NVIDIA entered the new generation video cards race. At the end of March this year, the GeForce GTX 680 appeared on sale – the main competitor of AMD Radeon HD 7970. Now we will tell you what is hidden under the code name Kepler, how many nuclei received a fresh crystal and at what frequency it works. Well, of course, we will answer the eternal question: who is faster, “green” or “red”?

Reorganization

Architecture Fermi , on which they bet the last couple of years, she exhausted itself. She was replaced by Kepler. At first glance, everything is still: stream processors are combined into SM-modules, they, in turn, form GPC clusters, and GPU is already collected from them. But if in GF110 ( GeForce GTX 580 ) Each SM-module was equipped with 32 CUDA yaders, then in Kepler there were six times more-192 pieces. It was added to the crazy blocks (now they are called SMX) and auxiliary elements: “textures” from now on 16 instead of 4, and the sets of special operations – 32 versus 4 SFU in the previous flagship.

SMX pair and one rasterizer form GPC (Graphics Processing Cluster). Four such sets form the basis of the chip GK104 – hearts GTX 680. Thus, the new GPU carries 1536 nuclei and 128 “textures” on board, while GF110 was content with 512 processors and 64 TMU. But the number of Polymorph Engine, responsible for the tesselar, decreased: 8 versus 16 pieces of the past generation. Few? As if not. The developers not only revised the principles of the functioning of PM Engine, but also added megagers. So, according to engineers, the total performance increased by about 28%.

But what Kepler refused – the separation of frequencies. In Fermi, stream processors worked at the doubles of stone, now the numbers were equal. On the one hand, this is good: the main elements GK104 received a non -fingered dose of adrenaline. On the other hand, the final power of a single Cuda Yadra fell noticeably.

Another loss is a memory bus. GK104 is equipped with four 64-bit GDDR5 controllers, so the discharge is 256 versus 384 bits in GF110 and Thaiti on Radeon HD 7970. Less now and ROP: 32 instead of 46 units in GTX 580.

New opportunities

Kepler introduced several fresh technologies. The first – GPU Boost. She monitors temperature and energy consumption and, if the indicators do not exceed critical values, automatically increases the speed and voltage of the crystal. All works at the hardware level and does not turn off even during manual acceleration.

The second is the ability to connect four monitors at once. Moreover, for 3D Vision Surround (games on three displays) no longer needs a SLI tie, one GTX 680 is enough.

Third – discrete chip of video encoding Nvenc. NVIDIA cards had previously coped well with the processing of video streams, but used the computing power of the CUDA nuclear. Now they play only an auxiliary role: the main calculations performs a new module and does it four times faster Fermi. Novenc weak spot – it is only familiar with the Codec H.264.

Fourth – fresh smoothing algorithms, TXAA And FXAA. The latter is the "cheap" replacement of the traditional MSAA. The impact on performance is minimal, but you have to pay with easy blurry pictures. TXAA is a more serious approach, and, according to the developers, it not only exceeds MSAA in quality, but also requires less resources.

Finally, the fifth is the mode of adaptive vertical synchronization. Everything is simple here: Vsync is activated only when FPS exceeds the screens. As a result, Framretite subsidence below the treasured threshold no longer leads to a sharp drop in speed.

Cold

Despite a three -fold increase in the number of stream processors, the number of transistors used to build a chip has increased slightly – from 3 to 3.54 billion. Thanks to the new 28-nm technology process, the area of ​​the nucleus and the level of energy consumption turned out to be small. Even at the maximum load, TDP does not exceed 195 watts, and this is when the basic frequency reached a fantastic mark of 1006 MHz, and under the GPU Boost rises to 1058 MHz! For comparison: Radeon HD 7970, operating at 925 MHz and carrying 4.31 billion transistors on board, consumes 250 watts.

Cold disposition GK104 allowed to simplify the design of the board. GTX 680 does not have a 8-Pin PCIE connector, so loved by top-end solutions-a pair of 6-pin sockets was enough. Only four phases were allocated for food, and judging by the wiring, they refused the fifth at the last moment. The length of the card is 25 cm, which is slightly smaller than that of the flagship GeForce of previous generations.

