Gtx 560 ti power consumption. Game tests: Call of Duty: Black Ops. Game Tests: StarCraft II: Wings of Liberty

Probably nothing affects computer performance in games more than a video card. Indeed, the more powerful this component, the more enjoyable the game. Currently available on the market great amount video processors from different manufacturers. The choice is extremely wide: from low-power entry-level models to top-end “monsters” for crazy money. And, as always, among all this diversity, NVIDIA stands out. It has long been famous for its high-quality and productive video cards.

Relatively recently, the company released a new video card from the mid-price segment - NVIDIA GeForce GTX 560 Ti. Its characteristics are such that it would be hard to call it average. However, with the current pace of technology development, it will soon become so. What is its peculiarity? At least her main feature the fact that they don’t ask for fabulous money for it. The high cost of products from this brand is common. All the more remarkable is the release of a product for an adequate amount with good characteristics.

general information

The GTX 560 Ti has replaced the successful, but hopelessly outdated GTX 460 model. Everything in the new video card is new: the technical process, the number of cores, the core frequency, memory, and the number of processors. It should be noted that only the top versions of the company’s video cards are marked with the Ti index. The 560 Ti clearly shows how much better this video card is than its predecessor. Significant changes are also observed in the cooling system. Now you don’t have to worry about your video card overheating under high loads. The appearance of the gadget has also been slightly changed. And for the better.

It must be said that NVIDIA always had a terrible confusion with video card indexes. What letters did they not use? As a result, the company achieved that it completely confused the user. Therefore, you should not rely on the Ti index. It is better to study the documentation for the GTX 560 Ti separately, the characteristics of which will not necessarily be top-end. But this does not mean that the video card is no good. Moreover, judging by the reviews, it is capable of “pulling out” even the most heavy games. How? Now we'll figure it out.

Design and appearance

Gone are the days when video cards were soulless pieces of board. Now each of them is a work of art. The appearance of the video card in question is also not satisfactory. Cool looks are the hallmark of the GTX 560 Ti, but the performance hasn't suffered at all.

The insides of the video card are covered with a black casing with engraving. And in the middle of all this splendor, a powerful cooler of the cooling system is conveniently located. At the rear end there are power connectors, and at the front there are DV connectors for two monitors and one HDMI connector. Nothing extra. This asceticism in the number of connectors is explained very simply. Firstly, this is not a top model. Therefore, it is not practical to provide it with a full set of connectors. Secondly, as you know, old schemes are the most reliable. Therefore, the absence of a full set of connectors is a plus in this case.

GPU

The GTX 560 Ti uses a GPU abbreviated as GF 114. This is a completely new graphics processor that was developed specifically for the GTX 560 Ti. Its characteristics are such that it is head and shoulders above all previous models. The new product is characterized by an increased number of cores and much lower heat generation. That is, the graphics processor not only works faster, but also heats up much less than previous models. This means that you can get good power from the video card. Especially if you consider powerful system cooling GTX 560 Ti.

Interestingly, such GPUs are also used in top-end video cards from this company. This alone indicates a “semi-royal” status of this device. Perhaps the reason for the high performance lies also in this.

Video memory capacity

Video memory capacity - most important characteristic video cards. According to some, it is the volume that determines the degree of “coolness” of a video card. But this is far from true. So how much memory is available in the GTX 560 Ti? The characteristics will not greatly affect its performance, but still. in this card is only 1024 MB. Not enough, of course. Where is she going? flagship models. However, not by volume alone... The main advantage lies in the frequency of the core and memory.

Core and memory frequency

The higher the frequency, the more powerful the video card. The faster it processes information. The GTX 560 Ti is fine with this. The core frequency of this video card is 822 MHz. And the memory frequency is 4000 MHz. This means that you can achieve high performance in the most power-hungry games. Even though the memory capacity is only gigabytes. As already mentioned, memory is not the main thing here. By the way, the ASUS GTX 560 Ti has exactly the same characteristics. Which is quite natural. GPU is the same. The difference is only in the details.

Thanks to the high frequency of the core and memory bus, the performance of the video card is much better than it might seem at first glance. Of course, the number of universal processors themselves is of considerable importance here. There are 384 of them. This is more than enough to increase productivity. Here is the answer to how a video card can cope with “heavy” games without having top-end characteristics.

The series of failures that haunted NVIDIA throughout almost the entire 2010, apparently, has finally been interrupted. A deep redesign of the GF100 graphics core allowed NVIDIA to launch a couple of fairly powerful accelerators on the market, thanks to which the chipmaker was able to regain the title of absolute champion in the class of senior-level single-chip solutions. The public, interested in the situation on the video card market, expected a lightning-fast and crushing response from the “red giant” - AMD, however, there was no full answer. Architecture changes, the introduction of new technologies and an increase in clock frequencies did not help AMD prepare a product capable of defeating the GeForce GTX 580. The Radeon HD 6970 accelerator, which is the senior representative in the line of single-processor video cards from AMD, is able to withstand only the GeForce GTX 570. However, a strong position in the upper segment market in no way means for NVIDIA full victory. We remember that the majority of buyers choose the golden mean or even budget solutions, so NVIDIA has something to work on, just like AMD. And such work is certainly underway. For example, towards the end of last year, AMD introduced its Radeon HD 6850/6870 series accelerators, which, judging by the results of our tests, demonstrated very good performance indicators in their class. Well, today it’s time for NVIDIA to make a response move, so in the arena is the GeForce GTX 560 Ti accelerator.

One can’t help but remember the times of GeForce 2/3 video cards, because it was then that NVIDIA first began using the Ti postfix, meaning Titanium, in the names of its products. The old-new term not only adds euphony to the name, it should be associated by a potential buyer with the reliability and high performance of the product. So, what has NVIDIA prepared this time? Let's look at the features of the new product in more detail.

