Card test in new gtx 970 games. Video cards

Test stand, testing methodology
Test bench configuration
CPU Intel Core i7-3960X @ 4.6 GHz (100x46) Intel Core i7-3970X @ 4.6 GHz (100x46)
Motherboard ASUS P9X79 Pro
RAM DDR3 Kingston HyperX 4x2 GB @ 1600 MHz, CL9
ROM Intel SSD 520 240 GB
power unit Corsair AX1200i, 1200 W Seasonic Platinum-1000, 1000 W
CPU cooling Thermalright Archon
Frame CoolerMaster Test Bench V1.0
operating system Windows 7 Ultimate X64 Service Pack 1
Software for AMD GPUs AMD Catalyst 13.4
NVIDIA GPU Software 344.07

To measure system power, a stand with a Corsair AX1200i power supply is used. Energy-saving CPU technologies are disabled in all tests. PCI-Express bus works in 3.0 mode. To activate PCI-E 3.0 on GeForce 600 and 700 series video cards in a system based on the X79 chipset, a patch from NVIDIA is used.

In the NVIDIA driver settings, the CPU is always selected as the processor for PhysX calculations. For AMD cards, the Tesselation setting is always transferred from the AMD Optimized state to Use application settings. In CrossFire configurations, the Frame Pacing option remains enabled.

Benchmarks: synthetic
Program API Settings Permission
3DMark 2011 DirectX 11 Extreme profile
3DMark DirectX 11 Fire Strike test (not Extreme)
TessMark DirectX 11 Set 4 (2048x2048), tessellation x8/x16/x32/x64 AF Off 1920x1080
Unigine Heaven 4 DirectX 11 DirectX 11, max. quality, tessellation in Extreme mode AF 16x, MSAA 4x 2560x1440 / 3840x2160
Benchmarks: games
Program API Settings Anisotropic filtering, full screen anti-aliasing Permission
Far Cry 3 + FRAPS DirectX 11 DirectX 11, max. quality, HDAO. Start of the Secure the Outpost mission AF, MSAA 4x 2560x1440 / 3840x2160
Tomb Raider. Built-in benchmark DirectX 11 Max. quality AF 16x, SSAA 4x 2560x1440 / 3840x2160
Bioshock Infinite. Built-in benchmark DirectX 11 Max. quality. Postprocessing: Normal AF 16x, FXAA 2560x1440 / 3840x2160
Crysis 3 + FRAPS DirectX 11 Max. quality. Start of the Post Human mission AF 16x, MSAA 4x 2560x1440 / 3840x2160
Metro: Last Light. Built-in benchmark DirectX 11 Max. quality AF 16x, SSAA 4x 2560x1440 / 3840x2160
Company of Heroes 2. Built-in benchmark DirectX 11 Max. quality AF,SSAA 4x 2560x1440 / 3840x2160
Batman: Arkham Origins. Built-in benchmark DirectX 11 Max. quality AF, MSAA 4x 2560x1440 / 3840x2160
Battlefield 4 + FRAPS DirectX 11 Max. quality. Beginning of the Tashgar mission AF 16x, MSAA 4x + FXAA 2560x1440 / 3840x2160
Thief. Built-in benchmark DirectX 11 Max. quality AF 16x, SSAA 4x + FXAA 2560x1440 / 3840x2160

Test participants

The following video cards took part in performance testing:

  • AMD Radeon R9 280 (918/5500 MHz, 2 GB)
  • GIGABYTE GeForce GTX 970 G1 Gaming (1050/7012 MHz, 4 GB)

Clock speeds, power consumption, temperature, overclocking

Although the GTX 970 has a lower base clock than the GTX 980, it actually has a higher GPU Boost range. In games, the GPU reaches frequencies up to 1213 MHz (163 MHz above the base), and, therefore, in this regard, it is not far behind the GTX 980 (1253 MHz). The GTX 970 is slightly voltage limited at 1.212V, while the GTX 980 goes up to 1.225V.

Regarding the video GIGABYTE cards, then in it with standard settings GPU operates at frequencies from 1178 MHz (Base Clock) to 1342 MHz (the maximum value observed under load).

The BIOS allows the power limit and GPU temperature to be increased to 112% and 91 °C, respectively. You can also raise the maximum voltage on the GPU to 1.256 V. At the same time, the frequency range increases by several steps. We managed to overclock our copy to 1278 MHz base frequencies, and frequencies up to 1467 MHz were observed in dynamics. The video memory increased its effective frequency by 1 GHz without any problems.

Base Clock, MHz Max. Boost Clock, MHz Base Clock, MHz (overclocking) Max. registered Boost Clock, MHz (overclocking)
GeForce GTX 980 1127 1253 (+126) 1387 1526 (+139)
GeForce GTX 970 1050 1213 (+163) 1278 1467 (+189)
GeForce GTX TITAN Black 889 1032 (+143) 1100 1262 (+162)
GeForce GTX TITAN 836 1006 (+145) 966 1150 (+184)
GeForce GTX 780 Ti 876 1020 (+144) 986 1130 (+144)
GeForce GTX 780 863 1006 (+143) 1053 1215 (+162)
GeForce GTX 770 1046 1176 (+130) 1190 1333 (+143)

At reference frequencies, the GeForce GTX 970 is practically no different in power from its flagship brother under load in games. The new product is also slightly more economical compared to the GeForce GTX 780, although the GTX 770 demonstrates less power. The GTX 970 also looks incomparably better compared to any AMD video cards- from R9 280X to R9 290X.

WITH using FurMark managed to catch a slight difference between the GTX 970 and GTX 980, and the ratio of GTX 970 and GTX 770 shifted in favor of the new product. Top-end AMD video adapters are still much more power-hungry, but the Radeon R9 280X paradoxically demonstrates lower power in FurMark than in the gaming test.

The GIGABYTE board has overclocking power to spare. After increasing the frequencies, power consumption in both tests increased by about 30 W compared to the results of the card at the reference frequencies for the GTX 970.

The cooling system of the GIGABYTE card is so effective that the card took first place with the lowest GPU temperatures under load by a large advantage over the rest of the test participants. However, we cannot remain silent about the annoying drawback of the cooler: when idle, the fans constantly either accelerate to high speeds or die down. Apparently this annoying behavior is the result unsuccessful firmware our test sample. After all, this is one of the first samples of the board.

Performance: Synthetic Benchmarks

  • The gap between the flagship card, the GTX 980, and the GTX 970 is quite large.
  • However, the new product leaves behind both the GeForce GTX 780 and its main rival, the Radeon R9 290.

  • The GTX 970 is still far behind the GTX 980, but it's not all that different from the GeForce GTX 780 Ti.
  • AMD's top video adapters remained behind the GTX 970, although the R9 290X was defeated by a minimal advantage.

TessMark

  • Disabling three SMMs in the GM204, coupled with a decrease in GPU clock speeds, affected performance in a predominantly geometric load. The GTX 970 is slightly slower compared to the GeForce GTX 980 and GTX 780 Ti.
  • And yet, even in this form, the GTX 970 left behind all competing AMD video adapters along with the GeForce GTX 770.

Unigine Heaven 4

Tomb Raider

  • The gap between the GTX 970 and the GTX 980 is hard to miss at both resolutions.
  • The GTX 970 and GTX 780 are equivalent at Ultra HD resolution, but at WQHD the new product is significantly faster.
  • The GTX 970 also competes head to head with the R9 290X in Ultra HD and even scores slightly more FPS at lower resolutions.

Bioshock Infinite

  • The GTX 970 is again far from the GTX 980.
  • But you can record a confident victory of the new product over the GTX 780 and Radeon R9 290X.

Crysis 3

  • Due to the low frame rate, the difference between the GTX 980 and GTX 970 is not noticeable.
  • The new product is also practically no different in performance from the GTX 780 and Radeon R9 290X.

Metro: Last Light

  • As in the previous test, there is a difference between the GTX 970 and GTX 980, but from a practical point of view it is small.
  • The GTX 970 also beats the GTX 780 by a narrow margin and is hot on the heels of the Radeon R9 290X.

Company of Heroes 2

  • The GTX 980 is noticeably superior to the GTX 970 only at WHQD resolution; in Ultra HD mode the distance decreases.
  • The GTX 970 is also slightly faster compared to the GTX 780.
  • But two top AMD video cards have moved ahead, although the gap between the GTX 970 and R9 290 is minimal.

Batman: Arkham Origins

  • GeForce GTX 970 is significantly different from its older friend.
  • The GTX 780 has only a minimal advantage over the GTX 970, and only in WQHD mode.
  • The new product is almost as good as AMD's flagship.

Battlefield 4

  • There is a huge difference between the GTX 970 and GTX 980.
  • The GTX 970 isn't that much better than the GTX 780, but it's about the same amount behind the Radeon R9 290.

  • The advantage of NVIDIA's flagship over the GTX 970 is more pronounced at WQHD resolution. In 4K the difference is not as noticeable due to the low frame rate.
  • The performance of the GTX 780 is 1-2 FPS less.
  • Radeon R9 290 is a fraction of FPS ahead of the new product.

