX 1080 Ti video card will certainly make it a top solution for home users.

As usual, the models of the so-called reference design were the first to be proposed, so let’s start our acquaintance with the video card with the reference version of the GTX 1080 Ti Founders Edition. The video card is packaged in a beautifully designed medium-sized box. The package includes, in addition to the video card itself, a DisplayPort to DVI video output adapter and a paper express manual. The external design remained the same as in the previous well-known reference models from NVIDIA: the same top cover made of aluminum, with chopped, high-quality edges, and the cooling plates are clearly visible through a transparent window on the top panel.

The heat-removing element of the base is a rather massive evaporator chamber, and air is blown and the radiator is blown through by a single centrifugal fan. The advantage of this constructive solution is that almost all the hot air leaves the «system unit». This powerful video adapter option is considered the best solution for compact or poorly ventilated cases. Optimized with a larger evaporator volume and usable internal interface panel, the cooler is more efficient than the TITAN X cooler.

The GTX 1080 Ti has four video outputs (one HDMI 2.0b and three DisplayPort 1.4 outputs). The developers placed all interface connectors in one tier. To save the area of ​​the panel, the DVI-D connector is not placed on the “second floor”, and if necessary, the monitor is connected via the included DisplayPort to DVI adapter. Thanks to this layout of the connectors, it was possible to significantly increase the cooling area of ​​​​the ventilation grille (it occupies about ⅔ of the mounting plate), simplify heat dissipation and increase the overall efficiency of the cooling system.

For the GTX 1080 Ti, the manufacturer used the same circuit board as for the GeForce TITAN X, supplemented by a GPU power stabilization subsystem. Due to the fact that the VRM module includes seven dual phases (dualFET paired chips), not only the required power is achieved, but also the efficiency of the converter is increased. The graphics card uses Micron’s GDDR5X memory chips that are arranged around the GPU case in a unique U-shape. The GeForce GTX 1080 Ti has 11 chips of 1 GB each. All of them receive additional cooling from a massive bracket of a common cooler through special heat-conducting pads.

To connect an additional 12 volt power supply to the device, two special connectors are installed in the video adapter: one for 8, the other for 6 pins. Both are located on the top edge of the adapter. According to the manufacturer, the value of the thermal package of the new video card (TDP) is about 250 watts. The TITAN X models are of the same importance, as both cards are based on identical GPUs, including 12 billion transistors. For the previous model GeForce GTX 1080, made on a simpler GP104 chip, the TDP value is up to 180 watts. But the increase in energy consumption is justified by the increased productivity. As NVIDIA recommends, when using systems with a GTX 1080 Ti graphics card, you must have a power supply with a power of at least 600 watts.

To prevent damage to the PCB surface, the manufacturer covered the reverse side of the main printed circuit board with a massive plate. Thanks to the composite design of the protective plate, part of the metal plate can be easily removed, increasing the free gap between the components of the devices, which improves the flow of cold air into the cooling system of the second video card or other device in the adjacent slot.

The top panel of the video adapter is decorated with the GeForce GTX logo, complemented by the company’s traditional green backlight, which turns on when the system is running. The length of the video card does not differ from the dimensions of other top models from NVIDIA — 267 mm.

First impressions of the GeForce GTX 1080 Ti

At rest (with the case cover open), the GPU temperature did not exceed 32 degrees. At the same time, the cooler rotation speed was about 1100 rpm. Despite the fact that the installed cooling system does not support hybrid mode, and the fan is constantly rotating, with a minimum graphic load of the system, the device remains as quiet as possible. But with an increase in system load, the temperature on the processor case increased to 84 degrees, and the fan speeded up to 2400 rpm. In peak modes, the noise level of the video adapter can be estimated as average.

With built-in GPU Boost 3.0, the internal frequency of the GP102 graphics chip in the reference GTX 1080 Ti constantly changes in response to the chip’s power consumption and current temperature. When running some games, the GPU frequency reaches 1800-1850 MHz, but in other operating situations it can drop to 1700 MHz or increase to 1886 MHz (at peak load).

Gaming performance

When experimenting with overclocking this GPU, its speed under real gaming load reached 1923-1974 MHz, and the actual frequency tended to 2000 MHz. However, the system of dynamic auto-acceleration does not allow to achieve stable operation at this frequency. When operating in this mode, heating and power consumption increase significantly: the internal temperature of the chip rises to 85 degrees, and the cooling fan speed exceeds 3000 rpm.