The novelty is equipped with eight GDDR5 modules with a total capacity of 2 GB. Memory speed is 6008 MHz record, which partially compensates for the low tire discharge. The back panel is equipped with one HDMI 1.4A with DisplayPort 1.2 and pair DVI. PCIe 3 is used as an interface.0, with reverse compatibility with PCIE 2.0.

A two -story turbine is responsible for cooling. It includes a large aluminum radiator with three heat tubes and an unusual cylindrical turntable for “green”, in the design of which sound -absorbing materials are used.

The first in the world

We received a fee directly from the Russian office NVIDIA. The appearance of the card is standard: the entire surface is closed with a black plastic casing and decorated with branded logos. The GTX 680 dimensions are really smaller than its predecessor and looks very neat. Of the non-standard solutions, we note the unusual location of the power contacts: two 6-pin nests are not in the line, but over the other.

For tests, we collected a motherboard -based stand Gigabyte GA-X58A-UD3R. As a processor used Core i7-920 , Three planks put in memory Kingston Hyperx DDR3-1666 MHz 2 GB each, Windows 7 Ultimate 64-bit And all the programs recorded on Kingston SSDNOW.

The list of applications included 3dmark11 , Unigine Heaven Benchmark 2.5 , Just Cause 2 , Dirt 2 , Aliens vs. Predator , Batman: Arkham City And Total War: Shogun 2. The competitors were assigned GeForce GTX 590 , GTX 580 , AMD Radeon HD 6970 And HD 7970.

Who is the first?

In 3dmark11, the novelty turned out to be 17% faster than HD 7970 and approached the results of the double GTX 590! I liked Kepler and greedy to the resources of UNIGINE Heaven Benchmark 2.5: superiority over AMD – 26%. In games, the numbers turned out even more interesting.

In Aliens vs. Predator Flagship "Green" showed 57.7 FPS, ahead of the GTX 580 from HD 6970 and retired from HD 7970 and GTX 590 by 1.5 and 15.8 frames, respectively. The newcomer also lost in Just Cuse 2, which is optimized for AMD drivers – the difference with HD 7970 was 21.7 FPS. It was possible to differ in Dirt 2, the "red" leader missed 37.5 FPS.

The next victory Nvidia – Batman: Arkham City. With the maximum Physx, MSAA 8x and a resolution of 1920×1080, the hero of our review issued completely playable 26 FPS, while HD 7970 barely reached 20 frames. After the Physx, the situation did not change: with MSAA 4x, the gap was 4-6 FPS, and after the transition to 8x-15-18 fps. The triumphal procession GTX 680 stopped Total War: Shogun 2. Senior GeForce reached 23.2 FPS, HD 7970 – to 24.9.

The last word

Let's just say, from the GeForce GTX 680 we were waiting for more. Look at the final ratio of FPS: only 8% faster HD 7970. The same difference two years ago was between GTX 580 and HD 6970. Yes, in many applications the "green" flagship on the head exceeds its competitor. But all this is just optimized for nvidia games. In impartial tests – aliens vs. Predator, Batman: Arkham City, Total War: Shogun 2 – Opponents go in a par.

Today's victory GTX 680, quite possibly, the merit of the late exit. Game to the GeForce earlier HD 7970, and AMD engineers from the nose, but would pull up Graphics Core NEXT to Kepler frequencies. What would come of it – see our signs. We dispersed HD 7970 to the average speed GK104 – 1035 MHz – and got the results similar to GTX 680.

It is impossible to call GTX 680 an unconditional leader. By performance, the card is not much ahead of HD 7970. Perhaps the only advantage of Kepler is technology. Automatic acceleration on the TDP reserve, original smoothing modes, the ability to play on three monitors, a separate video decoding chip, Physx, 3D vision, noise insulation of the cooling system. And all this with extremely modest energy consumption in 195 watts. Otherwise AMD and NVIDIA are approximately equal. And now it all depends on competent support, setting up drivers and, of course, the price policy of companies. For example, HD 7970 can already be bought for 16,500 rubles, an official price tag on GTX 680 – 17,990 rubles.