The technical characteristics of the MSI GeForce GTX 970 Gaming video card are shown in the table in comparison with the reference options

NVIDIA GeForce GTX 980, NVIDIA GeForce GTX 780 Ti, NVIDIA Ge...

PCB design and features
The design of the newest MSI GeForce GTX 970 Gaming has changed dramatically in comparison with its predecessor, for example, the MSI GeForce GTX 780 Ti Gaming. Now the video card looks lighter in appearance, thanks to the plastic rather than metal casing of MSI’s most advanced cooling system – Twin Frozr V. But in style, in our opinion, the product has only benefited from this and has become even more attractive.


The entire front side of the video card is covered by a cooler with two fans with very interesting blades, which we will talk about in the section describing the cooling system. The heat pipes coming out of the radiator are visible from above, and back side The printed circuit board is not covered by anything. The dimensions of the MSI GeForce GTX 970 Gaming are 269x130x37 mm.

The video card is equipped DVI-I outputs and DVI-D (both Dual-Link), HDMI versions 1.4a/2.0 and DisplayPort version 1.2, combined with a grille for partial release of air heated by the video card outside the system unit case.


And yet, the bulk of the air heated by the video card will remain inside the system unit case. However, as we will soon see, this is an insignificant minus in the case of this particular MSI video card.

The MSI GeForce GTX 970 Gaming has two connectors for combining up to four video cards in various SLI modes, as well as two connectors for additional food, one of which is eight-pin.


The maximum power consumption level of the video card is stated at 148 watts, which is only 3 watts more than indicated in the specifications of the reference NVIDIA GeForce GTX 970. The recommended power supply for a system with one such video card is 500 watts.
Despite its massiveness, the cooling system is attached to printed circuit board with only four screws around the perimeter of the GPU, so removing it was not difficult. Below it you can see a radiator on the power circuits and a heat distribution plate on the memory chips and individual batteries.


Now we unscrew them from the printed circuit board. The MSI GeForce GTX 970 Gaming uses an original PCB design. The double-sided arrangement of memory chips and the empty contact pads for them are immediately striking. There is no point in doubting that eight-gigabyte versions of this video card model will soon be waiting for us.

Cooling systems

Almost simultaneously with the announcement of new NVIDIA GPUs, MSI introduced the next generation of original cooling systems for its video cards - Twin Frozr V. As was the case with Twin Frozr IV, we are likely to see several versions of coolers under the same name, designed for video cards of different classes. But the MSI GeForce GTX 970 Gaming most likely has a top version of this cooler installed. According to the developers, the design and implementation of the new Twin Frozr model took 18 months, and its advantage over the Twin Frozr IV Advanced version can reach 10 degrees Celsius at peak load. It is based on a nickel-plated radiator with a copper base, copper heat pipes and aluminum fins .

There are four heat pipes in total - two with a diameter of 6 mm and two more with a diameter of 8 mm. The tubes, alternating with each other, emerge from the base into different sides, almost evenly penetrating the radiator.

The radiator fins are connected to the heat pipes by soldering, as befits high-end cooling systems.

Thin aluminum plates are placed on heat pipes with a distance of just over 1 mm. Objectively speaking, we didn’t see any serious innovations in the radiator, so let’s see what kind of interesting radiators are installed on the Twin Frozr V.
The cooling system is equipped with two 100 mm Torx Fans, the original manufacturer of which is Power Logic (labeled PLD10010S12HH). The actual size of the impellers is 94 mm. Distinctive feature This type of fan has blades of different types, which not only alternate with each other, but also have different angles of attack.

Thanks to this shape of the blades and their increased number, the developers were able to achieve an increase in air flow with a 5% reduction in noise levels. In addition to design changes in the fans, their operating algorithm has also been optimized. The fans support Zero Frozr technology, which consists of stopping them completely in low load modes or 2D mode. Judging by the monitoring, the fans start spinning when the GPU reaches 60 degrees Celsius. Moreover, the fans start very softly, without a sharp surge in speed, as happens. Let us add that the fans are based on plain bearings with an extended service life.
On the power transistors and memory chips located on the front side of the printed circuit board, an aluminum radiator and a heat distribution plate with thermal pads are installed, respectively.

A nice bonus in Twin Frozr V, which will certainly please fans of modding, is the illuminated Gaming series logo in the upper part of the cooler casing.

She looks just great! However, see for yourself:

In the diagrams, the results of two video cards on NVIDIA GPUs of the now previous generation - ASUS STRIX GTX 780 OC Edition and EVGA GeForce GTX 770 Superclocked ACX - are highlighted in light green, and on the AMD GPU - Radeon R9 290X - in traditional red. To highlight the results of the heroine of today’s article, MSI GeForce GTX 970 Gaming, we “filled” them with dark turquoise, and the video card from the future article - NVIDIA GeForce GTX 980 - is highlighted in dark blue in the diagrams. Tests


The performance of the MSI GeForce GTX 970 Gaming in 3DMark is at a relatively high level. Not only is it more than 40% ahead of its model predecessor, the GeForce GTX 770 (in the original version, we note, and on higher frequencies), and the GeForce GTX 780 is not any serious competitor to the MSI GeForce GTX 970 Gaming in this test. In addition, the new product demonstrates the same performance as AMD’s single-processor flagship Radeon R9 290X. The lag of the MSI video card from the NVIDIA GeForce GTX 980 at the nominal frequency is about 10-12%, but is almost completely compensated by the impressive overclocking of the GPU and video memory of the MSI GeForce GTX 970 Gaming. So so far everything is going very well for a new product.

Unigine Valley Bench

The second semi-synthetic benchmark showed us a slightly different picture.


If in non-anti-aliasing mode the MSI GeForce GTX 970 Gaming is still faster than ASUS STRIX GTX 780 OC, then when MSAA8x is activated, the latter wins back the position due to a wider memory bus and its bandwidth. At the same time, the difference with the other video cards of today's testing in this test remained the same. It should be noted that the AMD Radeon R9 290X showed itself to be very confident when using anti-aliasing.

Total War: SHOGUN 2 – Fall of the Samurai

Game tests open with the good old Total War: SHOGUN 2 – Fall of the Samurai.


In three test modes out of four, video cards on NVIDIA GPUs were arranged according to height. The MSI GeForce GTX 970 Gaming delivers a 30% performance boost over the factory overclocked GeForce GTX 770, up to 14% faster than the ASUS STRIX GTX 980 OC, and 8-17% faster than the AMD Radeon R9 290X. That is, the predecessor and possible competitors have been beaten, the only thing left is not to lose to the new NVIDIA flagship, and the MSI GeForce GTX 970 Gaming is fully successful, since its maximum gap from the GeForce GTX 980 does not exceed 14% and is covered by excellent overclocking.

Sniper Elite V2 Benchmark

But in the game Sleeping Dogs the balance of power can be called “classic” for today’s test.


In the most resource-intensive mode, the MSI GeForce GTX 970 Gaming leaves behind the EVGA GeForce GTX 770 SC by 27%, is ahead of the ASUS STRIX GTX 780 OC by 6-7% and is not inferior to the AMD Radeon R9 290X. By overclocking the MSI GeForce GTX 970 Gaming to 1330/8072 MHz, you can get the same level of performance as the reference NVIDIA GeForce GTX 980, although at nominal frequencies the difference between these video cards ranges from 4 to 14%.

Hitman: Absolution


Hitman: Absolution is one of two and a half games in testing where the price has noticeably dropped in recent months AMD Radeon R9 290X beats MSI GeForce GTX 970 Gaming. Moreover, AMD's flagship single-processor video card is capable of competing with the NVIDIA GeForce GTX 980 (in the most resource-intensive mode). However, one cannot fail to note the noticeable superiority of the MSI GeForce GTX 970 Gaming over the EVGA GeForce GTX 770 SC, its equality with the ASUS STRIX GTX 780 OC and excellent scalability when overclocked, thanks to which even the reference GeForce GTX 980 does not look overly fast.

In Crysis 3 we can see slightly different results.


The advantage of the MSI GeForce GTX 970 Gaming over the factory overclocked GeForce GTX 770 here is more modest than in all previous tests and reaches only 18% at its peak. At the same time, the new product was slightly behind the ASUS STRIX GTX 780 OC in all quality modes and resolutions and noticeably lagged behind the AMD Radeon R9 290X. True, overclocking the MSI GeForce GTX 970 Gaming radically improves the situation and, along with the NVIDIA GeForce GTX 980, brings it to overall first place in the test. Of course, the GeForce GTX 980 also overclocks well, but we will tell you about this in the next article.

Tomb Raider (2013)

But in Tomb Raider new MSI The GeForce GTX 970 Gaming regains a more than 30 percent advantage over the EVGA GeForce GTX 770 SC and a slight increase in speed compared to the ASUS STRIX GTX 780 OC.


The reference AMD Radeon R9 290X is at the same level of performance, for which it deserves honor and praise (at such and such a low current cost). NVIDIA GeForce GTX 980 is 10-12% faster than MSI GeForce GTX 970 Gaming, but is easily achievable when overclocking the new product.

BioShock Infinite

The test built into the BioShock Infinite game was once again noted for its low minimum FPS on video cards with NVIDIA GPUs, in contrast to AMD.