Conclusions

Contrary to the established practice of NVIDIA when introducing new models of top GPUs in previous generations of video adapters, this time the developer did everything possible to ensure that the GeForce GTX 1080 Ti was not inferior in performance to its main competitor TITAN X, and also received a number of additional benefits. Despite some limitations in the set of functional blocks of the GPU and the absence of one of the RAM controllers, eight ROP chips and one of the 12 GB of RAM, this practically does not affect the performance of the video card in games and is compensated by a slight increase in the GPU clock speed and the use of new chips GDDR5X, with an effective frequency of 11 GHz.

These minor specifications differences are more likely to reduce the cost of the device than to differentiate the GeForce GTX 1080 Ti model from its older sister TITAN X. In addition, the performance of the GTX 1080 Ti is also affected by improved «physical» parameters of the video adapter: higher efficiency cooling systems and power supply efficiency. To date, the GTX 1080 Ti is the first GeForce device to be fully capable of gaming at 4K resolution. And the relatively low price makes it the most cost-effective solution compared to any multi-processor configurations.

Choose and buy your graphics card from the technical experts!

Buy a video card for a computer in the online store

NVIDIA GeForce

ASUS

graphics cards ZOTAC 9000

Contents

1. Introduction
2. Features
3. Tests
4. Games
5. Key differences
6. Conclusion
7. Comments

Video card

Video card

Introduction

We compared two graphics cards: NVIDIA TITAN Xp vs NVIDIA GeForce GTX 1080 Ti. On this page, you will learn about the key differences between them, as well as which one is the best in terms of features and performance.

NVIDIA TITAN Xp is a Pascal-based GeForce 10 generation graphics card released Apr 6th, 2017. It comes with 12GB of GDDR5X memory running at 1426MHz, has 1x 6-pin + 1x 8-pin power connector and consumes up to 165 W.

The NVIDIA GeForce GTX 1080 Ti is a Pascal-based GeForce 10 generation graphics card released on Mar 10th, 2017. It comes with 11GB of GDDR5X memory running at 1376MHz, has a 1x 6-pin + 1x 8-pin power connector and consumes up to 165 watts.

354.4 GFLOPS (1:32)

CLOCK Speeds

Basic frequency

1405 MHz

1481 MHZ

Frequency

1582 MHZ

9000 Render Config

SHADING Units

3840

3584

Texture Units

240

224

Raster Units

9000

88

SM Count

30

28

Graphics Features

DirectX

12 (12_1)

12 (12_1)

OpenGL

4. 6

4.6

OpenCL

3.0

3.0

Cuda

6.1

6.1

VULKAN

1.2

1.2

Board Design

Heating0004 1x 6-pin + 1x 8-pin

Slot width

Dual-slot

Dual-slot

Benchmarks

3DMark Graphics

3DMark is a benchmarking tool designed and developed by UL to measure the performance of computer hardware. Upon completion, the program gives a score, where a higher value indicates better performance.

NVIDIA TITAN XP
+4%

NVIDIA GeForce GTX 1080 Ti

Blender bmw27

Blender is the most popular 3D content creation software. It has its own test, which is widely used to determine the rendering speed of processors and video cards. We chose the bmw27 scene. The result of the test is the time taken to render the given scene.

NVIDIA TITAN Xp

NVIDIA GeForce GTX 1080 Ti
+1%

Th200 RP

Th200 RP is a test created by Th200. It measures the raw power of the components and gives a score, with a higher value indicating better performance.

NVIDIA TITAN XP
+10%

NVIDIA GeForce GTX 1080 Ti

Games

1920×1080, Ultra

Game	TITAN Xp	GeForce GTX 1080 Ti
Anno 1800
Assassin’s Creed Odyssey
Battlefield V
DOOM Eternal
Far Cry 5
Hitman 2
Metro Exodus
Red Dead Redemption 2
Shadow of the Tomb Raider
The Witcher 3
Average	0. 00 fps	110.04 fps

2560×1440, Ultra

Game	TITAN Xp	GeForce GTX 1080 Ti
Anno 1800
Assassin’s Creed Odyssey
Battlefield V
DOOM Eternal
Far Cry 5
Hitman 2
Metro Exodus
Red Dead Redemption 2
Shadow of the Tomb Raider
The Witcher 3
Average	0. 00 fps	84.40 fps