The heart of the new GeForce GTX 560 Ti accelerator is the GF114 video processor, which, in fact, is a significantly simplified modification of the older GPU of the Fermi family. Moreover, unlike the GeForce GTX 460, which is being replaced by the GTX 560 Ti, the GF114 chip is a derivative of the GF110, and not the GF100, like the GTX 460. To compare the configurations of functional blocks of modern NVIDIA GPUs, as well as to familiarize yourself with the characteristics of the reference versions of video cards Based on them, we present to your attention a comparative table:

GF104 GPU architecture

So, the GeForce GTX 560 Ti GPU contains two GPC clusters (Graphics Processing Clusters), which, in turn, consist of 8 streaming multiprocessors (Streaming Multiprocessors or SM), 4 for each GPC. The SM configuration is similar to that used in the GeForce GTX 460. In total, the GF114 core contains 384 CUDA cores (48 per SM), four dispatch units, 8 texture units and one special functionality unit (SF Unit) for each streaming multiprocessor. The GeForce GTX 560 Ti memory bus width is 256 bits. Like its GF104-based predecessor, the new GTX 560 Ti supports Dolby True HD and DTS-HD Master Audio output via HDMI. In fact, the new GF114 is nothing more than a noticeably accelerated GF104 with another activated streaming multiprocessor. Operation at such a high frequency was made possible thanks to a serious “transistor” upgrade of the graphics core, similar to what NVIDIA engineers did in the case of the GF110. Another difference between the new accelerators is the higher memory frequency compared to the GTX 460, which naturally has a positive effect on memory bandwidth.

The relative simplicity of the GF114 chip, as well as the modifications that NVIDIA made to increase the frequency potential, will certainly appeal not only to enthusiasts, but also to those who want an energy-efficient product.

Judging by NVIDIA's presentation, the new accelerator is not only faster than the GeForce GTX 460, but also has a better performance per watt ratio, including in comparison with some competitor solutions. Of course, we still have to check this, but one thing is clear now: the appearance of overclocked products with a modified GPU power supply and memory system is not far off, fortunately, the characteristics of the new video cards are conducive to this.

The list of supported technologies has not changed compared to other modern NVIDIA solutions. The new product supports NVIDIA proprietary technologies: CUDA, PhysX, Surround, 3DVision and others.

Besides technical information, cost is an important factor. According to NVIDIA, the recommended retail price for GeForce GTX 560 Ti accelerators is 9,999 rubles. If we assume that domestic sellers will keep the cost of video cards in this series within the specified limits, then the following should be considered as competitors for the GTX 560 Ti:

• GeForce GTX 470 (retail cost about 8,000 rubles)
• Radeon HD 6870 (retail cost about 8,000 rubles)
• Radeon HD 5870 (retail cost about 9,500-10,000 rubles)
• Radeon HD 6950 (retail cost about 10,500 rubles)

Testing will show what the balance of power will be, and we move on to getting acquainted with the reference sample GeForce GTX 560 Ti.

The design of the cooling system (CO) of the GeForce GTX 560 Ti can be called typical for the new generation of NVIDIA graphics accelerators. Black gloss of the CO casing adds appearance solidity accelerator. The length of the GeForce GTX 560 Ti reference sample is just under 23 cm, which means there should be no problems with placing the accelerator in compact cases.

To power the GeForce GTX 560 Ti, two six-pin PCI-Express power connectors are required.

On the end panel, in addition to the holes through which hot air exits the body, there are two Dual Link DVI-I connectors, as well as one Mini-HDMI port.

Dismantling the cooling system is not difficult. The cooler is in contact with the elements of the power system and video memory through special thermal pads, and the heat GPU CO is transferred to the copper base through a thin layer of thermal paste.

Let's disassemble the cooler. Under the casing lies a cooling system consisting of a metal plate, onto which heat is removed from the video memory and power system elements, and a structure responsible for cooling the GPU, consisting of a copper base soldered to several radiators. Three heat pipes with a diameter of 6 mm transfer heat from the base, and they evenly distribute it over the surface of the radiators located on the sides of the central one.

In some cases, when upgrading a video system, users leave separately purchased coolers in order to later install them on a new video card. If you are going to do this, then you will probably find information about the distance between the mounting holes around the GPU useful. For convenience, we have numbered the holes around the GF114 GPU.

• The distance between the centers of holes 1 and 3 is 6.0 cm
• The distance between the centers of holes 3 and 6 is 5.0 cm
• The distance between the centers of holes 6 and 1 is 7.9 cm

The power supply system of the GeForce GTX 560 Ti accelerator has the following circuit:

Four phases are dedicated to the GPU and one phase is allocated to the memory, as on the GeForce GTX 460. The controllers are marked on the PCB as U501 for the GPU and U5 for the memory. Most likely this is ON Semiconductor NCP5388 (or NCP5395) for the GPU and Anpec Electronics APW7165 for memory.

Memory chips are manufactured by Samsung. The chip marking is K4G10325FE-HC04, the nominal frequency is 5 GHz (effective frequency). Despite such a high nominal frequency, the GeForce GTX 560 Ti memory operates at a lower frequency - 4 GHz.

This concludes our study of the design features of the GeForce GTX 560 Ti accelerator and moves on to practical tests. Before moving on to the test results, please review the list of test packages as well as the test bench configuration.

Nvidia Geforce GTX 560 Ti:

description of video cards and results of synthetic tests

It makes sense to recall once again that 460/560 class cards require additional food, with two 6-pin connectors.

About the cooling system.

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

Nvidia Geforce GTX 560 Ti 1024 MB 256-bit GDDR5, PCI-E

Nvidia Geforce GTX 560 Ti o/c 922/1844/4400 MHz 1024 MB 256-bit GDDR5, PCI-E

We think that there is no point in explaining why there are two monitoring schedules. Yes, we overclocked the card's operating frequencies from 822/1644 MHz core to 922/1844 MHz. At the same time, the card works stably, there are no problems. And the maximum core heating in both cases is clearly not great for this kind of card.

By the way, in our diagrams with test results we will show the performance of the card not only at nominal, but also at precisely these increased frequencies.

Equipment. Considering that reference samples never have a complete set, we will omit this question.

Installation and drivers

Test bench configuration:

• Computer based on CPU 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 900 W.
• operating room Windows system 7 64-bit; DirectX 11;
• Dell 3007WFP monitor (30″);
• ATI drivers version Catalyst 10.12; Nvidia version 266.56 / 266.35.

VSync is disabled.