However, in terms of average frames per second, the GeForce GTX and, in particular, the MSI GeForce GTX 970 Gaming are the leaders in testing, and the AMD Radeon R9 290X is somewhere between the EVGA GeForce GTX 770 SC and the ASUS STRIX GTX 780 OC.

Metro: Last Light

The activated Advanced PhysX technology leaves no chance for the AMD Radeon R9 290X and at the same time puts the overclocked MSI GeForce GTX 970 Gaming in the lead.


The situation looks different without using Advanced PhysX.


However, the MSI GeForce GTX 970 Gaming is on par with the AMD Radeon R9 290X, significantly faster than the GeForce GTX 770 with a factory-overclocked EVGA graphics card, and slightly faster than the ASUS STRIX GTX 780 OC. Getting ahead of the reference GeForce GTX 980 for the overclocked MSI GeForce GTX 970 Gaming is not particularly difficult.

Company of Heroes 2

It is very difficult to deal with video cards based on AMD GPUs in the game Company of Heroes 2, as the Radeon R9 290X demonstrated to us today.


But here the evolution of video cards on NVIDIA GPUs is very clearly visible, when from the EVGA GeForce GTX 770 SC to the NVIDIA GeForce GTX 980 or the overclocked MSI GeForce GTX 970 Gaming, the results of all video cards are arranged in almost equal steps.

Total War: Rome II


In Total War: Rome II, the new MSI GeForce GTX 970 Gaming is 39-54% faster than the EVGA GeForce GTX 770 SC and 3-16% faster than the ASUS STRIX GTX 780 OC, the same as with the AMD Radeon R9 290X. The lag from the reference NVIDIA GeForce GTX 980 is 9-17%, but again it is almost completely compensated by overclocking the Gaming video card.

Batman: Arkham Origins

Energy consumption
The energy consumption of the system with various video cards was measured using the Zalman ZM-MFC3 multifunctional panel, which shows the system consumption “from the outlet” as a whole (excluding the monitor). The measurement was carried out in 2D mode, during normal work in Microsoft Word or “Internet surfing”, as well as in 3D mode. In the latter case, the load was created using four consecutive cycles of the introductory scene of the “Swamp” level from the game Crysis 3 in a resolution of 2560x1440 pixels with maximum graphics quality settings, but without using MSAA anti-aliasing.

Let's compare the power consumption of systems with video cards tested today.


The system with the MSI GeForce GTX 970 Gaming video card turned out to be the most economical among other participants in today's testing. The level of its energy consumption as a whole reached 472 watts, and on average remained at 450 watts. It is noteworthy that the much weaker video card from the previous generation EVGA GeForce GTX 770 SC in the system consumes 30 watts more, so in terms of power consumption a serious step forward is obvious. Compared to the GeForce GTX 780 and Radeon R9 290X, the new GeForce GTX 970 looks even more convincing in this regard, outperforming systems with these video cards by 81 to 87 watts at peak load. The reference NVIDIA GeForce GTX 980 is also impressive, the configuration with which consumes barely more than the MSI GeForce GTX 970 Gaming. When overclocking the latter, system power consumption increases by only 21 watts at peak load.

Conclusion
There is a well-known Russian proverb “a fly in the ointment”, when some little thing spoils something big and exceptionally good. So today we have a solid “barrel of honey” without any spoons of black substances, and its name is MSI GeForce GTX 970 Gaming. The video card has an unprecedentedly high level of performance for its class, providing a significant increase in performance over the model it replaces from the previous generation, the GeForce GTX 770, and does this at a lower recommended price at the start of sales. The impressive overclocking potential of the new GM204 graphics processor of the Maxwell architecture and matching high-frequency memory can and do bring the MSI video card to the performance level of the flagship GeForce GTX 980, and in some graphics modes allows it to outstrip it. At the same time, the cost of the GeForce GTX 970 is much more affordable than that of the GeForce GTX 980, and, most likely, this particular model will be in greatest demand. Of course, here we must pay tribute to the MSI engineers, who equipped their video card not only with a high-quality element base with durable components, but also with the most advanced MSI Twin Frozr V cooler, which provides effective cooling with extremely low noise levels. Previously, it was not possible to achieve such results with air cooling at such high frequencies for almost top-end video cards. We note the relatively low level of power consumption of the new product and more modest power requirements for power supplies. In general, all that remains is to wait for such video cards to appear on wide sale in order to become the happy owner of one of these “barrels of honey”.

The NVIDIA GeForce GTX 970 video accelerator can hardly claim the title of “people's video card”. But the fact that this video card is from various manufacturers is in high demand among less wealthy gamers, there is no denying it. Neither the well-known scandal with 0.5 GB of slow video memory, nor the increased dollar exchange rate prevented the NVIDIA GeForce GTX 970 from becoming the optimal gaming graphics accelerator among video cards in the upper price range, which gives definite reasons to consider this product again.

In this review we will introduce you to the video card MSI GeForce GTX 970 Gaming 4G, which is quite popular among the entire set of GeForce GTX 970 graphics accelerators. The most important MSI features GeForce GTX 970 Gaming 4G are, of course, solid factory overclocked and proprietary system air cooling TWIN FROZR V with semi-passive operating mode.

Specifications

  • GPU - NVIDIA GeForce GTX 970
  • GPU codename - GM204
  • Technological process - 28 nm
  • Number of supported monitors - 4
  • Maximum resolution - 4096x2160
  • GPU frequency - 1140 MHz
  • Video memory capacity - 4096 MB
  • Video memory type - GDDR5
  • Video memory frequency - 7010 MHz
  • Video memory bus width - 256 bits
  • Support for SLI/CrossFire mode - yes
  • Connectors - DVI x2, HDCP, HDMI, DisplayPort support
  • Number universal processors - 1664
  • Shader version - 5.0
  • Number of texture blocks - 104
  • Number of rasterization blocks - 56
  • Standards support - DirectX 12, OpenGL 4.5
  • Additional power required - yes, 8 pin + 6 pin
  • TDP - 145 W
  • Number of fans - 2
  • Dimensions - 269x141 mm

Packaging and equipment

The black and red colors of MSI Gaming series products have long become familiar and recognizable, so the color of the MSI GeForce GTX 970 Gaming 4G box is fully consistent with this design. On the front panel of the package there is the name of the video card, its main key characteristics are indicated, and supported technologies are also listed.

Already on the back of the package, the new TWIN FROZR V air cooling system is described in detail, as well as a brief specification of the video card.

Inside the box, the MSI GeForce GTX 970 Gaming 4G video card is packaged in a translucent antistatic bag and placed in a molded polypropylene bath. With this level of protection, the user certainly does not have to worry about the integrity of the video card during transportation.

The MSI GeForce GTX 970 Gaming 4G package is modest and consists only of a user manual, a disk with software and drivers, a DVI-VGA adapter and a 6-pin to 8-pin power adapter.

Product appearance

MSI GeForce GTX 970 Gaming 4G is a video card with an open-type cooling system. The interplay of black and red plastic in the design of the cooling system made it possible to give the MSI GeForce GTX 970 Gaming 4G video card an expressive appearance.

Despite the massive cooling system, the MSI GeForce GTX 970 Gaming 4G video card, when installed in the system case, occupies only two slots. This is largely due to the new cooling system, in which the manufacturer made several subtle but important engineering decisions. First of all, the cooling system casing of the MSI GeForce GTX 970 Gaming 4G is somewhat recessed behind the metal rear connector strip, and secondly, low-profile fans with an impeller diameter of 100 mm are used here. These seemingly unimportant decisions made it possible to make the MSI GeForce GTX 970 Gaming 4G more compact in thickness, which can make it possible to install two such video cards in adjacent PCI-E x16 expansion slots.

The connectors on the rear panel include HDMI, DVI-I, DVI-D and DisplayPort.

Like all modern video accelerators, the MSI GeForce GTX 970 Gaming 4G video card has a PCI-E version 3.0 connector. However, work in old motherboards with PCI-E slots 2.0 is also supported by this video card.

The end part of the video card is not deprived useful elements- here, according to the manufacturer’s idea, illuminated corporate symbols are implemented.

She works in automatic mode, however, you can configure its operation in the proprietary utility MSI Gaming App.

Also on the front side you can find two 6+8 pin additional power connectors for the video card.

And on the other side there are a couple of connectors for SLI bridges. The MSI GeForce GTX 970 Gaming 4G video card supports multi-GPU combinations of two or three video cards simultaneously.

To control and monitor the cooling system, the MSI GeForce GTX 970 Gaming 4G uses a 6-pin connection connector, five of the six pins of which are used.

The back side of the MSI GeForce GTX 970 Gaming 4G video card does not have any reinforcing metal plate on the outside, and the board elements are available for detailed inspection.

On this side of the printed circuit board you can see four memory chips, wired in an L-shape. This is only half of the entire onboard memory of the video card; the remaining half is located on the other side of the video card and is hidden under the cooling system.

Printed circuit board

The TWIN FROZR V cooling system is mounted on the MSI GeForce GTX 970 Gaming 4G with four screws, unscrewing which you can see the graphics card's printed circuit board in more detail. A reinforcing metal plate with a radiator for power supply to the graphics processor immediately catches your eye.

