Video cards. Video cards Characteristics of the nvidia geforce gtx 550 card

• GPU: GeForce GTX 550 Ti (GF116)
• Interface: PCI Express x16
• 920 (nominal - 900)
• 1025 (4100) (nominal - 1025 (4100))
• 192
• 192/1840 (nominal - 192/1800)
• 192
• 32 (BLF/TLF/ANIS)
• Dimensions (mm):
• PCB color: blue
• RAMDACs/TMDS: integrated into GPU
• Output Jacks: 2×DVI (Dual-Link/HDMI), HDMI-mini
• SLI (Hardware)

• GPU: GeForce GTX 550 Ti (GF116)
• Interface: PCI Express x16
• GPU operating frequency (ROPs) (MHz): 1000 (nominal - 900)
• Memory operating frequency (physical (effective)) (MHz): 1100 (4400) (nominal - 1025 (4100))
• Memory bus width (bits): 192
• Number of computational units in the GPU/block operating frequency (MHz): 192/2000 (nominal - 192/1800)
• Number of operations (ALU) in block: 1
• Total number of operations (ALU): 192
• Number of texturing units: 32 (BLF/TLF/ANIS)
• Number of rasterization units (ROP): 24
• Dimensions (mm): 195×100×33 (the last value is the maximum thickness of the video card)
• PCB color: red
• RAMDACs/TMDS: integrated into GPU
• Output Jacks: 1×DVI (Dual-Link/HDMI), HDMI, VGA
• Multiprocessor support: SLI (Hardware)

The card has 1024 MB of GDDR5 SDRAM memory, located in six chips on the front side of the PCB. At the same time, four memory chips have a capacity of 1 Gbit each, and two chips have a capacity of 2 Gbit each.

It makes sense to remind you once again that the card requires additional power, with one 6-pin connector.

About cooling systems.

Inno3D Geforce GTX 550 Ti 1024 MB 192-bit GDDR5, PCI-E
The product has a simple CO, which is a banal rounded central radiator made of aluminum alloys. A large fan is placed above it, and the whole thing has a sort of stylized plastic frame. The fan operates at low speeds, so there is practically no noise even under heavy load. The memory chips are not cooled, only the core.
Gainward Geforce GTX 550 Ti Golden Sample 1024 MB 192-bit GDDR5, PCI-E
The cooler is essentially of a similar design, with a central radiator with a large fan in the center. And the long casing is just a design, a decoration. The fan also operates at low speeds, making it almost silent. Memory chips are also not cooled.

We conducted a temperature study using the MSI Afterburner utility (author A. Nikolaychuk AKA Unwinder) and obtained the following results:

Inno3D Geforce GTX 550 Ti 1024 MB 192-bit GDDR5, PCI-E
Gainward Geforce GTX 550 Ti Golden Sample 1024 MB 192-bit GDDR5, PCI-E

It is obvious that even with such a very significant core overclock, like the Gainward card, the heating does not exceed critical values ​​and is normal for cards of this class.

Now about the configuration. The video card from Inno3D arrived to us without packaging or kit, so we will only study the kit from Gainward.

The basic package should include a user manual, a disk with drivers and utilities.

Package.

Installation and drivers

Test bench configuration:

• Computer based on Intel Core i7-975 (Socket 1366)
  • Intel Core i7-975 processor (3340 MHz);
  • Asus P6T Deluxe motherboard based on Intel X58 chipset;
  • RAM 6 GB DDR3 SDRAM Corsair 1600 MHz;
  • hard drive WD Caviar SE WD1600JD 160 GB SATA;
  • power supply Tagan TG900-BZ 900W.
• operating system Windows 7 64 bit; DirectX 11;
• Dell 3007WFP monitor (30″);
• ATI drivers version Catalyst 11.2; Nvidia version 266.58/266.66/267.59.

VSync is disabled.