3840×2160, Ultra

Game	TITAN Xp	GeForce GTX 1080 Ti
Anno 1800
Assassin’s Creed Odyssey
Battlefield V
DOOM Eternal
Far Cry 5
Hitman 2
Metro Exodus
Red Dead Redemption 2
Shadow of the Tomb Raider
Average	0. 00 fps	47.91 fps

Key differences

Why is NVIDIA TITAN Xp better than NVIDIA GeForce GTX 1080 Ti?

Newer — released 27 days later

Has 10% better performance

Has 1 GB more memory

Has a slightly larger memory bus

Has 14% more bandwidth — 547.6 GB/s vs 484.4 GB /s

Has a 10% higher pixel fill rate — 151.9 GPixel/s vs 139.2 GPixel/s

Has an 8% higher texture fill rate — 379.7 GTexel/s vs 354.4 GTexel/s

Has more shading units +16

Has 256 more shading units

Conclusion

Which is better TITAN Xp or GeForce GTX 1080 Ti?

The TITAN Xp delivers 10% better performance, consumes up to 0% less energy and holds 1 GB more memory. According to our research, TITAN Xp 3990X is more powerful than GeForce GTX 1080 Ti.

What AMD graphics cards is equivalent to TITAN Xp?

The Radeon RX 6700 XT is AMD’s closest competitor to the TITAN Xp. It is 1% more powerful, uses 8% less energy, and holds the same amount of memory.

What NVIDIA graphics cards is equivalent to TITAN Xp?

The GeForce RTX 2080 Ti is NVIDIA’s closest competitor to the TITAN Xp. It is 3% more powerful, uses the same amount of energy, and holds 1 GB more memory.

How does TITAN Xp perform compared to GeForce GTX 1080 Ti and other graphics cards?

Relative performance

General performance

Bench performance

Testing GeForce GTX 1080 Ti Founders Edition vs. GeForce GTX 1080

Saturday, 20 May 2017

• 1
• 2
• 3
• 4
• 5

(13 votes)

Video cards

Author

Alexander

About a year ago, NVIDIA introduced older gaming solutions based on the new Pascal architecture, paving the way for a new generation. It’s time to update the GeForce GTX lineup with a new flagship product. Initially, two versions of the new TITAN were presented, which are produced in small quantities at a high price. And now a new mainstream champion is entering the arena in the face of the GeForce GTX 1080 Ti, seriously raising the performance bar relative to its predecessors. Everything about the architecture, technological features and performance of the new flagship in this article. Consider the standard reference version of the GeForce GTX 1080 Ti Founders Edition, check it for overclocking and compare it with the GeForce GTX 1080.

The GeForce GTX 1080 Ti is based on the GP102 GPU, which consists of 6 GPC clusters instead of 4 clusters on the GP104 (GeForce GTX 1080). And in general, GP102 resembles the first chip of the Pascal generation represented by GP100. But if the latter was used strictly for NVIDIA Tesla specialized computing devices, then GP102 is already an optimized chip for gaming solutions. Within the Pascal architecture, the cluster structure is identical, and one GPC includes 5 TPC blocks from the SM multiprocessor array and the PolyMorph Engine geometry processing block. The SM has 128 CUDA cores, which is the minimum computing unit. In total GP102 can handle 3840 CUDA cores. Also, the total potential of the core includes 240 TMU texture units and 96 ROPs. 12 memory controllers form a common 384-bit bus.

The first GP102-based graphics accelerator was the TITAN X, and the GeForce GTX 1080 Ti can be called its successor. Both products use an incomplete configuration of computing units. The GeForce GTX 1080 Ti has 3584 active CUDA processors and 224 texture units. The memory bus has also been partially truncated — its capacity is 352 bits. The amount of memory reaches 11 GB instead of 12 GB for TITAN X. Probably, the production of massive GPUs at first was associated with certain difficulties. Therefore, a full-fledged GP102 saw the light of day in the form of real products quite recently — this is an updated TITAN Xp with a full configuration of computing units and a slight increase in frequencies. And the GeForce GTX 1080 Ti demonstrates exactly the same frequencies: for the core, the base level is 1480 MHz with an average Boost Clock of 1582 MHz, and the effective memory frequency is at 11 GHz.