Removing the reinforcing plate is also not difficult, after which you can already see the MSI GeForce GTX 970 Gaming 4G PCB in all its glory.

Based on a detailed inspection of the printed MSI boards GeForce GTX 970 Gaming 4G shows that this graphics accelerator is not an overly complex device and is not overloaded with a dense arrangement of elements on the board. Figuratively speaking, the MSI GeForce GTX 970 Gaming 4G could be made more compact in PCB size if desired.

The heart of the MSI GeForce GTX 970 Gaming 4G video card is the NVIDIA GM204 graphics chip, made using a 28 nm process technology. Unlike the reference GeForce GTX 970, where the graphics chip operates at 1050 MHz and increases its frequency to 1178 MHz using Boost Clock technology, on the MSI GeForce GTX 970 Gaming 4G graphics card the graphics chip operates at 1140 and 1279 MHz, respectively. Such a solid factory overclock allows not only to increase the overall performance of the video card, but also to become a strong argument in favor of choosing this particular video card.

The power system on the MSI GeForce GTX 970 Gaming 4G is made according to the (3+2+2) scheme, where the phases are distributed between the GPU, MEM and PLL, respectively.

As for the onboard memory, it communicates with the graphics processor via a 256-bit bus. The total memory capacity of 4 GB is achieved using eight Elpida W4030BABG-70-F chips.

Cooling system

We would like to touch on the issue of the cooling system of the MSI GeForce GTX 970 Gaming 4G, starting with the reinforcing metal platinum. This engineering solution not only adds rigidity to the printed circuit board of the video card, but also serves as a kind of heat sink from the RAM chips and some power supplies of the processor and memory. Heat transfer from appropriate heat sources is carried out through special thermal pads.

The main pride and asset of the MSI GeForce GTX 970 Gaming 4G is its proprietary TWIN FROZR V cooling system, which is capable of operating in both active and passive mode fan operation. This design is based on a massive radiator with two heat pipes with a diameter of 6 mm and two heat pipes with a diameter of 8 mm. This metal design quickly removes heat from the GPU and dissipates it throughout the heatsink.

Active heat dissipation from the radiator is carried out by two 100 mm fans, which have a multi-stage operating mode, controlled fully automatically. This includes support for complete fan stop on the MSI GeForce GTX 970 Gaming 4G. This usually happens when there is no load, and according to the manufacturer, the fans on the MSI GeForce GTX 970 Gaming 4G do not become active until the GPU temperature threshold reaches 64 degrees Celsius. However, we will check this issue during practical testing.

The plastic casing of the cooling system carries not only an aesthetic function, but also an illuminated logo area. Therefore, the casing is connected to the printed circuit board of the video card using a separate 2-pin backlight power connector.

A pair of proprietary fans with a fourteen-bladed impeller are secured to the radiator structure with three screws each.

Both fans have modified blade shapes, the design of which is referred to by the manufacturer as Torx Fan.

The fan model is designated 10010S12HH with operating characteristics of 0.4 A and 12 V. The manufacturer of these fans is Power Logic.

Test bench configuration

CPU

Intel Core i5-6600k 3.5GHz LGA 1151 (TurboBoost on)

Cooling system

RAM

G.Skill DDR4-2666 16Gb (8Gb*2)

Video card

MSI GeForce GTX 970 Gaming 4G

power unit

HDD

SATA-3 1Tb Seagate 7200 Barracuda (ST1000DM003)

ASUS PB298Q, 29" (2560x1080)

Motherboard

ASUS Maximus VIII Hero

Thermal interface

Gelid GC-Extreme

operating system

Windows 7 x64 SP1

Other software

CPU-Z ROG 1.74, Aida64

At the core test platform we used a bunch of Intel processor Core i5-6600k and ASUS Maximus VIII Hero motherboard.

For the MSI GeForce GTX 970 GAMING 4G video card, the latest ForceWare driver version 361.43 was installed at the time of testing. Testing was carried out on the Windows 7 x64 operating system.

Performance Evaluation

Such synthetic test packages as 3D Mark13, Valley Benchmark and Heaven Benchmark were called upon to reveal the issue of evaluating the performance of the MSI GeForce GTX 970 GAMING 4G, in which the tested graphics accelerator showed quite high results.


The issue of gaming performance was left to the judgment of such popular games as Far Cry 4, Dying Light, Wolfenstein The New Order, Battlefiled 3 and Batman Arkham Origins. In all these games, maximum use was made of possible settings graphics and resolution 1920x1080 pixels.

Analyzing the results obtained, it is worth noting that the performance of the MSI GeForce GTX 970 GAMING 4G video card in a resolution of 1920x1080 pixels is enough for any of the games tested at maximum graphics settings. This state of affairs allows us to say that this graphics accelerator is more than enough for a modern gaming computer with a FullHD monitor.

Temperature and power consumption

The Maxwell architecture and video cards based on it have already accustomed us to low heating and power consumption. The MSI GeForce GTX 970 GAMING 4G video card is once again a clear confirmation of this. Possessing great system Cooling TWIN FROZR V This graphics accelerator operates in passive mode without load and heats up to only 33 degrees Celsius. Of course, when the fans are stopped, there is no noise coming from the computer at all.

With the advent of 100% load on the video card, the temperature graphics chip MSI GeForce GTX 970 GAMING 4G grows in passive mode of operation of the cooling system up to 64 degrees Celsius, after which the cooling system fans are started according to a given algorithm. As a result, in a good ventilated case, the MSI GeForce GTX 970 GAMING 4G video card does not heat up in games above 64 degrees Celsius, and throughout the entire load, even despite the fans running at about 1000 rpm, the noise from the graphics accelerator is also not audible at all.

The MSI GeForce GTX 970 GAMING 4G video card performs excellently in terms of power consumption of the gaming system. Having a low appetite, the system based on Intel Core i5-6600k and MSI GeForce GTX 970 GAMING 4G graphics cards consume less than 300 W under load. Such modest numbers allow you not to worry about the power of the power supply when using such a video card.

Conclusion

The participant in this review can be accurately described as follows: if you use a GeForce GTX 970 video card for games, that’s already good, but if you use an MSI GeForce GTX 970 GAMING 4G, that’s great! Possessing an absolutely excellent combination of positive features and the absence of significant drawbacks, this graphics accelerator can easily lay claim to the title of one of the best video cards among the GeForce GTX 970. Of course, the MSI GeForce GTX 970 GAMING 4G is not the only example of an excellent video card of this kind, but what this video card It turned out great, there’s no denying it. The strengths of the MSI GeForce GTX 970 GAMING 4G will be its proprietary cooling system with semi-passive operation, high energy efficiency, excellent performance and striking appearance.

Pros:

  • 3 year warranty;
  • beautifully designed appearance of the cooling system;
  • high factory overclocking;
  • high efficiency of the cooling system;
  • practically silent operation video cards thanks to the semi-passive operation of the cooling system;
  • low GPU heating;
  • low power consumption;
  • high performance in modern games.

Minuses:

  • Higher price compared to other models of video cards for the GeForce GTX 970 cooling system.

For excellent consumer qualities of the graphics accelerator MSI GeForce GTX 970 GAMING 4G receives a fair assessment from our editors - gold!

Gainward GeForce GTX 970 Phoenix Test results One page

We continue to compensate for the lack of reviews of some current video cards on our website. The GeForce GTX 970 was periodically present in various tests for individual games. It's time for a summary review. We will describe general features GeForce GTX 970 and consider specific model performed by Gainward. Tests will be carried out at the recommended frequencies, with factory overclocking and at the maximum frequencies that we managed to achieve on this card. We will compare the performance of the GeForce GTX 970 with some flagship solutions and competitors from AMD.

The GeForce GTX 970 video card is a junior solution based on the GM204 graphics processor of the second generation Maxwell architecture, being essentially a stripped-down version of the GeForce GTX 980. There is an unpleasant story associated with this card when changes were made to the official specifications after the product was released. The product itself did not change in any way; NVIDIA simply did not initially disclose all its technical features.

In the full version, GM204 operates with 2048 CUDA-cores stream processors and 128 texture units, which are organized into two clusters of 8 SMM units each. In the GeForce GTX 970, 3 out of 16 SMMs are disabled, which gives 1664 active CUDA cores and 104 texture units. The 256-bit memory bus remained unchanged. But, as it turned out later, the GeForce GTX 970 is a little simplified in terms of ROPs - there are 56 of them instead of 64. At the same time, the associated L2 cache memory blocks were cut off, and instead of 2 MB, the asset of the younger card is 1792 KB. One of the eight 64-bit memory controllers has lost its allocated cache area. It works with the last memory segment. As a result, of the 4 GB of video buffer in the GeForce GTX 970, 3.5 GB work at maximum bandwidth, and the last 512 MB act as an additional segment, and the speed of operation with it is significantly lower. To simplify, we can talk about 3.5 GB effective memory and 0.5 GB of auxiliary, which is not always used.

The general organization of such a structure is depicted in a block diagram. For clarity, you can give a more familiar block diagram of the GM204 core, which you may have seen in various reviews. On it we deleted the deactivated blocks.