Synthetic tests

The synthetic test packages we use can be downloaded from the following links:

• D3D RightMark Beta 4 (1050) with a description on the website http://3d.rightmark.org.
• D3D RightMark Pixel Shading 2 and D3D RightMark Pixel Shading 3— tests of pixel shaders versions 2.0 and 3.0 link.
• RightMark3D 2.0 with a brief description: , .

Synthetic tests were carried out on the following video cards:

• GeForce GTX 550 Ti GTX 550);
• GeForce GTX 460 with standard parameters, model with 768 MB of video memory (hereinafter GTX 460);
• GeForce GTS 450 with standard parameters (hereinafter GTS 450);
• Radeon HD 6850 with standard parameters (hereinafter HD 6850);
• Radeon HD 5770 with standard parameters (hereinafter HD 5770).

To compare the results of the new GeForce GTX 550 Ti model, we chose these particular video cards for the following reasons: Radeon HD 6850 and Radeon HD 5770 are based on different GPUs and are the closest solutions in price from AMD, which competes with Nvidia. The Geforce GTS 450 is a video card based on a similar graphics processor from the previous generation, and the GTX 460 with 768 MB of memory is a solution similar in characteristics to the previous line, based on a more powerful GF114 chip.

Direct3D 9: Pixel Filling tests

The test determines the peak texture sampling performance (texel rate) in FFP mode for a different number of textures applied to one pixel:

In our first test, video cards traditionally show numbers that are far from theoretically possible values, and therefore we will double-check them again in a similar 3DMark Vantage test. The results of this synthetic for the GTX 550 Ti do not reach the peak values; it turns out that the new video card selects up to 21 texels per clock cycle from 32-bit textures during bilinear filtering in this test, which is significantly lower than the theoretical figure of 32 filtered texels.

As for comparison with other boards, in terms of texturing speed the new GTX 550 Ti was exactly in the middle between the GTS 450 and GTX 460. In other words, the new chip works somewhat more efficiently than its older brother from the previous line - GF104.

Well, both compared video cards from AMD are significantly ahead of all Nvidia solutions, including the new one, in all modes with any number of textures superimposed on a pixel. Interestingly, the difference is most obvious in cases with a small number of textures, although there the bandwidth limitation should have a greater effect. It seems that Radeons perform the task more efficiently, although they also do not achieve realistically possible performance. Let's look at the fill rate test:

In the second synthetic test, which shows the fill rate, everything is the same, but taking into account the number of pixels recorded in the frame buffer. And the diagram shows a strange limitation in the rendering speed of Nvidia solutions with up to four textures inclusive, while AMD video cards have only up to two or three textures per pixel.

The maximum result still remains with AMD solutions, which cope much more efficiently with our synthetic test. It’s surprising that in simple conditions the Radeon HD 5770 shows the best result, and in difficult conditions the HD 6850 comes out ahead. The Nvidia solution being reviewed today is inferior to everyone except its predecessor GTS 450, although its corresponding theoretical characteristics are close to those of the younger AMD video card.

Direct3D 9: Pixel Shaders tests

The first group of pixel shaders that we are considering is very simple for modern video chips, it includes various versions of pixel programs of relatively low complexity: 1.1, 1.4 and 2.0, found in older games.

These tests of pixel shaders of lower versions are very, very simple for modern GPUs, even of this level, and they clearly cannot show the capabilities of modern video chips. Performance in these tests is limited mostly by the speed of the texture units, taking into account block efficiency (see above) and texture data caching. There is also some influence of video memory bandwidth.

It is clearly noticeable that the GF116 repeats the results of the GF106, but taking into account the higher GPU clock speeds for the GTX 550 Ti. The new model is again located approximately in the middle between the GTS 450 and GTX 460. In the simplest shaders, the difference between the GTX 550 Ti and GTX 460 was 10-14% in favor of the old model, which is lower than the theoretical difference in power of these GPUs. It appears that the GTX 460's performance in this test is more limited by video memory bandwidth, and the GTX 550 Ti is quite close in terms of tasks.