Thoughts out loud

If you distract a little from Kepler and take a closer look at the characteristics, you can notice a lot of interesting things. GTX 680 looks too strange for a top video card. The first thing that catches your eye is a noticeable simplification of the memory subsystem: only 256 versus 384 bits at GTX 580. Further – more. The energy consumption of the new items is surprisingly enough – 195 watts. And although this corresponds to current trends in increasing productivity on watts, we all know that NVIDIA was never shy about making large and hot chips that consumed 250 watts. The GDDR5 is also embarrassed by the GDDR5 – only 2 GB, and this is a board that should provide a crazy resolution of 5760×1080 when playing three monitors. By the way, about the latter: GK104 supports only four displays at the same time, although he could well have pulled six, like senior representatives of AMD. And finally, a negligible increase in elements and a modest core area. But Nvidia never sought minimalism.

Conducting parallels with past generations, it seems that we have a cut version of the crystal in front of us and soon we may well expect something more powerful. Of course, these are just thoughts out loud, and perhaps Nvidia really went against its own principles, but very much GTX 680 is similar to some GTX 670.

Interesting technologies

In the main part of the article, we mentioned several new technologies. We will tell you more about them. Let's start with Adaptive Vsync.

Any monitor has its own scan frequency. For modern LCD panels, it is 60 or 120 Hz with the support of 3D Stereo. These figures indicate the maximum number of personnel, which the screen can show in one second. However, in some games, FPS crosses this border. On the one hand, this is good – no brakes. On the other hand, exceeding the scan leads to twitching the image.

Vsync allows you to adjust the frequency of update to the possibilities of the monitor. If the matrix is ​​limited to 60 Hz, then the video card will not give more than 60 FPS, which means that there will be no distortion of the picture. But this approach has one significant minus. Когда GPU не может обеспечить нужное число кадров, VSync выставляет планку «не больше» 30, потом 20 и 10 fps. And does not pay attention that the card easily holds, for example, 40 fps. Nvidia solved this problem.

Adaptive VSync monitors the system performance, and if the speed drops below 60 FPS, it turns off VSYNC, preventing the board from idle. As soon as the game is gaining momentum, the restriction returns. Bottom line: lack of artifacts and artificial freezes.

Smoothing

The next on the agenda is new smoothing methods designed to replace the usual Multisampling Antialiasing. To understand what the standard MSAA is bad, you need to deal with the principles of its work.

The technology was invented to get rid of aliasing in the angles of objects. Unpleasant Zazubins appear for a very simple reason. LCD panel can be represented as a sheet of notebooks in a small cell, each of which is allowed to paint in one color. Try to draw some house according to these rules: horizontal and vertical lines will turn out even, and inclined lines-"Christmas tree". The computer is all the same. The video card turns the figures into pixels (this is done by ROP blocks) and fills the screen. Since there are few points of the display, characteristic “cloves” come out on uneven planes. To get rid of them is relatively simple: you need to repaint the neighboring pixel objects with the edges of the edges. To determine them – and use anti -aliasing.

At first, for this, Supersampling was used: the scene rendered in resolution 2/4/8 times higher than the necessary, then the picture was compressed to the capabilities of the matrix and was displayed on the monitor – the Zazubrins disappeared. One thing is bad – this approach consumed indecently a lot of resources, so Multisampling soon replaced it. With the help of cunning algorithms, he processed exclusively problem areas and did not touch the rest of the image, which allowed to significantly increase performance.

Unfortunately, with an increase in the number of landfills and the emergence of new complex effects, this technology was ineffective: the loss in speed ceased to correspond to an increase in quality. Plus many problems began to create engines with deferred rendering. They first calculate the geometry and only then apply lighting. A striking example is “Metro 2033”, where the inclusion of MSAA led to a monstrous fall of FPS. In general, there are enough difficulties with modern anti -aliasing, and manufacturers are looking for an alternative to him.

The first swallow was Morphological AA, announced with AMD Radeon HD 6970. The video card renders frame in standard resolution and applies a light “photoshop” to it – blurs an image. As a result, no “Zazubrin”, but at the same time the clarity of the textures is lost. Nvidia now uses the same technique, however, it calls its fxaa.