Synthetic tests

The synthetic test packages we use can be downloaded here:

• 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 560 Ti GTX 560)
• GeForce GTX 460 with standard parameters, model with 1 GB of video memory (hereinafter GTX 460)
• GeForce GTX 570 with standard parameters (further GTX 570)
• Radeon HD 6950 with standard parameters (further HD 6950)
• Radeon HD 6870 with standard parameters (further HD 6870)

To compare the results of the new Geforce GTX 560 Ti model, we chose these particular video cards for the following reasons: Radeon HD 6950 and Radeon HD 6870 are the closest solutions from the competitor in price, Geforce GTX 460 is a video card on a similar previous generation GPU, and GTX 570 This is the closest solution to the current generation in terms of characteristics, based on the more powerful GF110 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 this test, all video cards traditionally show numbers that are far from theoretically possible values ​​(we will recheck them later, in the 3DMark Vantage test). The results of this synthetic for the GTX 560 Ti fall far short of the peak values; it turns out that the new video card selects up to 34 texels per clock cycle from 32-bit textures during bilinear filtering in this test, which is significantly lower than the theoretical figure of 64 filtered texels.

This happens, most likely, due to the limitation of performance by video memory bandwidth, since the same GTX 570 was ahead, despite the fact that, according to theoretical figures, the older solution should be inferior to the one announced today. However, the GTX 460 is still left behind, although not too noticeably.

But both AMD video cards are head and shoulders above Nvidia's new solution in modes with a large number of textures superimposed on the pixel. And in cases with a small amount textures, the bandwidth limitation affects even more, and up to three textures all video cards show similar results. This test clearly does not achieve the realistically possible performance of the new GPU, but let’s see them in 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 clearly shows that the rendering speed of many solutions in simple conditions is seriously limited by memory bandwidth.

The maximum result remains with AMD solutions, which have a significantly larger number of TMUs and are more efficient in achieving high efficiency in our synthetic test. The HD 6950 shows the maximum result, almost double the numbers of the GTX 560 Ti. It is interesting that even in cases with 0-4 overlaid textures, the solution we are considering today is inferior to the others, except for the GTX 460, although its memory bandwidth is almost the same as that of the HD 6870.

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.

Tests of pixel shaders of lower versions are very, very simple for modern even mid-level GPUs and cannot show all the capabilities of modern video chips. In these tests, performance is limited mostly by the speed of texture modules, taking into account the efficiency of blocks and caching of texture data in real tasks, and there is also the influence of video memory bandwidth.

It can be seen that the GF114 completely repeats the results of the GF104, only taking into account a larger number of ALUs and TMUs and their operation at higher frequencies in the case of the GTX 560 Ti. In the simplest shaders, the difference between the GTX 560 Ti and the GTX 460 was 23-28%, which is lower than the theoretical ALU and TMU power gain figures. It seems that the GTX 560 Ti's performance in this test is limited by video memory bandwidth and fill rate, since the difference between the solutions is significantly lower in these indicators.

What’s more interesting is that in three simple tests, the GTX 560 Ti was able to compete even with the GTX 570. However, in the most difficult tests, the GF114 solution still lagged behind the top-end GF110. As for comparison with AMD video cards, both of them outperformed the GTX 560 Ti; it was only able to compete with the HD 6870, and even then only in the simplest tests. Let's look at the results of more complex intermediate pixel programs:

But these tests turned out to be much more interesting. We are interested in the difference in the results of the GTX 560 Ti (and the GTX 460) and the GTX 570 in these two tests. 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. And in this test, the GTX 560 Ti shows a theoretically justified result, outperforming even the GTX 570. It was not possible to get the best among AMD video cards, but the HD 6870 showed a similar result, which is quite consistent with the theory (the peak texturing speed of these solutions is close).

The results of the second test are quite different, in which the GTX 560 Ti already loses to everyone except its younger sister GTX 460. This test is more computationally intensive, and the influence of mathematical performance is felt in it. Therefore, the test is better suited for AMD video cards with a large number of ALU units. The difference between the GTX 560 Ti and the GTX 460 in these two tests was 32-37%, which is approximately in line with theoretical figures.

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 simpler shaders version 2.0:

• 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 the cards in the Frozen Glass test is similar to what we saw above in Cook-Torrance, and the new GTX 560 Ti is again noticeably inferior to the GTX 570, which has the top-end GF110 GPU. Both AMD solutions were also far ahead.

In the second “Parallax Mapping” test, the results are also somewhat similar to the previous ones, but this time both Radeons are not so far ahead of Nvidia video cards. Solutions based on the GF114 and GF104 chips are again left behind everyone, and the GTX 560 Ti is ahead of the GTX 460 in these tests by 24-28%, which again indicates that the potential of the new GPU has not been fully realized, caused by a not too large increase in the video memory frequency and fill rate compared with GTX 460 1 GB.

Let's consider these same tests in a modification with a preference for samples from textures over mathematical calculations, where the new solution should show a stronger result:

As in the case of the GTX 460, the position of the new solution relative to the GTX 570 from the top series has improved somewhat. True, Nvidia video cards have become even more inferior to the HD 6870 and HD 6950, which have many texture modules. But now the new GTX 560 Ti even beats the GTX 570 in the Frozen Glass test, which is more dependent on TMU performance. And in the second test, the results of the GTX 560 and GTX 570 were close. The difference between video cards based on GF104 and GF114 was 30-32%, which is already closer to 38% of the theoretical difference in texturing speed.

These were all legacy 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.

It looks like the new Nvidia solution is doing great when testing pixel shaders version 3.0. Both PS 3.0 tests are quite complex, they are almost independent of memory bandwidth and texturing and are purely mathematical, but with a large number of transitions and branches, which the new Nvidia architecture copes with very well.

In the most complex Direct3D 9 tests, the GTX 560 Ti presented today shows results noticeably higher than the HD 6870, and in one of the tests it is ahead of the HD 6950. When compared with the GTX 570, the video cards are close in the Fur test, but in the advanced parallax mapping test there is a new solution Nvidia is inferior to its older brother GTX 570, and quite significantly. It seems that the test results are greatly affected by both the lack of bandwidth and the lower efficiency of GF114/GF104 compared to GF110/GF100 in this test (affected by the increased number of ALU units in each multiprocessor). Although the result for the GTX 560 Ti is still excellent - it is close to the more expensive HD 6950 from its competitor.