However, if we take a comparison with AMD video cards, then again there is nothing positive - the GTX 550 Ti loses to both the HD 5770 and HD 6850 in all tests. However, the difference between them is often not that great. Let's look at the results of more complex intermediate pixel programs:

These tests are traditionally more curious. The highly texturing-speed-dependent procedural water rendering test "Water" uses dependent sampling from highly nested textures, and therefore its maps are usually ranked by texturing speed. In this test, the GTX 550 Ti shows a theoretically quite reasonable result, trailing the GTX 460 by 22%, which is slightly less than the theoretical difference. This time, the new GTX 550 Ti is closer to the GTS 450. Unfortunately, AMD video cards still remain ahead, although they are not so far ahead.

The results of the second subtest are different, although here the GTX 550 Ti is still inferior to all solutions except its younger sister GTS 450. The test is more computationally intensive, and the influence of mathematical performance is already evident in it. And therefore, the test is much better suited for AMD video cards, which have a noticeably larger number of ALU units. The difference between the GTX 550 Ti and GTS 450 in the second test was less than 20%.

Direct3D 9: pixel shader tests Pixel Shaders 2.0

These DirectX 9 pixel shader tests are more complex than the previous ones, they are close to what we now see in multi-platform games, and are divided into two categories. Let's start with the simpler version 2.0 shaders:

• Parallax Mapping- a method of texture mapping familiar to most modern games, described in detail in the article.
• Frozen Glass- a complex procedural texture of frozen glass with controllable parameters.

There are two variants of these shaders: those with a focus on mathematical calculations and those with a preference for sampling values ​​from textures. Let's consider mathematically intensive options that are more promising from the point of view of future applications:

These are universal tests that depend on both the speed of the ALU units and the texturing speed; the overall balance of the chip is important in them. The performance of video cards in the “Frozen Glass” test is similar to what we saw above in “Cook-Torrance”, and the new GTX 550 Ti is again inferior to the Geforce GTX 460 by about 15%, and both AMD solutions were also ahead as usual. The GTX 550 Ti is again roughly midway between the GTS 450 and GTX 460 in terms of speed.

In the second test “Parallax Mapping” the results are also not much different from the previous ones. Solutions based on the GF116 and GF106 chips are usually behind the rest, and the GTX 550 Ti is ahead only of the GTS 450, lagging the same distance behind the GTX 460. The difference between them is slightly higher - 17%, which confirms the fact that the new Nvidia solution is most not sufficient mathematical performance. Let's consider these same tests in a modification with a preference for samples from textures over mathematical calculations:

The position of the new solution regarding the GTX 460 and GTS 450 has not changed; they are located on a flat ladder in order: GTS 450, GTX 550 Ti, GTX 460. This time, Nvidia video cards are even a little more inferior to the HD 6850 and HD 5770. The difference between the video cards on GF104 and GF116 it turned out to be 15-20%, and compared to the mathematical analogue of the test it increased slightly. The new GTX 550 Ti also has nothing special to boast about in terms of texturing speed.

But these were all outdated tasks, mostly focusing on texturing or fillrate, and not particularly complex. Next, we'll look at the results of two more pixel shader tests - version 3.0, the most complex of our pixel shader tests for the Direct3D 9 API, which are much more indicative of modern PC games. These tests differ in that they place a greater load on both the ALU and the texture modules; both shader programs are complex and long, and include a large number of branches:

• Steep Parallax Mapping- a much more “heavy” type of parallax mapping technique, also described in the article.
• Fur— a procedural shader that renders fur.

But with tests of pixel shaders version 3.0, the new Nvidia solution does much better. Which is no wonder, since both PS 3.0 tests are quite complex, almost independent of memory bandwidth and texturing, and although they are mathematical, they have a large number of transitions and branches, which the new Nvidia architecture copes well with. Although the latest AMD video cards show very strong results here.