But TXAA is already something new. It combines both traditional methods of anti -aliasing and post -cutting based on HDR and information from previous frames. The final quality result exceeds MSAA 8x, and consumes less resources than MSAA 4X.

Note that all these innovations are not the games of marketers. We conducted a test in which we compared MSAA and FXAA. The difference in performance was 10%. You can see the result on the screenshots given by a number. FXAA does not look as beautiful as the traditional MSAA, but in dynamics it is almost not noticeable.

Unfortunately, we could not check the brand new TXAA. It should be built into the game; His support in future projects has already been announced Epic , Crytek , Gearbox And CCP.

Acceleration

GPU Boost, implemented in the GTX 680, has long asked for video cards. In central processors, automatic acceleration has been used since the first Core i7. In GPU, it penetrated only with the advent of GTX 580, and even then in a cut -out form. The technology was intended to protect against overheating: it lowered the speed when exceeding TDP. Thus, NVIDIA protected its chips from special grenes like Furmark and could significantly increase the final frequencies of work. The disadvantage of the old approach was implementation: the drivers who reacted to pre -prescribed applications were controlled by all.

This error was corrected in AMD. PowerTune, presented with the Radeon HD 6970, also fought overheating, but the sensors built into the crystal, which turned out to be significantly large than software solutions and did not depend on specific programs, followed the temperature.

In GTX 680, engineers went even further. Hardware controls now accelerate GPU. Когда игра не может полностью нагрузить камень и есть запас по TDP, быстродействие платы увеличивается. As soon as the maximum level of energy consumption is achieved, the card reduces momentum. Interestingly, the declared 1056 MHz is not the limit for the GPU Boost. In our tests, Kepler often rose to the mark of 1100 MHz.

The only negative point is that “doping” cannot be disconnected. Even with an increase in the basic frequency, GPU Boost continues to work. So, by putting Base Clock at 1100 MHz, we witnessed how GK104 reached 1300 MHz! True, only professional testers will complain about this, you and I will never hurt a “free” increase.

Physics

Another news is Physx update. With GTX 680 he received two fresh effects. The first is a modified version of hair calculation. If separate strands were shown on the GTX 580, then at the Kepler presentation they showed us a technical demo with natural goets, each hair of which reacted to external influences like wind or stroking.

The second is the possibility of destroying objects. Nvidia taught the engine to break up objects in real time. And not just calculate the animation of pre -prepared blocks, but naturally break entire columns, apply textures on the resulting fragments and spread them according to all the laws of physics.

Table 1.

Technical characteristics

The core

Nvidia GeForce GTX 690

Kepler

AMD Radeon HD 7970

Thaiti

AMD Radeon HD 7950

Thaiti

AMD Radeon HD 6970

Cayman

Nvidia GeForce GTX 580

GF110

Nvidia GeForce GTX 590

2x GF110

The number of transistors

Nvidia GeForce GTX 690

3.54

AMD Radeon HD 7970

4.31 billion

AMD Radeon HD 7950

4.31 billion

AMD Radeon HD 6970

1.95 billion

Nvidia GeForce GTX 580

3 billion

Nvidia GeForce GTX 590

2x 3 billion

Process process

Nvidia GeForce GTX 690

28-nm

AMD Radeon HD 7970

28-nm

AMD Radeon HD 7950

28-nm

AMD Radeon HD 6970

40-nm

Nvidia GeForce GTX 580

40-nm

Nvidia GeForce GTX 590

40-nm

The number of stream processors

Nvidia GeForce GTX 690

1536 pcs.

AMD Radeon HD 7970

2048 pcs.

AMD Radeon HD 7950

1792 pcs.

AMD Radeon HD 6970

1536 pcs.

Nvidia GeForce GTX 580

512 pcs.

Nvidia GeForce GTX 590

2x 512 pcs.