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, which is clearly seen in the close figures of HD 6870 and HD 6950. The results in “High” are approximately one and a half times lower than in “Low”, as and according to theory it should be. In Direct3D 10 tests of procedural fur rendering with a large number of texture samples, Nvidia's solutions have always been stronger, but the latest AMD architecture also shows good results.

How good they are that the new Nvidia video card is even slightly behind both Radeons, although not very much. And the leader of the test is the GTX 570. This again indicates some influence of fill rate, and possibly also bandwidth. Although in the case of the GTX 560 Ti and GTX 460, the difference in speed exactly corresponds to the theoretical difference in the speed of ALU and TMU - about 38%.

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:

Enabling supersampling theoretically increases the load by four times, and in this case absolutely all Nvidia solutions lose ground, and both AMD video cards look even stronger in such conditions. Now both Radeons outperform even the GTX 570. But what's strange is that the HD 6870 is ahead of the HD 6950. In other words, in the case of AMD video cards, performance is limited by ROP performance, which is higher in the HD 6870. The GTX 560 Ti still lags behind the GTX 570, but outperforms the GTX 460 by the same 39-40%, corresponding to theoretical figures.

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 difficult test mode with supersampling and self-shadowing selects from 80 to 400 texture values, that is, eight times more than in simple mode. Let's first check simple options without supersampling:

This test is more interesting from a practical point of view, since varieties 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 Crysis and Lost Planet. 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 without SSAA, and the results are close even in absolute numbers. In the updated D3D10 version of the test without supersampling, the new GTX 560 Ti model copes with this task 36-37% faster than its sister GTX 460 on the GF104 chip. But they still lag behind both AMD video cards, although the leader is the GTX 570, based on the GF110. Although the top card has a clear advantage in terms of theoretical characteristics, we did not expect such a large gap.

Let's see what difference enabling supersampling makes, it should again cause a larger drop in speed on Nvidia cards.

When supersampling and self-shadowing are enabled, the task becomes noticeably more difficult; enabling both 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 clearly improved their performance compared to the Nvidia solution.

And now the HD 6950, with low detail, slightly outperforms the GTX 570, just as slightly behind it by more difficult conditions. And the HD 6870 isn't too far behind. The same cannot be said about the GTX 560 Ti and GTX 460. Nvidia’s new solution loses to both competitors and is only ahead of the younger GTX 460. Moreover, by exactly the same theoretically justified 38-40%.

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 amount 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 GF114 GPU is architecturally no different from its predecessor GF104, the difference between the GTX 560 Ti and the GTX 460 corresponds to a theoretical 38% in comparative performance ALU blocks. And all other solutions were located approximately in accordance with the theoretical indicators.

AMD video cards are clearly faster in this synthetic test, since in computationally complex tasks the modern AMD architecture has big advantage ahead of competing Nvidia graphics cards. This time the gap between Nvidia and AMD cards remains huge, the HD 6870 and HD 6950 show the same result, even ahead of the top-of-the-line GTX 570. Well, the difference with the GTX 560 Ti turned out to be one and a half times, which is also close to theory, taking into account the lower efficiency of AMD video chips.

In our past studies, we noted that this test is not entirely dependent on ALU speed, and the best performing solutions are limited by video memory speed. So let's look at the second shader calculation test, 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 very few changes, only the HD 6950 is even further ahead, as it should be in theory. In the second test, the rendering speed is already limited solely by the performance of the shader units, and the difference between the GTX 560 Ti and the GTX 460 became even a little more theoretical - 40%, and the new model almost caught up even with the GTX 570. But this is still too little for the new mid-range video card level has caught up with competitors represented by the Radeon HD 6870 and even more so the HD 6950 in mathematical tests.

The result of mathematical calculations has been unchanged for several years - AMD solutions have a clear advantage, explained by the large number of ALU units, the speed of which is not greatly affected by even the relatively low efficiency. Let's move on to the results of testing geometry shaders, they will be interesting because the main performance limiter in them is the speed of geometry processing, and it will be interesting to compare the GTX 560 Ti with the HD 6870 and HD 6950.

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, performance corresponds to the number of points, with each step the FPS drops by about two times. Task for modern video cards not particularly complex, performance in general is limited not only by geometry processing speed, but also by memory bandwidth.

The difference between the GeForce GTX 560 Ti and the GTX 460 turned out to be even higher than theoretical - 46%. The new video card showed results approximately at the level of both competitors from AMD, and the leader was the Nvidia video card, based on the top-end GF110 graphics processor. The Geforce GTX 570 in all modes noticeably outperformed all other video cards, including the GTX 560 Ti. This is due to the fact that the GTX 570 has more geometry processing units.

This time, AMD video cards showed good results - the optimization of geometric blocks made by AMD engineers clearly had an impact. Their execution speed for geometry shaders turned out to be close to the performance of the GF114, which is not bad. Let's see if the situation changes when we transfer part of the calculations to the geometry shader:

When the load in this test changed, the numbers for Nvidia solutions remained almost unchanged, and the older Radeon video card slightly improved the results, and now they are both just a little faster than the GTX 560 Ti. 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. Let's see what changes in 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 creation geometry using rendering in two buffers, 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 the theoretical parameters for which each next level"Polygon count" should be less than twice as slow.

It is in this test, with a balanced load, that the rendering speed for all solutions is less clearly limited by geometric performance. The new GeForce GTX 560 Ti this time lags less behind the GTX 570, and as the complexity of the geometry increases, the lag gets smaller. And compared to Radeon cards, the new GPU shows very close results. The difference with the GTX 460 this time is only about 30%, in contrast to 46% in the two previous diagrams, which clearly indicates an emphasis on memory bandwidth.

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.

In this test, the difference between GF114 and GF104 again returned to the theoretically justified 40%. And the GF110 is far ahead of all others in terms of execution speed of geometry shaders - this is clearly due to the presence of four rasterizers, in contrast to two in the GF114 and GF104. It is clearly seen that the GTX 570's geometry processing capabilities and geometry shader execution speed are almost twice as high as those of the GTX 560 Ti.