So, in one of the most complex Direct3D 9 tests (Fur), the GTX 550 Ti presented today showed a result noticeably higher than both Radeons, and in the second (Steep Parallax Mapping) it was at the level of the HD 5770 and slightly inferior to the HD 6850. As for comparison with other Nvidia boards, then in both tests the new solution is again located exactly between the GTS 450 on the GF106 and the GTX 460 on the GF104. The new Nvidia solution is inferior to its older brother GTX 460 by 20-23%, which is explained by the smaller number of ALU units. In general, the result for the GTX 550 Ti for this pair is not bad, because it successfully competes with both representatives from AMD.

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

The second version of RightMark3D included two 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 this test depends not only on the number and efficiency of TMUs, but also on the fill rate. The results in the “High” mode are approximately one and a half times lower than in the “Low” mode, as it should be according to theory. In Direct3D 10 tests of procedural fur rendering with a large number of texture samples, Nvidia's solutions were previously stronger, but AMD's latest architecture shows similar results.

As a result, the new Nvidia video card lags slightly behind the younger Radeon and is very much inferior to the competitor's older model, which became the test leader. This indicates a clear influence of memory bandwidth on rendering speed; this figure is maximum for the HD 6850. And in the case of the GTX 550 Ti and GTX 460, the difference in speed corresponds to the theoretical difference in the speed of ALU and TMU - just above 30%.

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 a different effect:

Enabling supersampling quadruples the theoretical load, and in this case, all Nvidia solutions lose ground, and both AMD video cards look even stronger. Now both Radeons outperform all Nvidia cards. At the same time, the HD 6850 is significantly ahead of the HD 5770. Probably, performance is still limited by ROP performance, which is noticeably higher in the HD 6850. But in the case of Nvidia cards, everything is different: the GTX 550 Ti lags behind the GTX 460, despite the higher bandwidth and fill rate.

The second test, which measures the performance of complex looped pixel shaders with a large number of texture samples, 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. When you enable heavy mode with self-shadowing, the number of samples doubles, 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 first check simple options without supersampling:

This test is more interesting from a practical point of view, since many types of parallax mapping have been used in games for a long time, and heavy variants, like our steep parallax mapping, are 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 SSAA), and the results are close even in absolute numbers. In the updated D3D10 version of this test without enabling supersampling, the new GTX 550 Ti performs this task approximately 30% slower than the GTX 460 based on the GF104 chip. In this test, it is clearly closer to the GTS 450 in speed, since the speed of texturing and mathematical calculations is closer to the younger model.

Both motherboards based on the GF106 and GF116 chips still lag behind both video cards produced by AMD, and only the GTX 460 competes with the Radeon HD 5770. Well, the HD 6850, based on the Barts chip, again becomes the clear leader. This video card has a clear advantage in terms of theoretical fill rate and memory bandwidth. Let's see what difference turning on supersampling will make.

When SSAA and self-shadowing were enabled, the task turned out to be noticeably more difficult; enabling these two options together increases the load on the cards by almost eight times, causing a large drop in performance. The difference between the speed performance of several video cards has changed, the inclusion of supersampling has the same effect as in the previous case - AMD cards have further strengthened their performance compared to Nvidia solutions.

The Radeon HD 6850 remains ahead without options, and the second place is taken by the HD 5770. Our today's hero GTX 550 Ti lags behind both competitors and is traditionally located between the GTS 450 and GTX 460. Not a very high result, although the new Nvidia video card lags behind the board on the GF104 chip exactly everything close to the theoretical 25-27%.

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.

Purely mathematical tests confirm that the GF116 GPU is architecturally no different from its predecessor GF106, the difference between the GTX 550 Ti and GTS 450 almost corresponds to the theoretical 15% in terms of comparative ALU performance. The practical gap between solutions is somewhat larger due to the influence of bandwidth, especially on the low-end model. All other solutions were also located approximately in accordance with the theoretical indicators.