The frequency of the graphic nucleus

Nvidia GeForce GTX 690

1006 MHz

AMD Radeon HD 7970

925 MHz

AMD Radeon HD 7950

800 MHz

AMD Radeon HD 6970

880 MHz

Nvidia GeForce GTX 580

772 MHz

Nvidia GeForce GTX 590

607 MHz

Frequency of stream processors

Nvidia GeForce GTX 690

1006 MHz

AMD Radeon HD 7970

925 MHz

AMD Radeon HD 7950

800 MHz

AMD Radeon HD 6970

880 MHz

Nvidia GeForce GTX 580

1544 MHz

Nvidia GeForce GTX 590

1214 MHz

Type, memory volume

Nvidia GeForce GTX 690

GDDR5, 2 GB

AMD Radeon HD 7970

GDDR5, 3 GB

AMD Radeon HD 7950

GDDR5, 3 GB

AMD Radeon HD 6970

GDDR5, 2 GB

Nvidia GeForce GTX 580

GDDR5, 1.5 GB

Nvidia GeForce GTX 590

GDDR5, 3 GB

Memory frequency

Nvidia GeForce GTX 690

6008 MHz

AMD Radeon HD 7970

5500 MHz

AMD Radeon HD 7950

5000 MHz

AMD Radeon HD 6970

5500 MHz

Nvidia GeForce GTX 580

4008 MHz

Nvidia GeForce GTX 590

3414 MHz

Data tire

Nvidia GeForce GTX 690

256 bits

AMD Radeon HD 7970

384 bits

AMD Radeon HD 7950

384 bits

AMD Radeon HD 6970

256 bits

Nvidia GeForce GTX 580

384 bits

Nvidia GeForce GTX 590

2x 384 bits

The number of texture blocks

Nvidia GeForce GTX 690

128 pcs.

AMD Radeon HD 7970

128 pcs.

AMD Radeon HD 7950

112 pcs.

AMD Radeon HD 6970

96 pcs.

Nvidia GeForce GTX 580

64 pcs.

Nvidia GeForce GTX 590

2x 64 pcs.

The number of rasterization blocks

Nvidia GeForce GTX 690

32 pcs.

AMD Radeon HD 7970

32 pcs.

AMD Radeon HD 7950

32 pcs.

AMD Radeon HD 6970

32 pcs.

Nvidia GeForce GTX 580

46 pcs.

Nvidia GeForce GTX 590

2x 46 pcs.

Energy consumption

Nvidia GeForce GTX 690

195 t

AMD Radeon HD 7970

250 watts

AMD Radeon HD 7950

200 watts

AMD Radeon HD 6970

250 watts

Nvidia GeForce GTX 580

244 watts

Nvidia GeForce GTX 590

365 watts

The length of the board

Nvidia GeForce GTX 690

250 mm

AMD Radeon HD 7970

270 mm

AMD Radeon HD 7950

260 mm

AMD Radeon HD 6970

270 mm

Nvidia GeForce GTX 580

267 mm

Nvidia GeForce GTX 590

280 mm

Interface

Nvidia GeForce GTX 690

PCIE 3.0 x16

AMD Radeon HD 7970

PCIE 3.0 x16

AMD Radeon HD 7950

PCIE 3.0 x16

AMD Radeon HD 6970

PCIE 2.0 x16

Nvidia GeForce GTX 580

PCIE 2.0 x16

Nvidia GeForce GTX 590

PCIE 2.0 x16

Price for January 2012

Nvidia GeForce GTX 690

17 990 rubles

AMD Radeon HD 7970

16 500 rubles

AMD Radeon HD 7950

15,000 rubles

AMD Radeon HD 6970

10,000 rubles

Nvidia GeForce GTX 580

12 500 rubles

Nvidia GeForce GTX 590

23,500 rubles

Table 2.

Synthetic tests

3dmark11

Graphics

Nvidia GeForce GTX 680

9659

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

7326

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

VTX Radeon HD 6970

5506

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

6182

Point of View GeForce GTX 590

9729

Physics

Nvidia GeForce GTX 680

5286

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

5210

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

VTX Radeon HD 6970

5200

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

4653

Point of View GeForce GTX 590

5287

Score

Nvidia GeForce GTX 680

7930

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

6603

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

VTX Radeon HD 6970

5367

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

5746

Point of View GeForce GTX 590

7922

%

Nvidia GeForce GTX 680

100%

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

83%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

VTX Radeon HD 6970

68%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

72%

Point of View GeForce GTX 590

99%

Unigine Heaven Benchmark 2.5

Fps

Nvidia GeForce GTX 680

43.7

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

32.6

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

37.4

VTX Radeon HD 6970

24.1

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

29.6

Point of View GeForce GTX 590

46.9

Score

Nvidia GeForce GTX 680

1102

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

822

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

941

VTX Radeon HD 6970

606

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

747

Point of View GeForce GTX 590

1181

%

Nvidia GeForce GTX 680

100%

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

74%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

88%

VTX Radeon HD 6970

55%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

68%

Point of View GeForce GTX 590

107%

Table 3.