For clarity, we have given the characteristics of all three video cards based on the GP102 processor in one table.

All graphics cards demonstrate performance in about the same class. TITAN Xp is by far the fastest gaming video accelerator, but the GeForce GTX 1080 Ti will be slightly behind. And if we talk about TITAN X, then the GeForce GTX 1080 Ti even turns out to be a little faster due to the increased frequencies.

In comparison with the GeForce GTX 1080, the progress is very serious, because the old flagship operated only 2560 CUDA cores with a 256-bit bus. The GeForce GTX 1080 Ti has 40% more compute units and about 51% more memory bandwidth. L2 cache increased to 2.8 MB instead of 2 MB for GP104. The high memory capacity of 11 GB is a record for mass gaming devices. The only simplification regarding the GeForce GTX 1080 can be called lower frequencies. However, the experience of a real comparison will show that in the game load, their average Boost is approximately at the same level.

The GP102 has 12 billion transistors instead of 7.2 billion for the GP104. Therefore, the level of heat dissipation and energy consumption has increased. At the same time, the thermal package TDP is at the level of 250 W, which is usual for old NVIDIA flagships. The GeForce GTX 980 Ti with the GM200 processor (8 billion transistors) showed the same TDP as the older flagship GeForce GTX 780 Ti. New chips are being produced according to the already familiar 16-nm process technology.

NVIDIA estimates the performance difference between the GeForce GTX 1080 Ti and GeForce GTX 1080 at 35%, which is higher than the difference between similar products of the older series in previous generations. You can directly compare the characteristics of the GeForce GTX 1080 Ti with its predecessors in the table below.

GeForce GTX 1080 Ti works with the fastest GDDR5X memory. To advance to the record-breaking 11GHz, NVIDIA engineers worked closely with memory manufacturers to improve signal quality and minimize electronic noise. Therefore, the memory subsystem has undergone some refinement and improvement. And although such memory frequencies during overclocking are achievable on the old GeForce GTX 1080, one can also expect a general increase in the frequency potential of GDDR5X from the flagship solution. A practical acquaintance will show that the maximum memory overclocking bar for the GeForce GTX 1080 Ti is really higher. Combined with a wide 352-bit bus, this results in a throughput of 484 GB/s. It should be noted that the only game card with progressive HBM memory, the AMD Radeon R9Fury X shows a throughput of 512 GB / s, that is, the difference is minimal. At the same time, the Pascal architecture uses progressive methods of color data compression, allowing to reduce the amount of transmitted data and increase the efficiency of memory use. More about this was described in our detailed material on the technological capabilities of Pascal.

Pascal and Maxwell solutions also support Tiled Caching based on a modified tile rendering method. The latter proposes to rasterize the image when splitting it into fragments (tiles), which in some cases allows more efficient use of resources. Tile caching partially uses this approach in combination with standard rendering. This reduces access to external memory, increasing the load on the fast L2 cache.

The combination of data compression and tile caching can significantly increase the efficiency of data transfer, bringing the throughput to a theoretical 1.2 TB / s. In reality, the effectiveness of all improvements depends on the software support in each application, so the declared value is very conditional. While it’s clear that the GeForce GTX 1080 Ti graphics card is the fastest gaming solution that only TITAN X can compete with in terms of memory.

There are several graphics APIs currently in use in the industry. DirectX 11 or OpenGL remains the main one for many games, but more and more new projects are being developed under DirectX 12 and Vulkan. The former refer to high-level APIs, where optimization largely depends on the video driver. The latter are low-level, which give more opportunities to the application itself, partially limiting optimization at the driver level.

The GeForce GTX 1080 Ti is ready to work efficiently with all APIs. In the case of DirectX 11, we have years of software optimization and refinement of NVIDIA’s GPU architecture. Therefore, the current GeForce GTX models demonstrate the best results in DirectX 11. In the case of DirectX 12, everything is still ambiguous. The first games showed a steady drop in performance compared to DirectX 11. But the gradual optimization of applications and drivers has improved the situation. The situation has become familiar when several months or even six months pass, and only then the game starts to work faster in DirectX 12. The latest versions of the GeForce video driver brought good acceleration. And the next slide from NVIDIA shows that as of March of this year, some games have received a speedup of 4% to 30% when compared to the performance at the time of release. Among the most notable examples are Hitman, Rise of the Tomb Raider. Deus Ex: Mankind Divided was not included in the list, but even there it was already possible to achieve higher fps in DirectX 12 compared to the old DirectX 11.