AMD video cards are traditionally much faster in this synthetic test, since in computationally complex tasks the modern AMD architecture simply has a huge advantage over competing Nvidia video cards, even despite its lower efficiency. This time again, the difference between Nvidia and AMD cards remains quite large, the HD 6850 and HD 5770 show results approximately twice as high as the GTX 550 Ti, which is in line with the theory. The new model lags behind the GTX 460 by close to the theoretical 27%.

In our previous studies, we noted that this test does not entirely depend on the speed of the ALU; performance solutions are also limited by the speed of the video memory. Therefore, let's look at the second test of shader calculations, which is called Fire. It is even 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:

There are few changes in the ranking of cards, only the difference between the HD 6850 and HD 5770 has narrowed slightly. In the second math subtest, rendering speed is limited almost exclusively by shader unit performance, and the difference between the GTX 550 Ti and GTS 450 has decreased, although it remains higher than theoretical. There is only a 27% difference between the new model and the GTX 460, although in theory it should be a little more.

The result of mathematical calculations is still the same - AMD solutions have a clear advantage, explained by the large number of ALU units. Next, we will look at the results of testing geometry shaders; they will be interesting to us because the main performance limiter in them is often the speed of geometry processing, and it will be interesting to compare the GTX 550 Ti with its competitors.

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 be widely used 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. The performance corresponds to the number of points, with each step the FPS drops by about two times. The task for modern video cards is not particularly difficult; performance is limited not only by the speed of execution of geometry shaders, but also by memory bandwidth.

The Geforce GTX 550 Ti again found itself exactly in the middle between the GTX 460 and GTS 450, and the difference between the GTX 550 Ti and GTX 460 was about 22%. Compared to its competitors, the new Nvidia video card showed a result slightly higher than that of the Radeon HD 5770, but one and a half times lower than that of the HD 6850. At the same time, the HD 6850 does not have a large number of geometry processing units. This indicates a large influence on the results of the speed of ROP and/or PSP blocks.

It is also possible that the optimization of geometric blocks made by AMD engineers “worked” for the HD 6850. In this test, the speed of execution of geometry shaders was even higher than that of the GTX 460, which is very good. Let's see what changes when we transfer part of the calculations to the geometry shader:

When the load in this test changed, the numbers for the Nvidia and Radeon HD 6850 solutions remained almost unchanged, but the junior Radeon HD 5770 video card slightly improved its results. And Nvidia boards in this test do not notice changes at all in the GS load parameter, which is responsible for transferring part of the calculations to the geometry shader, and show results similar to the previous diagram. As a result, the HD 5770 is neck and neck with the GTX 550 Ti unveiled today. Let's move on to the next test, which involves a heavy load on geometry shaders.

“Hyperlight” is the second test of geometry shaders, demonstrating the use of several techniques at once: instancing, stream output, buffer load. It uses dynamic geometry creation using dual buffer rendering, as well as a new feature of Direct3D 10 - stream output. The first shader generates the direction of the rays, the speed and direction of their growth, this data is placed in a buffer, which is used by the second shader for drawing. For each point of the ray, 14 vertices are built in a circle, up to a million output points in total.

A new type of shader programs is used to generate “rays”, and with the “GS load” parameter set to “Heavy” - also to draw them. In other words, in the “Balanced” mode, geometry shaders are used only to create and “grow” rays, the output is carried out using “instancing”, and in the “Heavy” mode, the geometry shader is also involved in output. First we look at the easy mode:

The relative results in different modes again correspond to the load: in all cases, performance scales well and is close to theoretical parameters, according to which each subsequent level of “Polygon count” should be less than twice as slow.

In this test, under balanced load, the rendering speed for all solutions is less clearly limited by geometric performance, and the numbers indicate a clear emphasis on bandwidth. The new GeForce GTX 550 Ti this time is already closer to the GTX 460 (the difference is only 9-13%), and with increasing geometry complexity its lag is getting smaller and smaller. What’s interesting is that now Nvidia’s new solution shows the result exactly between two AMD video cards.