Along with some simplifications, the GeForce GTX 970 managed to achieve high energy efficiency and a TDP of 145 W. The core Boost frequency is stated at an average level of 1178 MHz with a base value of 1051 MHz. The effective frequency of memory chips is 7 GHz. Compared to the GeForce GTX 980, the younger card is 23% weaker in terms of main computing units, and in terms of ROP the difference is 14%. The GPU frequency is also slightly lower. For better clarity, let's compare the characteristics of these two video cards in one table.

Architecture

GPU codename

Number of transistors, million

Technical process, nm

Core area, sq. mm

Number of texture blocks

Number of ROP blocks

Core frequency, MHz

Memory bus, bit

Memory type

Memory capacity, MB

Interface

TDP level, W

Now let's take a look at the specific Gainward GeForce GTX 970 Phoenix model.

The video card comes in a large box. Supplied with power adapter, DVI/D-Sub adapter and software disc.

The Gainward GeForce GTX 970 Phoenix is ​​a massive device with a cooling system that occupies more than two slots. The cooler uses two fans. The plastic case is covered with decorative elements in red and silver colors.

The card has two power connectors, which are located closer to the center. Moreover, the board itself is full-size. Gainward is protected by a warranty seal on one of the screws. For this reason, we did not have the opportunity to disassemble Phoenix and examine the board in detail.

There are four interface connectors for image output: HDMI, two DisplayPorts, DVI.

You can evaluate the design of the cooler without disassembling it. The plastic case is removed separately. Underneath is a large radiator with three copper heat pipes.

The radiator has a “window” for the power connectors. In the area of ​​the power supply there is a finned radiator with transistors.

Airflow is provided by two Apistek fans with a diameter of 92 mm. They are screwed to a plastic removable housing.

The operating core frequencies of the Gainward GeForce GTX 970 Phoenix are higher than the recommended level. Base value 1152 MHz with Boost Clock 1304 MHz. Memory frequencies remained at the standard level of 7012 MHz.

The Boost frequency is determined by power restrictions and temperature limits. If they are not exceeded, the frequencies are higher than the declared Boost Clock level. As it heats up or exceeds power limits, the frequency may drop to more low values. Our test unit had a peak Boost of 1367 MHz. The decreases relative to the maximum are small, in some applications only by 12-25 MHz, in some games more. For example, during the Metro Last Light benchmark, the core frequency sometimes dropped to almost 1300 MHz, but not lower than the declared Boost Clock.

The noise level is very moderate. The temperature was kept within 75 °C.

Overclocking was carried out through the MSU Afterburner utility. The base frequency was increased to 1277 MHz, allowing the boost to reach 1505 MHz after a slight increase in voltage. The memory was overclocked to 8100 MHz.

To maintain stability, we had to increase the fan speed, which is already fraught with the appearance of a certain hum. If the issue of noise is extremely critical, then without increasing the voltage and at a lower overclock, the video card will be able to operate at the same noise level as at nominal.

We are also interested in the performance of the simplest GeForce versions GTX 970. Based on the results of studying other reviews, we can come to the conclusion that the Boost frequency largely depends on cooling and operating conditions, even if specifications The same for different video cards. In the end we chose this base frequency so that the peak value does not exceed 1230 MHz. At the same time, they lowered the power limit, which reduced the core Boost to 1177 MHz in the most difficult moments. In simpler games, the resulting Boost value was closer to the maximum.

With these settings, the Gainward video card will replace the reference version in our comparison.

Characteristics of test participants

An old flagship in the form of the GeForce GTX 780 Ti has been added to testing. This video adapter is equipped with less memory, but can boast a 384-bit bus. It's interesting to compare its capabilities with the GeForce GTX 970 in modern games. Previously, we compared the flagship cards of two generations from NVIDIA and AMD in a separate review, but the junior model on the GM204 was not included in it. For clarity, the results of the GeForce GTX 980 Ti have been added. This is the most powerful serial card. Its default level is higher. But I wonder what the gap will be with the extremely overclocked GeForce GTX 970 card.

Among AMD's competitors are top solutions based on the Grenada (Hawaii) processor - Radeon R9 390X and Radeon R9 290X, plus Radeon R9 Fury.

Only the Gainward GeForce GTX 970 graphics accelerator was tested for overclocking; all other participants were tested at the recommended frequencies.

The technical characteristics of all participants are indicated in the table. The official data for Boost is given for frequencies. The performance charts show GeForce base and peak frequencies.

GeForce GTX 980 Ti

Gainward GeForce GTX 970 Phoenix

GeForce GTX 780 Ti

Architecture

GPU codename

Number of transistors, million

Technical process, nm

Core area, sq. mm

Number of stream processors

Number of texture blocks

Number of ROP blocks

Core frequency, MHz

Memory bus, bit

Memory type

Effective memory frequency, MHz

Memory capacity, MB

Interface

TDP level, W

Teststand

  • Processor: Intel Core i7-3930K @4.4 GHz
  • motherboard: ASUS Rampage IV Formula
  • memory: Kingston KHX2133C11D3K4/16GX, 1866 MHz, 4x4 GB
  • hard drive: Hitachi HDS721010CLA332, 1 TB
  • power supply: Seasonic SS-750KM
  • operating system: Windows 7 Ultimate SP1 x64
  • GeForce driver: NVIDIA GeForce 359.06
  • Radeon driver: AMD Radeon Crimson Edition 15.11

Tests were carried out in a resolution of 2560x1440. The diagrams on the left indicate the minimum fps, on the right the average fps. The list of tests is not very large, so the description of the testing methodology was not included in a separate block of the article. All descriptions and necessary notes appear before the results.


Gigabyte GeForce GTX 970 WindForce SuperOC 4096 MB 256-bit GDDR5 PCI-E
ParameterMeaningNominal value (reference)
GPUGeForce GTX 970 (GM204)
InterfacePCI Express x16
GPU operating frequency (ROPs), MHz1178-1380 1050-1178
Memory operating frequency (physical (effective)), MHz1750 (7000) 1750 (7000)
Memory bus width, bits256
Number of computational units in the GPU/block operating frequency, MHz13/1178-1380 13/1050-1178
Number of operations (ALU) per block128
Total number of operations (ALU)1664
Number of texturing units (BLF/TLF/ANIS)104
Number of rasterization units (ROP)64
Dimensions, mm300×105×35270×100×36
Number of slots in the system unit occupied by a video card2 2
PCB colorblackblack
Power consumption (peak 3D/2D/sleep), W159/68/21 147/62/22
Output jacks1×DVI (Dual-Link/HDMI), 1×DVI (Single-Link/HDMI), 1×HDMI 2.0, 3×DisplayPort 1.21×DVI (Dual-Link/HDMI), 1×HDMI 2.0, 3×DisplayPort 1.2
Multiprocessing supportSLI
Maximum number of receivers/monitors for simultaneous image output4 4
Additional power: number of 8-pin connectors1 No
Additional power: number of 6-pin connectors1 2
Maximum 2D resolution: DP/HDMI/Dual-Link DVI/Single-Link DVI
Maximum 3D resolution: DP/HDMI/Dual-Link DVI/Single-Link DVI3840×2400/3840×2400/2560×1600/1920×1200

Equipped with local memory

The card has 4096 MB of GDDR5 SDRAM memory, housed in 8 4 Gbit chips (4 on each side of the PCB).

We used examples from the SDKs as synthetic DirectX 11 tests Microsoft companies and AMD, as well as a demo program from Nvidia. First, there are HDRToneMappingCS11.exe and NBodyGravityCS11.exe from the DirectX SDK (February 2010). We also took applications from both video chip manufacturers: Nvidia and AMD. Examples were taken from the ATI Radeon SDK (they are also in the DirectX SDK). Additionally, Nvidia's demo program, also known as Island11, was used.

Synthetic tests were carried out on the following video cards:

  • GeForce GTX 970 GTX 970)
  • GeForce GTX 980 with standard parameters (abbreviated GTX 980)
  • GeForce GTX 770 with standard parameters (abbreviated GTX 770)
  • Radeon R9 290X with standard parameters (abbreviated R9 290X)
  • Radeon R9 290 with standard parameters (abbreviated R9 290)

To analyze the performance of the new Geforce GTX 970 video card model, we chose these solutions for the following reasons. The Geforce GTX 980 is also Nvidia's latest model based on the same GM204 GPU, but in a full configuration, and it will be interesting to see the difference between the two. And from the previous family, for comparison, we took the Geforce GTX 770 model, which is the predecessor of the new product in question in the same segment, and is based on the GK104 graphics processor, similar in positioning to the GM204, but with the previous Kepler architecture.

From rival AMD, for our comparison we chose two video cards based on the same graphics processor - Hawaii, but differing in speed. The Radeon R9 290 is the main price competitor for the GeForce GTX 970, and we will compare them from a market point of view, but the Radeon R9 290X will be the solution that the new product should ideally strive for, although this AMD video card is noticeably more expensive and consumes much more energy .

Direct3D 10: PS 4.0 pixel shader tests (texturing, loops)

We abandoned the outdated DirectX 9 tests, since all powerful solutions have not very impressive results in them, being seriously limited by memory bandwidth, fill rate or texturing. In addition, some video cards do not work correctly in such applications, and Nvidia has long since stopped optimizing D3D9 applications.