DirectX 12 development process is more resource intensive. It is important to have quality tools for debugging applications. Microsoft recently introduced a dedicated PIX utility for this purpose. NVIDIA will support developers with the release of nSight Visual Studio Edition 5. 3 and NVIDIA Aftermath for debugging and troubleshooting.

NVIDIA is also expanding the GameWorks toolset to implement various graphical effects with DirectX 12 support. lighting. An example is Fallout 4, which uses HBAO+, Volumetric Lighting, and shard effects powered by NVIDIA Flex. All this is discussed in detail in our special review.

Moving GameWorks to the DirectX 12 platform gives more freedom to developers, and also improves the implementation of the effects themselves, using the capabilities of the new API. Under DirectX 12, NVIDIA HairWorks technology has already been adapted, which allows you to visualize hair and wool with the study of individual hairs and reliable physics of such hairline. Dynamic fur and hair are familiar to players from The Witcher 3: Wild Hunt.

New GameWorks technologies include Turf Effects for rendering complex grass with more realistic physics and soft shading. This technology has been used in Tom Clancy’s Ghost Recon: Wildlands. This game runs under DirectX 11, but the technology is also compatible with DirectX 12.

Much attention is paid to the simulation of dynamic environments and destructibility. To implement such effects, there is a whole package of GameWorks tools. For example, Flex allows you to simulate the behavior of loose and liquid media, WaveWorks recreates dynamic waves on the surface of the water, Blast simplifies the implementation of individual destructible objects.

This list also includes NVIDIA GameWorks Flow for simulating gases and fire. An example implementation using DirectX 12 is shown in the following video:

Moving to DirectX 12 allows you to take advantage of new features, such as Async Compute. This speeds up the calculations for visualizing Flex and Flow effects. As an example, data is given on a Flex-based particle simulation using a GeForce GTX 1080 graphics card: in the DirectX 12 environment, calculations are performed almost twice as fast compared to DirectX 11. . GameWorks support is implemented in a number of major engines, including the latest versions of Unreal Engine and Unity, as well as the promising open source engine Amazon Lumberyard. GameWorks under DirectX 12 works through the DirectCompute API, so the effects must be supported not only by GeForce GTX, but also by AMD Radeon.

The direction of virtual reality is slowly moving forward. The GeForce GTX 1080 Ti delivers all the benefits of a VR experience with support for the NVIDIA VRWorks suite of tools. Multi-Res Shading and Lens Matched Shading technologies can significantly speed up image processing, more about this in the article on Pascal technologies. Combined with out-of-the-box performance, you get the ultimate VR solution with a GeForce GTX 1080 Ti.

Recently, technologies based on artificial intelligence systems with neural networks have been actively developed. Such self-learning systems make it possible to speed up the execution of many intellectual tasks. Specialized NVIDIA hardware solutions are actively used to create unmanned vehicles. Tesla accelerators are used in large computing centers, and a new generation of Tesla V100 accelerators has already been announced. NVIDIA is experimenting with the application of deep learning technologies in various areas, for example, for image recognition and enhancement, for recreating objects from images using photogrammetry. We recently wrote that NVIDIA has some developments to speed up image processing when rendering with ray tracing.

All this is still far from the average user. But the GeForce GTX also has a number of special features for ordinary players. GeForce Experience is being developed, this software center is getting new features.

The ShadowPlay video capture app will get new features in ShadowPlay Highlights. The program will track the user’s activity, automatically recording the most intense gaming moments.

NVIDIA Ansel, a special toolkit that allows you to take beautiful staged screenshots, is gaining more and more popularity. With Ansel, you can use the free camera for photo mode, get different angles, apply post-processing effects, and save images in ultra-high resolutions. Read more about Ansel features in a separate article.

Ansel support is already present in many popular games — The Witcher 3, Dishonored 2, For Honor, War Thunder, Watch Dogs 3 and others. The SDK will be released to the public, allowing any developer to incorporate Ansel support into their project. So there will be even more games with support for this application.