The numbers should change in the next diagram, in a test with more active use of geometry shaders. It will also be interesting to compare the results obtained in the “Balanced” and “Heavy” modes with each other.

But in this test the difference between GF116 and GF104 became closer to theoretically justified - about 25%. Unfortunately, the GF116 does not have an advantage in terms of execution speed for geometry shaders, since it has only one rasterizer, unlike two in the GF114 and GF104. And optimized geometry processing in Barts allowed the Radeon HD 6850 to be noticeably ahead.

A comparison with AMD video cards shows that the performance of the GTX 550 Ti in this test is at the level of only the younger model - the HD 5770. But the company's newer solution - the HD 6850 - unexpectedly became the test leader, even surpassing the GF104 card, which has two rasterization units.

Based on tests of geometry shaders, we conclude that the speed of rasterization and triangle setup can be a potentially weak indicator of the GF116 for overall performance in some modern games with a lot of geometry. Although the results shown are not bad for such a low-level solution, the Radeon HD 6850, which costs the same, clearly performed better. However, in tessellation tasks the speed will be limited not by rasterizers, but by tessellators, and in such cases the new GPU can show stronger results.

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 similar in essence and the ratio between the cards’ results in the “Earth” and “Waves” tests should be approximately the same. Both tests are 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:

Previous research has shown that the results of this test are affected by both texturing speed and memory bandwidth. The difference between the solutions is not very big, and they clearly show strange results, not accelerating in simple modes. The latest GTX 550 Ti is traditionally as much faster than the GTS 450 as it is slower than the GTX 460.

In difficult mode, the GTX 550 Ti outperforms both competitors from AMD (and the HD 5770 is almost twice as far behind), but in medium and easy mode the result of the new solution is clearly somewhat limited. Therefore, the HD 6850 works in medium and light modes, and the HD 5770 is faster only in light modes. Let's look at the performance in the same test with an increased number of texture samples:

The relative position of the cards on the diagram has not changed too much, but now the focus of video cards on Nvidia chips on something unknown has become even more noticeable, especially in the lightest mode. And, quite interestingly, the GTX 550 Ti is this time much closer to the GTS 450 than to the GTX 460, which is based on a different chip. The difference between solutions based on GF116 and GF104 was up to 34%, which is higher than the theoretical difference in texturing speed.

All other conclusions can simply be repeated. Since Nvidia is stronger in heavy mode, the GTX 550 Ti outperforms both AMD cards. In medium mode it outperforms only the HD 5770, and in easy mode it is completely inferior to both competitors. This obviously involved some specific optimizations of the Nvidia driver.

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. The number of bilinear texture samples in this case is 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.

But here the results are not similar to those we saw in the previous diagrams. In the “Waves” test under different conditions we already see a clear advantage of the Radeon HD 6850, but all other solutions show the expected results. Our today's hero GTX 550 Ti in this test shows performance only slightly higher than the GTS 450, and the new card on the GF116 chip lags behind the GTX 460 by as much as 36% in heavy mode. It appears that the test results are affected by caching efficiency. Let's consider the second version of the same test:

There are very few changes as the complexity of the conditions increases, and we didn’t notice anything interesting. The relative results of the GF116 graphics processor in the second vertex fetch test at high detail levels remained approximately at the level of the competing Radeon HD 5770; in heavy mode, the GTX 550 Ti is ahead, and inferior in light mode.

Well, it’s pointless to compare with the Radeon HD 6850 - this video card is clearly the strongest in this test, and this is especially noticeable in simple conditions. The difference between video cards based on GF116 and GF106 was 11-12%, which almost corresponds to the theoretical difference in texturing speed.

But as for the comparison of the GTX 460 and GTX 550 Ti, the practical difference of 37-41% is clearly higher than the theoretical 31%. It appears that performance in this test is not limited by texturing speed alone, and is already impacted by the performance of the triangle setup units.