Game tests (personnel per second)

Aliens vs. Predator (DX11)

Veryhight. 1680×1050, AF 16X, AA 2X

Nvidia GeForce GTX 680

69.5

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

76.6

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

82.7

VTX Radeon HD 6970

61.1

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

63

Point of View GeForce GTX 590

94.3

Veryhigh, 1680×1050, AF 16X, AA 4X

Nvidia GeForce GTX 680

64

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

65.9

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

70.8

VTX Radeon HD 6970

57.7

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

53.3

Point of View GeForce GTX 590

81.3

Veryhigh, 1920×1080, AF 16X, AA 2X

Nvidia GeForce GTX 680

71.8

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

68.9

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

74.2

VTX Radeon HD 6970

55.1

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

56.7

Point of View GeForce GTX 590

85.4

Veryhigh, 1920×1080, AF 16X, AA 4X

Nvidia GeForce GTX 680

57.7

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

59.2

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

63.7

VTX Radeon HD 6970

51.1

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

48

Point of View GeForce GTX 590

73.5

%

Nvidia GeForce GTX 680

100%

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

103%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

111%

VTX Radeon HD 6970

85%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

84%

Point of View GeForce GTX 590

127%

Dirt 2 (DX11)

Ultra. 1680×1050, AF 16X, AA 4X

Nvidia GeForce GTX 680

124.1

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

98

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

101.1

VTX Radeon HD 6970

78.5

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

125.2

Point of View GeForce GTX 590

117.8

Ultra, 1680×1050, AF 16X, AA 8x

Nvidia GeForce GTX 680

119.9

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

89.1

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

100.5

VTX Radeon HD 6970

75.5

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

116.3

Point of View GeForce GTX 590

122.9

Ultra, 1920×1080, AF 16X, AA 4X

Nvidia GeForce GTX 680

121.5

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

89

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

98

VTX Radeon HD 6970

76

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

117.5

Point of View GeForce GTX 590

119.2

Ultra, 1920×1080, AF 16X, AA 8x

Nvidia GeForce GTX 680

121

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

83.5

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

97.9

VTX Radeon HD 6970

73.7

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

109.9

Point of View GeForce GTX 590

121

%

Nvidia GeForce GTX 680

100%

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

74%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

82%

VTX Radeon HD 6970

62%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

96%

Point of View GeForce GTX 590

99%

Just Cause 2

High, 1680×1050, AF 16X, AA 4X

Nvidia GeForce GTX 680

72.3

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

119.1

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

119.9

VTX Radeon HD 6970

83.6

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

68.9

Point of View GeForce GTX 590

58.2

High, 1680×1050, AF 16X, AA 8x

Nvidia GeForce GTX 680

76.1

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

103.4

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

109.1

VTX Radeon HD 6970

74.8

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

62.9

Point of View GeForce GTX 590

63.5

High, 1920×1080, AF 16X, AA 4X

Nvidia GeForce GTX 680

74.8

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

111.9

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

112.8

VTX Radeon HD 6970

76.1

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

65.1

Point of View GeForce GTX 590

61.3

High, 1920×1080, AF 16X, AA 8x

Nvidia GeForce GTX 680

71.5

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

93.2

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

https://sister-site.org/bets-io/

99.8

VTX Radeon HD 6970

67.1

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

62.7

Point of View GeForce GTX 590

62

%

Nvidia GeForce GTX 680

100%

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

145%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

150%

VTX Radeon HD 6970

102%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

88%

Point of View GeForce GTX 590

83%

Batman: Arkham City (DX11, Full Physx)