More in this category:

« Testing GeForce GTX 1050 Ti and GeForce GTX 1050

Review and testing of Inno3D iChill GeForce GTX 1060 6GB X3 »

More in this category:

« Testing GeForce GTX 1050 Ti and GeForce GTX 1050

Review and testing of Inno3D iChill GeForce GTX 1060 6GB X3 »

NVIDIA introduced GeForce GTX Titan X on Pascal

architecture

07/22/2016 07:32

#2

dchekanov

When the sample arrives. The news came as a surprise, to be honest.

quote — reply

07/22/2016 08:22

#3

squanterfall

I wonder why nvidia is in such a hurry. Maybe a reaction to the accelerated exit 490?

quote — reply

07/22/2016 14:49

#4

Div

Well, finally, Nvidia rolls out a really powerful chip with non-corn performance. Life becomes more fun!
The number of transistors and PCBs is impressive, although GDDR5X memory instead of HBM is a little disappointing. 🙂

quote — reply

07/22/2016 16:54

#5

vorvort

Registered: 01/28/2015
Moscow
Guest

Posts: 615

There is no HBM, the cores have been cut, there are no calculations. They called Tishka Titan and put his price tag on? ))

quote — reply

07/23/2016 05:46

#6

Aksakal

Message squanterfall;22244
I wonder why nvidia is in such a hurry. Maybe a reaction to the accelerated exit 490?

(smiled) no matter what NV rolls out, there will always be people who will say — they say this is a reaction to a competitor (AMD)

and the Competitor, meanwhile, cannot even roll out one Low-card completely, everything pulls and pulls.
And apparently, apart from Low — 480 and TOP Vega, more single-chip is not expected.

This I mean, maybe this is still not HB’s reaction to anything — but the usual plan for the release of the line’s products.
Yes, I understand — this explanation is too simple, well, that’s how it is in life 90% of anything has a simple logical explanation.

quote — reply

07/23/2016 06:38

#7

Aksakal

Message Div;22251
Well, finally, Nvidia rolls out a really powerful chip with non-corn performance. Life becomes more fun!
The number of transistors and PCB is impressive, although GDDR5X memory instead of HBM is a little disappointing. 🙂

and what did HB once roll out corn???
as far as I know — AMD’s corn flops (flops — uuuu! in games zilch)

don’t care about memory — gddr5x is enough. Vaughn AMD ran with a fashionable memory like a cat with a bubble, but there was no point, because show-offs do not add speed, first you need to be able to figure out how to turn show-offs into a benefit.

Of course IMHO

quote — reply

07/23/2016 07:42

#8

Message vorvort;22253
No HBM, kernels cut, no calculations. They called Tishka Titan and put his price tag on? ))

1) What are your claims? 2) Who told you that she was cut on the chip? The Tesla P100 card has the same 3584 cores! Where did Huang promise you 3840 cores in the specifications??? 3) GTX 1080 outperforms Fury X by up to 50%, and here is an additional increase of up to 40%! 4) Fury X2 with about the same performance has a price even higher than $ 1500, and it also has less memory and higher power consumption !!! That is, on all counts there is nothing to compare. 5) Tesla is sold for 200-300 thousand rubles, but here they ask for only 76 thousand rubles for a complete analogue! Do you think that’s a lot??? I assure you, gamers would even pay twice as much for it than Huang requested at the start of sales.

quote — reply

23. 07.2016 22:16

#9

Aksakal

Message vorvort;22253
no calculations))

XM? (thoughtfully) and what’s wrong with that??? — on the contrary, well done NV — they shoveled the architecture and created a game CHIP
(instead of stupidly using the GP100) You should praise for this, not find fault.

quote — reply

23/07/2016 22:442872 Message vorvort;22253
cores cut )

Even if we assume that the full CHIP has 3840 cores.
So-o (thoughtfully scratching my mane) I suppose that the history of Titan is repeating itself
After all, AMD has not yet offered anything, NV is saving a trump card up its sleeve (in case of a fight with the older Vega / if it shoots)
If this is so, then the future Tishkovods can only be envied.

ZY
In one thing, I agree with you — apparently, AMD TOPs will again compete at the expense of «wild iron» / have a «wider» architecture, eat more.
The history of confrontation (FuriKh — 980Tishka) repeats itself.