3DMark Vantage: Feature tests

Although the synthetic tests from the 3DMark Vantage package are not new, they support D3D10 and are interesting because they differ from ours. When analyzing the results of Nvidia's new solution in this package, we will be able to draw some useful conclusions that eluded us in the RightMark family of tests.

The first test is a texture fetch speed test. This involves filling a rectangle with values ​​read from a small texture using multiple texture coordinates that change every frame.

In the Vantage texture performance test, the results are not the same as in our RightMark. And these numbers are more like the truth, and they are very close to the theory. In texture synthetics from 3DMark Vantage, all Nvidia video cards use the available texture units noticeably more efficiently, and therefore the GTX 550 Ti showed results only 20-30% worse than both Radeon video cards, which is confirmed by theoretical figures. But still, AMD video cards remain faster, having a larger number of texturing units.

As for the comparison with Nvidia video cards, the result here is quite understandable - the GTX 550 Ti outperforms its fellow GTS 450 and lags further behind the GTX 460, in full accordance with the theory. The difference between the performance of the GTX 550 Ti and the GTX 460 was more than 30%, which is almost equal to the theoretical difference in TMU performance. In general, the new video card based on the GF116 chip shows a good result, considering that Nvidia video cards are usually much inferior to their competitors in such tests.

Fill rate test. A very simple pixel shader is used 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 performance indicators in this test approximately correspond to the theoretical fill rate figures (performance of ROP blocks), but some results are also affected by the influence of video memory bandwidth. The numbers are not similar to ours because we use an integer buffer with 8 bits per component, while the Vantage test uses 16 bits of floating point.

The relative results of the announced GTX 550 Ti in this test are worse than the theoretical ones. Probably, the influence of bandwidth is on some of the solutions, since according to theory the GTX 550 Ti should be faster than the GTX 460 with 768 MB of video memory. But the new Nvidia model shows results worse than the GTX 460, which does not allow it to catch up with competitors from AMD, including the Radeon HD 5770, which has a lower theoretical fill rate. It is no wonder that there is a gigantic gap between the HD 6850 and the new Nvidia board. Well, compared to the GTS 450, the new model in this test was only 17% faster.

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.

The rendered surface is shaded using a heavy Strauss algorithm. This is a test of a very complex and heavy pixel shader for a video chip, containing numerous texture samples during ray tracing, dynamic branching and complex lighting calculations according to Strauss.

The test differs from other similar ones in that the results in it depend not exclusively on the speed of mathematical calculations or the efficiency of branch execution or the speed of texture fetches, but on a little bit of everything. And to achieve high speed, a proper balance of GPU and video memory blocks is important. Significantly affects the speed and efficiency of branching in shaders.

Unfortunately, all Nvidia chips show weak results in this test. The strongest of them (GTX 460) is not even enough for the Radeon HD 5770, not to mention the company’s simpler boards. The GTX 550 Ti unveiled today, aimed at the upper end of the lower price range, falls short of AMD's low-end selected solution by about 50%. And compared to Nvidia video cards, the situation is familiar - the new product is located exactly between the GTX 460 and GTS 450.

The 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.

And only here Nvidia solutions are rehabilitated. It seems that the rendering speed in this test is also affected by several different parameters. But still, the speed most of all depends on the performance of geometry processing and the efficiency of execution of geometry shaders. In this test, the GTX 460 is the fastest, being the clear leader in the comparison. Well, the GTX 550 Ti outperforms both AMD video cards, albeit with a slight advantage.

In this test, all Nvidia boards show better results when running complex shaders. In terms of execution of geometry shaders, speed of geometry processing and efficiency of execution of complex programs, the GF116 is doing very well. It's very interesting that the GTX 550 Ti is closer in speed to the GTX 460 than the GTS 450 in this test.

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.