But the most important thing here is the comparison with AMD video cards. Nvidia's new solution is still faster than the Radeon HD 6870 and HD 6950 in this test, but only slightly. The number of rasterization units in the GF114 is not as large as in the GF110, so the same GTX 570, having a large number of rasterizers, shows results that are 70-75% higher.

So, compared to the GF110, rasterization speed could be a potentially weak indicator for overall performance. Although for a mid-level solution it is important that its results are slightly higher than those of a top-end competitor. In tessellation tests, the speed is no longer limited by rasterizers, but by tessellators, and in such cases the new GPU should show even stronger results than its competitors.

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 both texturing speed and memory bandwidth affect the results of this test. The difference between all solutions is not very big, only the GTX 460 shows slightly strange results, being the slowest. The newly announced GTX 560 Ti is 30-40% ahead of the old solution in simple modes, and only 10% in complex modes. Probably, different balances were made for these video cards in different driver versions.

The GTX 560 Ti slightly outperforms both AMD competitors in medium and hard modes, but struggles with something (bandwidth again?) in easy modes. It seems that Nvidia cards are somewhat easier to do these tasks. 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. Now all video cards based on Nvidia chips run into something unknown in the lightest mode, and the HD 6950 becomes the leader in it. But in the heavy mode, the GTX 560 Ti has almost caught up with the GTX 570 and is noticeably ahead of its competitors HD 6870 and HD 6950. The difference between the GTX 560 Ti and GTX 460 ranged from 10% (in heavy mode) to 50% (in light mode)! There is clearly something wrong with the driver for the old model...

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.

The results in the Waves test are not similar to what we saw in previous charts. In it in different conditions We already see a slight advantage for AMD products, although in general all solutions, except for the GTX 460, perform evenly. Our today's hero GTX 560 Ti in this test shows performance only slightly better than the HD 6870 and HD 6950, and even lags behind the GTX 570 in easy mode. But the new card is faster than the GTX 460 by up to 37%. Let's consider the second version of the same test:

Changes with increasing complexity of conditions are again very few, almost non-existent. The relative performance of the GF114 GPU in the second vertex fetch test at high detail improved slightly, and the new GTX 560 Ti now leads both Radeons in the heavy mode, while continuing to lag slightly in easy conditions. The difference between video cards based on GF114 and GF104 was 36-39%, which corresponds to the theoretical difference in texturing speed.

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 the new Nvidia solution in this package, we will be able to draw some new and useful conclusions that eluded us in the RightMark family of tests. This is especially true for the TMU speed test, because our analogue shows strange results.

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 texture performance test from the Vantage package, the results are completely different than in our RightMark. These figures are more similar to the true state of affairs, and closer to the theory. In 3DMark texture synthetics, Nvidia cards use the available texture units more efficiently, and the GTX 560 Ti shows results at the level of the Radeon HD 6870, which is close to the theoretical difference. Naturally, the HD 6950 remains far ahead, as it has a large number of TMUs.

As for the comparison with Nvidia video cards, here too we see the correct result - the GTX 560 Ti outperforms the GTX 570, in full accordance with the theory. And the difference between the GTX 560 Ti and the GTX 460 was 39%, which is also equal to the theoretical difference in texture fetch performance. In general, the new video card based on the GF114 chip shows very good results, and among all the presented Nvidia video cards, it is the one that becomes the leader in texturing.

Fill rate test. 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 performance indicators in this test correspond to the theoretical fill rate figures (performance of ROP blocks), without taking into account the influence of video memory bandwidth. They are not similar to ours because we use an integer buffer with 8-bits per component, and in Vantage test- 16-bit floating point. And the 3DMark Vantage numbers show the performance of the ROP units, and not the amount of memory bandwidth.

The test results roughly correspond to theoretical figures, and most of all depend on the number of ROP blocks and their frequency. There is an influence of PSP, but it is small. The new GTX 560 Ti model shows good results, almost catching up with its younger competitor from AMD, which has approximately the same theoretical fill rate. But the new Nvidia board is not able to catch up with the HD 6950. The new mid-level solution lags behind the GTX 570 for the same reason - even the inferior GF110, in theory, has higher performance of ROP units. But compared to the GTX 460, the new model turned out to be noticeably faster.

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 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 successful balance of GPU and video memory memory blocks is important. Significantly affects the speed and efficiency of branching in shaders.

Unfortunately, the GF104 and GF114 do not show very good results in this test; the GTX 460 ended up being the slowest card and was more than two times behind the fastest HD 6950! Well, the video card presented today, intended for the middle price range, does not reach the youngest of those presented AMD boards and its older sister GTX 570. However, little consolation can be the fact that it is ahead of the previous generation model by as much as 41%.

We have already written about the fact that it is difficult to say which parameters most influence the results of this test. The culprit is probably the reduced efficiency of executing shader programs with branches in the GF104 and GF114, compared to the GF110 and GF100. In past studies, the GF104 has proven itself in physics simulation tests, and we hope that the GF114 will not fail there either.

Feature Test 4: GPU Cloth

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.

It seems that the rendering speed in this test is also affected by several different parameters. Most likely, overall speed depends on geometry processing performance and geometry shader execution efficiency. Even the GTX 460 performs well in this test, only slightly behind the HD 6950 - fastest card AMD in our review. It is clearly seen that in this test all Nvidia cards show much better results when executing complex shaders.

In this test, the GTX 560 Ti clearly has an advantage over both competing solutions in the form of the Radeon HD 6870 and HD 6950. With the execution of geometry shaders, the speed of geometry processing and the efficiency of executing complex programs, the GF114 is clearly all right, like all other chips from the company. The new model lags behind the top-end GTX 570, which is quite consistent with the theoretical characteristics.

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.

The results of this test are very similar to those we saw in the last chart, but the GTX 460 now scores even better than both Radeon cards. Not to mention other Nvidia video cards. The difference between the GTX 560 Ti and GTX 460 this time was slightly more than 30%.

In synthetic tests simulating tissues and particles of this test package, which use geometry shaders, the new GPU showed excellent results, significantly ahead of competing AMD GPUs. And the junior top solution based on the GF110 chip simply has a noticeably larger number of geometry processing units, which is why it has become the comparison leader in these tasks.