The second version of RightMark3D included two already familiar PS 3.0 tests for Direct3D 9, which were rewritten for DirectX 10, as well as two more new tests. The first pair added the ability to enable self-shadowing and shader supersampling, which further increases the load on video chips.

These tests measure the performance of pixel shaders running in cycles with a large number of texture samples (in the heaviest mode, up to several hundred samples per pixel) and a relatively small ALU load. In other words, they measure the speed of texture samples and the efficiency of branches in the pixel shader.

The first test of pixel shaders will be Fur. At the lowest settings, it uses 15 to 30 texture samples from the height map and two samples from the main texture. The Effect detail mode - “High” increases the number of samples to 40-80, the inclusion of “shader” supersampling - up to 60-120 samples, and the “High” mode together with SSAA is characterized by maximum “heaviness” - from 160 to 320 samples from the height map.

Let's first check the modes without supersampling enabled; they are relatively simple, and the ratio of results in the “Low” and “High” modes should be approximately the same.

Performance in the test depends on the number and efficiency of TMUs, and the efficiency of executing complex programs also affects. And in the version without supersampling, the effective fill rate and memory bandwidth also have an additional impact on performance. The results at the “High” level of detail are up to one and a half times lower than at the “Low” level.

In procedural fur rendering tasks with a large number of texture samples, with the release of video chips based on the GCN architecture, AMD took the lead in performance, and Radeon boards are now leaders in these comparisons, which indicates a significantly higher efficiency in executing these programs. This conclusion is confirmed by today's comparison - the recently introduced Nvidia video card lost to both competitors from AMD, including its main rival, the Radeon R9 290.

In our first test, the new Geforce GTX 970 video card turned out to be clearly faster than its price equivalent from the previous generation in the form of the GTX 770, but the advantage is not very large. The result can be explained by the more efficient execution of this task by the new GPU and a noticeably higher fill rate, since the texturing speed of the GTX 970 is not too high. The lag of 16% from the GTX 980 speaks in favor of the emphasis on the performance of the TMU and ALU units.

Let's look at the result of the same test, but with shader supersampling enabled, which increases the work by four times: perhaps in this situation something will change, and memory bandwidth with fill rate will have less effect:

In the changed conditions, the new Geforce GTX 970 video card is also ahead of the similar model from the previous generation - the GTX 770, finding itself between the latter and the GTX 980. It lags behind its more productive brother on the GM204 by 15-18%, and the gap is behind its main price competitor in the form of Radeon The R9 290 has only increased - AMD chips are clearly strong in such calculations, since they prefer per-pixel calculations.

The next DX10 test measures the performance of complex pixel shaders with loops with a large number of texture samples and is called Steep Parallax Mapping. At low settings it uses 10 to 50 texture samples from the height map and three samples from the main textures. Enabling heavy mode with self-shadowing doubles the number of samples, and supersampling quadruples this number. The most complex test mode with supersampling and self-shadowing selects from 80 to 400 texture values, that is, eight times more than the simple mode. Let's check first simple options without supersampling:

The second Direct3D 10 pixel shader test is more interesting from a practical point of view, since types of parallax mapping are widely used in games, and heavy options, like steep parallax mapping, have long been used in many projects, for example, in the games of the Crysis and Lost Planet series. In addition, in our test, in addition to supersampling, you can enable self-shadowing, which approximately doubles the load on the video chip - this mode is called “High”.

The diagram is very similar to the previous one (also without supersampling enabled), and this time the new GeForce GTX 970 turned out to be closer to the GTX 980 than to the GTX 770. The new product continues to be noticeably ahead of the previous generation model from Nvidia, and loses to the older one a little less - 14% . Looks like new graphics Nvidia processor in pixel shaders it does the job better than previous solutions.

Well, if we compare it with AMD video cards, then in this case everything is approximately the same as it was earlier in the previous test. The new GTX 970 is inferior to both the Radeon R9 290X and R9 290, although it comes somewhat closer to its price competitor. Let's see what difference enabling supersampling will make:

When supersampling and self-shadowing are enabled, the task becomes even more difficult, and enabling both options together increases the load on the cards by almost eight times, causing a serious drop in performance. The difference between the speed performance of the tested video cards has changed slightly; turning on supersampling has less of an impact than in the previous case.

Both AMD Radeon graphics solutions are clearly more efficient in this D3D10 test of pixel shaders compared to competing Geforce boards, but the new GM204 chip has slightly changed the situation for the better - the previously presented second of the top-end Geforce GTX 970 boards on the second Maxwell architecture chip has become noticeably faster than the previous GeForce GTX 770 and in the most difficult conditions lags behind its rival Radeon R9 290 by only 10%.

Direct3D 10: PS 4.0 Pixel Shader Tests (Compute)

The next couple of pixel shader tests contain a minimum number of texture fetches to reduce the performance impact of the TMU units. They use a large number of arithmetic operations, and they measure precisely the mathematical performance of video chips, the speed of execution of arithmetic instructions in a pixel shader.

The first math test is Mineral. This is a complex procedural texturing test that uses only two samples of texture data and 65 sin and cos instructions.

The results of limiting mathematical tests most often only approximately correspond to the difference in frequencies and the number of computational units, since the results are also influenced by the different efficiency of their use in specific tasks, and driver optimization, and the latest frequency and power management systems, and even an emphasis on memory bandwidth, especially in the case of the fastest solutions. In the case of the Mineral test, the new GeForce GTX 970 model turned out to be noticeably faster than the GTX 770 board from the previous generation, which it replaced in Nvidia's line. And it lags behind the older GTX 980 by only 12%.

Moreover, even a comparison of the GeForce GTX 970 with its competing board from AMD turns out to be in favor of the former! Updating Nvidia's architecture from Kepler to Maxwell has led to the fact that their new chip even has an advantage over competing solutions from AMD in such tests. In this mathematical test, the Radeon R9 290 video card, competing with the new product in price, was slightly behind, but still behind, which can be called an excellent result for the second board on the GM204 chip.

Let's look at the second shader calculation test, which is called Fire. It is heavier for an ALU, and there is only one texture fetch, and the number of sin and cos instructions has been doubled, to 130. Let's see what has changed with increasing load:

And in our second mathematical test from RigthMark, we see approximately the same results from video cards relative to each other - the difference between the tests turned out to be insignificant. Thus, the new GeForce GTX 970 is still far ahead of the GTX 770, lagging behind the GTX 980 by the same 12%. Yes, and the Radeon R9 290 is still the same, albeit slightly, but still lags behind the new product produced by Nvidia. In peak mathematical performance tests, the new Nvidia video chip performs just fine!

Direct3D 10: geometry shader tests

The RightMark3D 2.0 package has two geometry shader speed tests, the first option is called “Galaxy”, a technique similar to “point sprites” from previous versions of Direct3D. It animates a particle system on the GPU, a geometry shader from each point creates four vertices that form a particle. Similar algorithms should obtain wide use in future DirectX 10 games.

Changing the balancing in geometry shader tests does not affect the final rendering result, the final image is always exactly the same, only the methods of processing the scene change. The “GS load” parameter determines which shader the calculations are performed in - vertex or geometry. The number of calculations is always the same.

Let's look at the first version of the Galaxy test, with calculations in the vertex shader, for three levels of geometric complexity:

The ratio of speeds for different geometric complexity of scenes is approximately the same for all solutions, performance corresponds to the number of points, with each step the FPS drop is close to twofold. This task is very simple for powerful modern video cards, and performance is limited by the speed of geometry processing, and sometimes by memory bandwidth and/or fill rate.

The difference between the results of video cards based on Nvidia and AMD chips is due to differences in the geometric pipelines of the chips of these companies, and it is quite noticeable. If in previous tests with pixel shaders AMD boards coped with the work no worse, or even better solutions Nvidia, the very first geometry tests show that Nvidia boards are very competitive in such tasks.

The latest model of the GeForce GTX 970 video card is almost twice as fast as the similar GTX 770 from the previous generation, and the lag behind the GTX 980 on the same chip is the usual 15-16%. Radeon video cards lag far behind this time, and this can be explained by the fact that the geometric performance of AMD's GPUs is worse than that of Nvidia. The new GM204 especially distinguished itself, as it copes with the work noticeably faster than even the GK110. Let's see how the situation changes when we transfer part of the calculations to the geometry shader:

When changing the load in this test, the numbers improved slightly for AMD boards and for Nvidia solutions, but the difference is small and does not change anything much. Video cards in this test of geometry shaders react weakly to changes in the GS load parameter, which is responsible for transferring part of the calculations to the geometry shader, so the conclusions remain the same.

The Nvidia board based on a stripped-down Maxwell architecture GPU is more than one and a half times faster than a similar model of the previous generation, it lags behind the GTX 980 by 13-14%, and both competitor solutions (Radeon R9 290X and R9 290) remain far behind.