The results of this test are similar to those we saw in the previous diagram. But this time the Radeon HD 6850 found reserves to improve its results. The GTX 460 remains in the leading position, but our hero GTX 550 Ti shows speed at the level of the strongest of the two AMD boards. We also note that the difference between Nvidia video cards is again somewhat unusual - the result of the GTX 550 Ti is located on the diagram closer to the result of the GTX 460, and the GTS 450 is quite far behind them. Therefore, we are inclined to attribute the difference to the lack of memory bandwidth of the company's junior solution.

In general, based on two synthetic tissue and particle simulation tests from this test package, which use geometry shaders, we can conclude that the presented Nvidia graphics processor showed a very good result, ahead of both competing video cards from AMD. Even taking into account the presence of only one rasterizer in the GF116.

Vantage's latest feature test is a math-intensive GPU test that calculates several octaves of Perlin noise in a pixel shader. Each 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 involves a lot of math.

This test from the 3DMark Vantage package measures the peak mathematical performance of video chips in extreme tasks. And the speed of all solutions shown in it approximately corresponds to what is obtained from theory and our mathematical tests from the RightMark 2.0 package.

AMD video cards again outperform all Nvidia solutions in this test. We have already seen more than once that simple but intensive mathematical calculations are performed on Radeon video cards much faster than on GeForce. So in this mathematical test, the GeForce GTX 550 Ti, based on the improved GF116 chip, shows a speed almost half as low as both direct competitors, as it should be according to theory.

Compared to other Nvidia boards, the new model is 22% faster than the GTS 450, which is less than the theoretical difference. And the lag behind the GTX 460 is 27%, which is not far from the theoretical 31%. But nothing can be done with both AMD video cards; they are clearly stronger here, as in other extreme mathematical tests.

Conclusions on synthetic tests

Based on the results of synthetic tests of the new Nvidia Geforce GTX 550 Ti model, based on the improved GF116 graphics processor, as well as the results of other video card models from both video chip manufacturers, we can conclude that the modification of the GF106 chip of the original Fermi architecture turned out to be technically good, but market success in this price range will depend more on retail prices. In many synthetic tests, Nvidia's new solution performed well, sometimes on par with the Radeon HD 6850. And yet, it also has obvious weaknesses.

The updated GPU differs from the original GF106 by an increased clock speed and an increased number of active ROP units and an expanded memory controller. These modifications led to increased performance (20-30% in most cases), and the video card presented today based on this GPU is located exactly in the middle between the GTS 450 and GTX 460 in terms of speed. Especially noteworthy is the high fill rate and bandwidth - for this parameter the GTX 550 Ti theoretically ahead of even the fairly strong GTX 460 model with 768 MB of video memory.

Among the shortcomings, we note that the performance of geometric processing in the GF116 is low, largely because the chip has only one rasterizer, although it has four tessellation units. So, in some tasks the rasterization speed may be slower than competing GPUs. We consider the second potential drawback to be the relatively low mathematical and texture performance when comparing the new GPU with AMD solutions. The GTX 550 Ti has exactly the same 192 ALUs and 32 TMUs as the GTS 450, and their deficiency can affect many games.

The results of the Geforce GTX 550 Ti in synthetic tests should be confirmed in the next part of our material, dedicated to testing in gaming applications. It will be interesting to see how successfully Nvidia’s new solution will compete not even with the Radeon HD 5770, whose price was recently reduced, but with the HD 6850, a direct competitor in price. This card from AMD is based on a clearly higher-end GPU, and it can easily outperform the GTX 550 Ti in most games.

Indeed, in real gaming applications, rendering speed can depend on several characteristics at once. And quite often it depends on mathematical performance and texturing, which compares favorably with the solution from AMD, which competes with the GTX 550 Ti in price - the Radeon HD 6850.

It seems to us that the recommended price for the new Nvidia model is a bit too high. Perhaps it was initially assumed that its competitor would be the Radeon HD 5770, but now the new product will have to fight with the cheaper Radeon HD 6850, which will be much more difficult, and even factory overclocked versions of the GTX 550 Ti are unlikely to cope with it.