High, 1680×1050, AF 16X, AA 4X

Nvidia GeForce GTX 680

28

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

22

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

22

VTX Radeon HD 6970

21

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

26

Point of View GeForce GTX 590

27

High, 1680×1050, AF 16X, AA 8x

Nvidia GeForce GTX 680

27

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

21

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

21

VTX Radeon HD 6970

20

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

23

Point of View GeForce GTX 590

25

High, 1920×1080, AF 16X, AA 4X

Nvidia GeForce GTX 680

28

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

22

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

22

VTX Radeon HD 6970

21

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

25

Point of View GeForce GTX 590

26

High, 1920×1080, AF 16X, AA 8x

Nvidia GeForce GTX 680

26

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

20

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

20

VTX Radeon HD 6970

19

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

23

Point of View GeForce GTX 590

24

%

Nvidia GeForce GTX 680

100%

XFX R7970 DOUBLE DISSIPATION (925/5500 MHz)

78%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

78%

VTX Radeon HD 6970

74%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

89%

Point of View GeForce GTX 590

94%

Batman: Arkham City (DX11, No Physx)

Ultra. 1680×1050, AF 16X, AA 4X

Nvidia GeForce GTX 680

66

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

62

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

65

VTX Radeon HD 6970

51

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

53

Point of View GeForce GTX 590

49

Ultra, 1680×1050, AF 16X, AA 8x

Nvidia GeForce GTX 680

61

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

43

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

48

VTX Radeon HD 6970

39

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

41

Point of View GeForce GTX 590

42

Ultra, 1920×1080, AF 16X, AA 4X

Nvidia GeForce GTX 680

62

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

56

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

59

VTX Radeon HD 6970

45

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

48

Point of View GeForce GTX 590

48

Ultra, 1920×1080, AF 16X, AA 8x

Nvidia GeForce GTX 680

53

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

38

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

40

VTX Radeon HD 6970

35

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

38

Point of View GeForce GTX 590

41

%

Nvidia GeForce GTX 680

100%

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

82%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

88%

VTX Radeon HD 6970

70%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

74%

Point of View GeForce GTX 590

74%

Total War: Shogun 2 (DX11)

Ultra. 1680×1050, AF 16X, AA 4X

Nvidia GeForce GTX 680

42.8

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

42.1

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

42.8

VTX Radeon HD 6970

39.7

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

39

Point of View GeForce GTX 590

59.6

Ultra, 1680×1050, AF 16X, AA 8x

Nvidia GeForce GTX 680

27.8

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

28.6

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

31.1

VTX Radeon HD 6970

29.9

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

30

Point of View GeForce GTX 590

57.2

Ultra, 1920×1080, AF 16X, AA 4X

Nvidia GeForce GTX 680

36.9

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

34.3

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

37.8

VTX Radeon HD 6970

34.9

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

34.4

Point of View GeForce GTX 590

53.6

Ultra, 1920×1080, AF 16X, AA 8x

Nvidia GeForce GTX 680

23.2

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

24.9

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

26.2

VTX Radeon HD 6970

26

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

29.3

Point of View GeForce GTX 590

41.3

%

Nvidia GeForce GTX 680

100%

ASUS HD 7970 DirectCU II TOP (925/5500 MHz)

99%

ASUS HD 7970 DirectCU II Top (1035/5500 MHz)

105%

VTX Radeon HD 6970

100%

Zotac GeForce GTX 580 AMP! Edition (772/4008 MHz)

101%

Point of View GeForce GTX 590

161%

Table 4.

Price/performance ratio

Performance

Nvidia GeForce GTX 680

100%

AMD Radeon HD 7970 (925/5500 MHz)

92%

AMD Radeon HD 7970 (1035/5500 MHz)

100%

AMD Radeon HD 6970

77%

Nvidia GeForce GTX 580

84%

Nvidia GeForce GTX 590

106%

Price

Nvidia GeForce GTX 680

100%

AMD Radeon HD 7970 (925/5500 MHz)

91%

AMD Radeon HD 7970 (1035/5500 MHz)

91%

AMD Radeon HD 6970

56%

Nvidia GeForce GTX 580

69%

Nvidia GeForce GTX 590

131%