Feature Test 6: Perlin Noise

The last feature test of the Vantage package is a mathematically intensive test of the video chip; it calculates several octaves of the Perlin noise algorithm in the pixel shader. Each color channel uses its own noise function to increase the load 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. The speed of all solutions shown in it approximately corresponds to what should be obtained according to theory, and is very close to the picture that we saw earlier in our mathematical tests from the RightMark 2.0 package (at least in the second one).

Of course, AMD video cards outperform competitors from Nvidia this time too. Simple but intensive mathematics performed on Radeon video cards much faster, as we have already seen more than once. Although in other computing tests with more complex programs, such as physics calculations, Nvidia solutions look quite good, including the GTX 560 Ti.

In the same math test The Geforce GTX 560 Ti, based on the new GF114 chip, shows speeds slightly lower than the GTX 570, as it should be in theory (although the difference should be slightly less than it turned out). But the new model is 44% faster than the GTX 460, which is even more than the theoretical difference. But there is a large lag behind both Radeon video cards, and the leader in comparison is the HD 6950 model, as in other extreme mathematical tests.

Conclusions on synthetic tests

Based on the results of our synthetic tests of the new Nvidia Geforce GTX 560 Ti model, based on the GF114 graphics processor, as well as the results of other video card models from both video chip manufacturers, we can conclude that Nvidia has turned out to be an excellent replacement for the GTX 470. In many synthetic tests The new Nvidia chip performed very well, sometimes catching up with the Radeon HD 6950 and approaching the level of the GTX 570 in some cases.

The new GPU differs from the GF104 in the increased number of execution units and increased clock speed, which led to significantly increased performance (about 30-40% in most cases), and the video card based on it presented today looks very attractive. We especially note the texture fetch speed - in this parameter, the GTX 560 Ti is significantly ahead of even the GTX 570, based on the GF110! We also note that the GF114 has increased the number of active tessellators, which makes it possible to further increase the performance of geometric processing.

Among the shortcomings, we highlight that the architectural changes in the GF114 and GF104 led to a slight decrease in the execution efficiency of some shader programs. A second potential disadvantage could be the relatively low memory bandwidth compared to higher-end solutions. It is insufficient memory bandwidth that often limits the performance of the GTX 560 Ti, and this can have an even stronger effect in the case of overclocked card versions, since this GPU is capable of operating at significantly higher frequencies, but the GDDR5 memory used is not.

It can be assumed that the generally quite good results of the GeForce GTX 560 Ti in synthetic tests will be confirmed by positive results in the next part of our material, dedicated to testing in gaming applications. The new Nvidia solution should show very good results at the level of previous solutions like the Geforce GTX 470, and will be slightly slower than the Geforce GTX 570, its older brother based on the GF110 chip, which is quite logical.

But it’s difficult to say what will happen in games compared to competitors for several reasons. Competing in price, the Radeon HD 6870 and HD 6950 from AMD are too different even from each other, having different strengths and weaknesses. More often than not, the GTX 560 Ti should be ahead of the HD 6870, but it should still be slower than the HD 6950. Although in some tests it will be faster than both competitors, in others it may be inferior to them.

In games the situation is always more complicated than in synthetics; the rendering speed there often depends on several characteristics at once. And very often it depends on fillrate and texturing, which is what AMD solutions are strong at. In addition, you need to take into account the slightly inflated price (for Russian market) for the new GTX 560 Ti. It looks like it will have to fight with the cheaper Radeon HD 6950 1 GB and this comparison may turn out to be less rosy for the new Nvidia video card.

The GeForce GTX 560 Ti video card, which appeared on the market not so long ago, is in stable demand. Although it did not provide a revolutionary leap in performance in its class, it did offer a good combination of price and performance parameters, “pulling out” the vast majority of modern games at maximum settings.

At one time, while researching this video card, I noticed the original Ti suffix in the name. It indicates that the video card belongs to the high-performance Titanium series, but has not been used to designate NVIDIA products since the GeForce 4000 series (2003)! I noted its use simply as a kind of “hello from the past,” a kind of “return to the roots.” In reality, everything turned out to be more prosaic - in the current generation, the entire NVIDIA designation system has changed slightly.

This is not officially regulated in any way, but at one time the GTX prefix was considered to belong to the most powerful video cards: the top model and its “light” version. This time it has been significantly “democratized” - now even the modest GeForce 550 (the fifth oldest video card in the manufacturer’s lineup at the time of release) is proudly called GTX. Let's leave these marketing "pranks" to the conscience of NVIDIA specialists - it is more pleasant for the buyer to purchase an accelerator of the older series, even with a modest digital index.

The second change is the use of the same suffix Ti. The company decided to take this step in order not to create a bunch of digital designations for video cards of similar performance (remember how price lists once grew with the advent of all sorts of GTX 275 or GTX 465, produced to “plug holes” in the model line?). Calculations were also made on the psychology of the buyer - first a full-fledged video card with the designation Ti is launched on the market (so far there are two of them - GTX 560 Ti and GTX 550 Ti), and a little later a stripped-down version without the corresponding suffix.

I think many will be tempted by seeing “almost the same thing, but cheaper” on sale, and some buyers will not even pay attention to the slight difference in the name.

In general, it’s all marketing all around. Meanwhile, the first of the video cards “without titanium impurities” - GeForce GTX 560 - has already been in the laboratory; Below is a detailed research report on this product. I wonder whether this accelerator will be able to compete with the older model, offering a favorable price/performance ratio, or, on the contrary, it will turn out to be a weak “off-cut” that got its big name through a misunderstanding. Let's figure it out, starting with studying the architecture of the new product, because it is from this data that it is easiest to understand how different the two “five hundred and sixties” are from each other.

Architecture, cost and market position

This time I will not describe the features of the architecture in detail so as not to bore readers, because the GF104 and GF114 graphics processors and video cards based on them (GTX 460, GTX 560 Ti) have already been studied inside and out. The new product has a lot in common with both “sixties” GeForce – the current and previous generation. The easiest way is to present the data in the form of a table, which will provide information on all video cards using GF104/114 graphics processors, as well as the main characteristics of the GeForce GTX 550 Ti accelerator, which is one step below the new product in the lineup and will also be used in the tests.