Unfortunately, “Hyperlight” is the second test of geometry shaders, demonstrating the use of several techniques at once: instancing, stream output, buffer load, which uses dynamic geometry creation by drawing into two buffers, as well as new opportunity Direct3D 10 - stream output simply does not work on all modern AMD video cards. At some point next update Catalyst drivers led to the fact that this test stopped running on boards from this company, and this error has not been corrected for several years. So in this test we consider only the results of Nvidia video cards:

Here we see approximately the same thing as in the Galaxy test - the second video card on the GM204 chip is still faster than the GTX 770, although the difference between them has already narrowed, and the lag behind the GeForce GTX 980 has dropped to 6-9%. Perhaps in hard mode something will change:

Under such conditions, the results of different Nvidia video cards changed, and the difference between the GTX 980 and GTX 970 increased to 10-13%. New GeForce Although the GTX 970 is faster than a similar board from the previous generation in the form of the GeForce GTX 770, the lag behind the latter is clearly smaller. In any case, the new product has shown itself to be very good; it copes very effectively with various programs based on geometry shaders.

Direct3D 10: texture fetch speed from vertex shaders

The Vertex Texture Fetch tests measure the speed of a large number of texture fetches from the vertex shader. The tests are essentially similar, so the ratio between the cards' results in the Earth and Waves tests should be approximately the same. Both tests use displacement mapping based on texture sample data, the only significant difference is that the Waves test uses conditional branches, while the Earth test does not.

Let's look at the first "Earth" test, first in the "Effect detail Low" mode:

Our previous studies have shown that the results of this test can be affected by both fill rate and memory bandwidth, which is especially noticeable in simple mode. The new top-end Nvidia video card in it shows speeds higher than the GTX 770 board from the previous generation, which is similar in market positioning, and in easy mode it clearly runs into bandwidth problems. The gap from the older GTX 980 was 6-9%.

The main price competitor of the new product, the Radeon R9 290, is this time slightly ahead of the Nvidia board, especially in easy mode. The difference in the other two modes is not too great, and in the most difficult one it is only 4%. Let's look at the performance in the same test with an increased number of texture samples:

The situation in the diagram has changed noticeably; AMD's solutions in heavy modes clearly lost more than GeForce boards, and significantly. The new Geforce GTX 970 video card showed speeds higher than both competitors (with the exception of the lightest mode, where it was limited by memory bandwidth). If we compare the new product with the solution of the same company from the previous generation, then the GTX 970 is ahead of the GTX 770 board by a good margin. Well, the lag behind the GTX 980 is the usual 6-8%.

Let's look at the results of the second test of texture fetches from vertex shaders. The Waves test has a smaller number of samples, but it uses conditional jumps. Number of bilinear texture samples per in this case up to 14 (“Effect detail Low”) or up to 24 (“Effect detail High”) per vertex. The complexity of the geometry changes in the same way as the previous test.

The results in the second "Waves" vertex texturing test are not too similar to what we saw in the previous charts. The speed performance of all GeForces in this test deteriorated greatly, and although the new model Nvidia GeForce The GTX 970 is slightly faster than the GTX 770, but the difference is negligible, and the new model loses up to 19% to the older model GTX 980. If we compare it with competitors, both Radeon R9 290 series boards were able to show better performance in this test. Let's consider the second version of the same problem:

As the task became more complex in the second texture fetch test, the speed of all solutions became lower, and Nvidia video cards suffered the most. Not much changes in the findings, but the new GeForce GTX 970 model lags behind the GTX 980 by as much as 17-23%, and even in comparison with its predecessor from the GeForce GTX 700 series, the new product loses - this is clearly due to insufficient texturing speed. And the new Radeon R9 290, which costs the same money, was outperformed in this test by almost half.

3DMark Vantage: Feature tests

Synthetic tests from the 3DMark Vantage package will show us what we previously missed. Feature tests from this test package support DirectX 10 and are interesting in that they differ from ours and are still relevant. When analyzing the results of the GeForce GTX 970 video card in this package, we will draw some new and useful conclusions that eluded us in tests from the RightMark family of packages. Feature Test 1: Texture Fill

The first test measures the performance of texture fetch blocks. This involves filling a rectangle with values ​​read from a small texture using multiple texture coordinates that change every frame.

The performance of AMD and Nvidia video cards in Futuremark's texture test is quite high and the final figures for different models are close to the corresponding theoretical parameters. Alas, the difference in speed between the GTX 970 and GTX 770 is in favor of the previous generation solution, which is easily explained theoretically - the new product has fewer TMU units. If we compare the GTX 970 with the GTX 980, the model we are considering today, based on the stripped-down GM204 chip, is inferior in texturing by 16%, which is less than the theoretical difference.

As for comparing the texturing speed of Nvidia's new top-end video card with a competitor's solution of a similar price, the new product is also inferior to its main rival in the price niche - the Radeon R9 290, and very much so. Still, the relatively small number of texture units in the GM204 cannot be covered by the high frequency of GPU operation. Feature Test 2: Color Fill

The second task is a fill rate test. It uses a very simple pixel shader that does not limit performance. The interpolated color value is written to an off-screen buffer (render target) using alpha blending. The 16-bit off-screen buffer of the FP16 format is used, which is most often used in games that use HDR rendering, so this test is quite timely.

The numbers in the second subtest of 3DMark Vantage show the performance of ROP units, sometimes taking into account the amount of video memory bandwidth (the so-called “effective fill rate”), so the test measures throughput and/or ROP performance, and this time it looks like the latter. The result of the top GeForce GTX 970 board being reviewed today is much (more than one and a half times) better than that of the GTX 770, and the new product was also inferior to the older GTX 980 - 17%.

If we compare the scene filling speed of the new Geforce GTX 970 video card with AMD video cards, then the board we are considering today in this test shows a scene filling speed that exceeds that which is capable of the identically priced Radeon R9 290 and even the more expensive Radeon R9 290X - all- However, optimizations for the efficiency of frame buffer data compression and a large number of ROP blocks did their job. Feature Test 3: Parallax Occlusion Mapping

One of the most interesting feature tests, since a similar technique is already used in games. It draws one quadrilateral (more precisely, two triangles) using a special Parallax Occlusion Mapping technique that simulates complex geometry. Quite resource-intensive ray tracing operations and a high-resolution depth map are used. This surface is also shaded using the heavy Strauss algorithm. This is a test of a very complex pixel shader that is heavy for a video chip, containing numerous texture samples during ray tracing, dynamic branching and complex lighting calculations according to Strauss.

This test from the 3DMark Vantage package differs from the ones we conducted earlier in that the results depend not solely on the speed of mathematical calculations, the efficiency of branch execution or the speed of texture samples, but on several parameters simultaneously. To achieve high speed in this task, the correct balance of the GPU is important, as well as the efficiency of executing complex shaders.

In this case, both mathematical and texture performance are important, and in this “synthetics” from 3DMark Vantage, the new Geforce GTX 970 board turned out to be almost 40% faster than its counterpart based on the Kepler architecture GPU, although it was 17% behind the GTX 980 based on the same GM204 chip, but in its full version.

The new product is slightly (13%) inferior to its rival in price in the form of the AMD Radeon R9 290, which is explained by the fact that AMD graphics processors work very effectively in this task, and now new boards based on the GM204 graphics processor have come close to the Radeon at close range. Feature Test 4: GPU Cloth

The fourth test is interesting because it calculates physical interactions (fabric imitation) using a video chip. Vertex simulation is used, using the combined work of vertex and geometry shaders, with several passes. Use stream out to transfer vertices from one simulation pass to another. Thus, the execution performance of vertex and geometry shaders and the stream out speed are tested.

The rendering speed in this test also depends on several parameters at once, and the main influencing factors should be geometry processing performance and the efficiency of geometry shaders. That is, strengths Nvidia chips should have appeared right here, but no - all Nvidia boards showed not too high (and the same!) speed. As a result, the new GeForce GTX 970 model showed results approximately on par with the GTX 770 and GTX 980 - something is clearly wrong here.

Comparison with Radeon boards in this test leads to clear conclusions - despite the smaller number of geometric execution units and the lag in geometric performance compared to competing Nvidia solutions, both Radeon boards in this test work very efficiently and outperform all three Geforce boards presented in comparison. Feature Test 5: GPU Particles

Test of physical simulation of effects based on particle systems calculated using a video chip. Vertex simulation is also used, each vertex representing a single particle. Stream out is used for the same purpose as in the previous test. Several hundred thousand particles are calculated, all are animated separately, and their collisions with the height map are also calculated.

Similar to one of our RightMark3D 2.0 tests, particles are rendered using a geometry shader that creates four vertices from each point to form a particle. But the test mostly loads shader units with vertex calculations; stream out is also tested.

In the second “geometric” test from 3DMark Vantage, the situation has changed significantly, this time the new GeForce GTX 970 model performs noticeably better - the new product is 23% faster than its predecessor GTX 770, and 14% behind the older GTX 980.

Comparison of the new Nvidia product with competing video cards from AMD this time is much more positive, because both cards from the rival company show slightly worse results, the new product is 5% ahead of the Radeon R9 290, and even the more expensive model in the form of the Radeon R9 290X is also very slightly a little behind. The result: boards on chips AMD is better perform a synthetic tissue imitation test from the 3DMark Vantage test package, while Nvidia retains the particle calculation test, although both tasks use geometry shaders. Feature Test 6: Perlin Noise

Well, the last feature test Vantage package is a mathematically intensive test of the video chip, it calculates several octaves of the Perlin noise algorithm in the pixel shader. Every color channel uses its own noise function to put more stress on the video chip. Perlin noise is a standard algorithm often used in procedural texturing and uses a lot of math.