* For most video cards, there are also “non-reference” versions with double the amount of video memory.
** Preliminary data.
*** At the time of release.

For readers tracking the evolution of GeForce video cards, these numbers are enough to understand what's what. Everything is simple here. When releasing a full-fledged version of the GeForce GTX 560 Ti, NVIDIA did not “reinvent the wheel”, but used the design reserves of the GF104 graphics processor (GeForce GTX 460), laid down in the previous generation.

The new GPU GF114 did not offer anything fundamentally new in terms of architecture. 8 streaming multiprocessors (corresponding to 384 single stream processors, or, as the manufacturer calls them, “CUDA cores” and 64 texture units) were already implemented in the design of the previous GF104 processor. True, on the GeForce GTX 460 one of the multiprocessors was deactivated (which naturally led to a decrease in the number of stream processors from 384 to 336, and texture units from 64 to 56), and on the GeForce GTX 560 Ti it was finally “awakened”, completely realizing the potential of the GF 104/114 core.

NVIDIA also claims that the GF114 uses an optimized manufacturing process and minor changes have been made to the core design, resulting in lower heat dissipation and improved overclocking potential. This seems to be true - the processor frequency on the GTX 560 Ti can be increased much more than on the “old” 460s.

So - the key difference: GTX 460 - seven active multiprocessors, GTX 560 Ti - eight active multiprocessors, otherwise these video cards differ only in operating frequencies. I’ll continue this series: the new GeForce GTX 560 has seven active multiprocessors, which means that in architectural terms the accelerator is completely identical to the GeForce GTX 460 1 GB.

How can one not recall NVIDIA’s love for renaming its products. But not everything is so simple: since the GeForce GTX 560 uses the GF114 GPU, and not the old GF104, the video card should retain the good overclocking potential inherent in the older GTX 560 Ti model.

The “new product” (I think the quotation marks are appropriate here) is the already well-known GeForce GTX 460 1 GB, but with a new revision of the graphics processor. Its use made it possible to increase the operating frequency from 675 to 810 MHz (the difference is 20%), which should provide an increase in performance. The video memory on the new video card is also slightly overclocked - up to 1002 (4008) MHz, the same figure for GeForce GTX 460 - 900 (3600) MHz, the difference reaches ~11.3%. If we compare the GeForce GTX 560 Ti and GTX 560, the standard frequencies are extremely close.

Recommended video price GeForce cards The GTX 560 is priced at $199, while the GTX 560 Ti is priced at $249. It is interesting that NVIDIA has also set recommended prices for the new product for Russia, so the GTX 560 is recommended to be sold for 6299 rubles (~$225), and overclocked versions for 6699 rubles (~$235). Let's see how these recommendations are followed.

Video card design

The laboratory received a Palit GeForce GTX 560 Sonic Platinum video card. The original NVIDIA video card (“reference”) is almost completely identical to the GeForce GTX 560 Ti. Considering the close “kinship” of the GTX 560 and GTX 560 Ti accelerators, this is not at all surprising (a similar situation arose when AMD released the Radeon HD 6790 video card, which was presented in the form of the already released Radeon HD 6870 with minor changes). It is also logical that the “without Ti” modification from Palit is also structurally similar to the full-fledged 560 Ti accelerators from the same company.

Thus, in this material, in addition to studying the performance of the GeForce GTX 560 accelerator as such, the design of the original “non-reference” Palit model will be considered. Products from this manufacturer are widely represented in Russian retail, so data on temperature and noise characteristics should be useful for many buyers of the GeForce GTX 560. And adjusted for the slightly higher heat dissipation of the GPU, they allow us to judge the performance of the Palit GeForce GTX 560 accelerator, which is similar in design Ti.

The design of the video card in question is typical for the products of this company: PCB with a bright red mask, a “cunningly” shaped cooling system casing and a fan with an orange impeller.

Another characteristic point is that Palit designers really like to “cut” their video cards in length; several times I have come across their printed circuit boards that are a couple of centimeters shorter than standard ones. So the GeForce GTX 560 turned out to be compact - only ~188 mm in length (measured by the printed circuit board). In general, this is a positive point; a small accelerator can be placed in a very compact case, and in a full-fledged “tower” it should interfere less with air flow, slightly improving ventilation.

The thickness of the video card is standard for accelerators of this level of performance - “two slots”, the height is about 111 mm (this is also a standard).

There are three connectors on the rear panel - HDMI, DVI and classic VGA. A good set that allows you to connect all kinds of monitors to the accelerator (including old ones without a digital interface). Of the common ones, the only thing missing here is the Display Port connector.

At the top of the rear panel there is a “grill” grill for removing heated air from the system unit. Unfortunately, a simple palm test showed that even at maximum fan speeds, very little flow passes through the holes.

This problem is generally typical for cooling systems with a fan, in contrast to “turbines”, which effectively blow heated air out of the case. In addition, the Palit video card has a non-sealed casing with many cutouts.

In the “tail” of the video card, the casing generally rests on miniature “legs”:

In the same photo you can see two six-pin additional power connectors. On such a compact video card they look like an alien element; you subconsciously expect to see only one connector. But don’t forget that this is a fairly powerful mid-class accelerator.

The plastic casing of the cooling system is easy to dismantle along with the fan - just remove four screws.

An eleven-blade fan with an impeller diameter of 75 mm is “tightly attached” to the frame. This is not very good - if something happens it will be difficult to replace it.

The video card with the casing removed looks like this:

Not a bad idea for those with crazy hands: for fun, you can use the video card without a casing at all, by attaching a large 120 mm fan directly to the radiator. It will fit perfectly in width, but will protrude 1.5-2 centimeters beyond the top edge of the PCB.

The radiator is a simple design made of aluminum fins mounted on two 6 mm heat pipes.