In this case, the performance of the solutions is not entirely consistent with the theory, but it is close to what we have seen in other math tests. In the mathematical test from the Futuremark package, which shows the peak performance of video chips in extreme tasks, we see a different distribution of results compared to similar tests from our test package. The main difference is that the result of the new Geforce GTX 970 is now lower than that of the Radeon R9 290.

AMD video chips with GCN architecture still cope with such tasks better than their competitors, especially in cases where intensive “mathematics” is performed. Nvidia's second model from the GeForce GTX 900 family showed a good result in this test, outperforming the GeForce GTX 770 by as much as 42% and lagging behind the GTX 980 by only 13% (with a slightly larger theoretical difference), but this was not enough to catch up with the Radeon R9 290, which is the main rival of the new product in terms of price. The difference between them turned out to be about 10%, in favor of AMD's solution.

Direct3D 11: Compute Shaders

To test Nvidia's today-announced solution using DirectX 11 features like tessellation and compute shaders, we used samples from the SDKs and demos from Microsoft, Nvidia, and AMD.

First we'll look at tests that use Compute shaders. Their appearance is one of the most important innovations in latest versions DX API, they are already used in modern games to perform various tasks: post-processing, simulations, etc. The first test shows an example of HDR rendering with tone mapping from the DirectX SDK, with post-processing using pixel and compute shaders.

The speed of calculations in computational and pixel shaders for all AMD and Nvidia boards has long been approximately the same; differences were observed only in video cards based on GPUs of previous architectures. Judging by our previous tests, the results in a problem sometimes depend not so much on mathematical power and computational efficiency, but on other factors, such as memory bandwidth and performance of ROP blocks.

In this case, the speed of video cards still depends most of all on mathematical performance. The new board from Nvidia in this test is slightly ahead of its predecessor, the GeForce GTX 770, and is only 12% behind the GTX 980, which can be explained by theory. If we compare the GeForce GTX 970 with its direct price competitor in the form of the Radeon R9 290, then the board we are considering today from the Californian company lags even a little further behind it.

The second test of computational shaders is also taken from the Microsoft DirectX SDK, it shows the computational problem of N-body gravity - simulation dynamic system particles that are affected by physical forces such as gravity.

This test most often focuses on the speed of execution of complex mathematical calculations, geometry processing, and the efficiency of executing code with branches. And in the second DX11 test, the balance of power between the solutions of two different companies turned out to be completely different.

Nvidia's solutions this time are clearly stronger than their rival's motherboards, and the GeForce GTX 970 surprised by a large gap from the older model - it is already 27% slower than its older sister GTX 980. It seems that the seriously reduced capabilities in geometric and mathematical performance are having an effect.

Direct3D 11: Tessellation Performance

Compute shaders are very important, but another important innovation in Direct3D 11 is hardware tessellation. We looked at it in great detail in our theoretical article about the Nvidia GF100. Tessellation has been used for quite some time in DX11 games, such as STALKER: Call of Pripyat, DiRT 2, Aliens vs Predator, Metro Last Light, Civilization V, Crysis 3, Battlefield 3 and others. Some of them use tessellation for character models, others use it to simulate realistic water surfaces or landscapes.

There are several various schemes partitions graphic primitives(tessellation). For example, phong tessellation, PN triangles, Catmull-Clark subdivision. Thus, the PN Triangles partitioning scheme is used in STALKER: Call of Pripyat, and in Metro 2033 - Phong tessellation. These methods are relatively quickly and easily implemented into the game development process and existing engines, which is why they have become popular.

The first tessellation test will be the Detail Tessellation example from the ATI Radeon SDK. It implements not only tessellation, but also two different pixel-by-pixel processing techniques: simple normal map overlay and parallax occlusion mapping. Well, let's compare AMD and Nvidia DX11 solutions in different conditions:

In a simple bumpmapping test, the speed of boards is not so important, since this task has long become too easy and performance is limited by bandwidth or fill rate. Today's hero of the review is slightly superior to the previous model GeForce GTX 770, inferior to the GTX 980 by more than 20%, and the competitor's solution, similar in price to the new product, in the form of the Radeon R9 290, is clearly the fastest.

In the second subtest with more complex pixel-by-pixel calculations, the new product is still faster than the GeForce GTX 770 model, is inferior to the GTX 980 on a full-fledged GM204 by about 17%, and again noticeably lags behind the competing Radeon R9 290. Still, the efficiency of performing mathematical calculations in pixel shaders on the chips GCN architectures are higher than Kepler, although Maxwell has improved the position of Nvidia solutions.

In the light tessellation subtest, the new Nvidia board is almost on par with the previous generation Geforce GTX 770, and the top-end GTX 980 is 18% faster. If we compare the new product with the only board manufactured by AMD, it is still seriously inferior, since in this tessellation test the division of triangles is very moderate and the speed is not limited by the performance of the geometry processing units.

The second tessellation performance test will be another example for 3D developers from the ATI Radeon SDK - PN Triangles. Actually, both examples are also included in the DX SDK, so we are sure that game developers create their code based on them. We tested this example with different coefficient tessellation factor to understand how much impact its change has on overall performance.

This heavy-duty test uses more complex geometry, so comparing the geometric power of different solutions brings completely different conclusions. All modern solutions presented in the material cope quite well with light and medium geometric loads, showing high speed. And although in light conditions the AMD graphics processor in the Radeon R9 290 works great, even ahead of its Californian competitors, who are clearly stuck on something, in difficult conditions Nvidia boards come out ahead with a large margin. In the most difficult modes, the GeForce GTX 970 introduced last month shows noticeably better speed than the only Radeon.

As for comparing Nvidia boards of different generations with each other, the Geforce GTX 970 model being reviewed today also increases its advantage with increasing geometric load, as it should be according to theory. As a result, it is almost twice as fast as the previous GeForce GTX 770 in the most difficult mode! And the older model on the top Maxwell chip is not that much ahead of the new product in question - the difference between them is only 7-10%.

Let's look at the results of another test - the Nvidia Realistic Water Terrain demo, also known as Island. This demo uses tessellation and displacement mapping to render realistic-looking ocean surfaces and terrain.

The Island test is not a purely synthetic test for measuring exclusively geometric GPU performance, since it contains complex pixel and computational shaders, too, and such a load is closer to real games, which use all GPU blocks, and not just geometry, as in previous geometry tests. Although the main one still remains the load on the geometry processing units.

We test all video cards at four different tessellation ratios - in this case the setting is called Dynamic Tessellation LOD. At the first triangle partitioning factor, the speed is not limited by the performance of geometric blocks, and the Radeon R9 290 video card shows high results, even surpassing the speed of the recently announced Geforce GTX 970 board, but already at the next levels of geometric load the performance of the Radeon board decreases very seriously, and Nvidia's new product comes out forward.

The advantage of the new Nvidia board on the GM204 video chip in such tests is up to two times. It’s interesting that if you compare the GeForce GTX 970 with the GTX 770, the difference between their performance reaches one and a half times, which is difficult to explain by the increased speed of geometry processing alone. It is quite possible that the matter is also that the GM204 works noticeably more efficiently in mixed load mode, quickly switching from execution to graphic tasks to computing and back. The new product lags behind the older model GTX 980 by only 11-16%, which is close to theory.

Having analyzed the results of synthetic tests of the new Nvidia Geforce GTX 970 video card, based on the GM204 graphics processor with reduced functional units, and also having considered the results of other video card models from both manufacturers of discrete video chips, we can conclude that the video card being considered today will be able to compete not only price rival from AMD in the form of the Radeon R9 290, but it can also become one of the most successful video cards in its price segment. It is quite a bit (about 7-15%) inferior to the older model GTX 980, and costs much less than it.

Nvidia's new video card shows fairly high results in "synthetic" results, close to those of the Radeon R9 290. In general, these solutions have different strengths; in some tests, the more complex AMD graphics processor is stronger, including due to higher memory bandwidth and efficient execution of pixel shaders, while in others the Maxwell architecture graphics processor is faster. Of course, the situation is real gaming applications may differ from the results of synthetic tests, but the GeForce GTX 970 will in any case become a strong competitor to the Radeon R9 290 in games.

At the same time, the Geforce GTX 970 also consumes less energy than even the Radeon R9 285, not to mention the Radeon R9 290. In terms of energy efficiency, both new models of the company (Geforce GTX 970 and GTX 980) look amazing - Nvidia engineers have achieved the highest efficiency of the new GM204 . From a market point of view, everything will depend on prices in stores and comparison of performance in gaming applications.

And just in order to draw conclusions on the real state of affairs, in the next part of our material we will determine the performance of Nvidia’s new product in gaming projects, comparing it with competitors and evaluating, including the justification of the retail price of the solution, taking into account the rendering speed of the Geforce GTX 970 and Radeon R9 290 in our gaming suite.