The base is an ordinary rectangular copper plate. The heat pipes are in contact with its reverse side; at the point of contact they are flattened to increase the contact area. On this video card, the graphics processor is covered with a heat distribution cover, but the area of ​​the GF114 core is small, so two tubes in the center of the base are enough to cover the most thermally stressed area. The ends of the tubes go to the edges of the radiator; this is a standard design - this way you can more efficiently use the entire area of ​​the fins, using remote areas.

The entire structure is secured with four spring-loaded screws. There are no additional radiators to remove heat from the power switches of the power converter; In addition, these board elements do not fall into the direct fan airflow area. Thermal paste, as always, in abundance.

The printed circuit board of the video card in question is not particularly complex. In the center is the graphics processor, labeled GF114-325-A1. To the right of it is a five-phase power converter controlled by the NCP5395T controller.

Around the GPU are eight memory chips labeled Samsung K4G10325FE-HC04. This designation indicates an access time of 0.4 ns, which corresponds to an effective frequency of 5000 MHz (the actual frequency is 1250 MHz taking into account the GDDR5 QDR).

The reverse side of the PCB is not particularly interesting. I will only note that the GPU power converter controller is located here, so that fans of “hardvolt mods” with additional soldering of resistors can easily get to its legs without dismantling the cooling system.

Based on the results of this section, we can conclude that the design of the reviewed video card corresponds to the GeForce GTX 560 class. There are no special engineering delights here, but a rather complex GPU power converter and a cooling system with heat pipes are used. In general, everything is as it should be.

Test bench

• Motherboard: ASUS P8P67 PRO (BIOS v 1204);
• Processor: Intel Core i5-2500K @ 4500 MHz (base frequency 3300 MHz);
• CPU cooling system: Noctua NH-D14 (2 x Scythe Slip Stream SY1225SL12SH; ~950-1800 rpm);
• RAM: Corsair TR3X6G1600C7 (DDR3-1600, 7-7-7-20, 2x2 GB, dual-channel mode);
• Video cards:
• Palit GeForce GTX 560 Sonic Platinum
• NVIDIA GeForce GTX 560 Ti,
• Leadtek GeForce GTX 550 Ti,
• NVIDIA GeForce GTX 460 768 MB;
• HDD: Western Digital WD1001FALS (1000 GB);
• Power unit: Cooler Master Real Power M1000 (1 kW);
• Housing: open stand.

Software

• Operating system: Windows 7 x64 Ultimate
• Video card drivers: nVidia Display Driver v. 266.66 for GeForce GTX 560 Ti video card, nVidia Display Driver v. 267.59 for GeForce GTX 550 Ti video card, nVidia Display Driver v. 266.58 for video card GeForce GTX 460 768 MB, nVidia Display Driver v. 275.20 Beta for GeForce GTX 560 video card.
• Supporting utilities: MSI Afterburner v. 2.2.0 Beta 2, GPU-z v. 0.5.3, FurMark build 1.8.0, OCCT GPU v. 0.7.

The test bench processor operated at a frequency of 4500 MHz. This level of overclocking for the new 32 nm Intel CPUs is not the maximum; the frequency of 4.5 GHz was chosen as “typical” and achievable for most users of unlocked processors Sandy Bridge.

Testing tools and methodology

To overclock video cards, as well as monitor temperatures and fan speeds, the MSI Afterburner v utility was used. 2.2.0 Beta 2.

To warm up and check the stability of video cards during overclocking, we used the OCCT GPUw (Error Check mode, 1024 x 768) and FurMark (Stability Test, Extreme burning mode, 1920 x 1200, AA0) utilities. The obtained frequencies were additionally checked by running the Heaven Benchmark v 2.1 test with an extreme level of tessellation and graphics tests from the 3DMark 06 and 3DMark Vantage packages.

To check the temperature conditions of video cards in conditions close to everyday conditions, Heaven BenchMark v was used. 2.1 (shader: high, tessellation: normal, AA4x, 1920 x 1200).

Performance in the game Crysis Warhead was studied using the Framebuffer Benchmarking Tool. In the games Lost Planet 2, Resident Evil 5, Mafia 2, Dragon Age 2, Assassin’s Creed: Brotherhood, Crysis 2, testing was carried out using the FRAPS v. utility. 3.2.3. In other cases, built-in performance measurement tools were used.

The noise level was measured using a Becool BC-8922 digital sound level meter with a measurement error of no more than 0.5 dB. Measurements were taken from a distance of 1 m. Level background noise indoors – no more than 27 dB. The air temperature in the room was 22-23 degrees.

Overclocking

The video card used for tests belongs to the Palit Sonic Platinum series. This is an accelerator with “factory” overclocking, and the frequencies of the graphics processor and memory are increased quite significantly. Thus, the GPU was overclocked from standard 810 to 900 MHz (the increase is ~11.1%). The video memory operates at a frequency of 1050 MHz (effective frequency 4200 MHz taking into account the quadrupling of QDR GDDR5), while the NVIDIA specifications indicate a value of 1002 (4008) MHz (an increase of ~5%).

Please note that not all Palit GTX 560 video cards that look like those shown in the illustrations belong to the series with factory overclocking; there are also regular ones that operate at standard frequencies. On the one hand, their purchase is more profitable, since they charge for a Sonic or Sonic Platinum sticker extra money, and readers of the site know how to overclock video cards on their own. On the other hand, it is generally accepted that more successful GPU instances can be selected for such accelerators.

I started by lowering the GPU and video memory frequencies to the standard values ​​​​set by the manufacturer. This kind of “downclocking” was needed to test the video card “in its pure form.”

I note that the MSI Afterburner v 2.2 beta 2 utility was used to adjust the frequencies and fan speeds. The only thing it can’t do is adjust the GPU supply voltage. This is a serious obstacle to successful overclocking; you will have to wait for the next versions that fully support the GeForce GTX 560. We can only rely on the Palit designers, who should have used a higher default voltage value on the factory overclocked video card.

And they did not disappoint; the graphics processor was overclocked from 900 to 960 MHz without raising the voltage. The difference of 60 MHz in relation to the nominal does not seem significant, but if you remember that the standard core frequency of the GeForce GTX 560 is only 810 MHz, then you get a significant increase of 18.5%.

This is quite a suitable option for performance testing. Surely many overclockers will use a video card with a close core frequency; it will be difficult to go further “in the open air”; and even if I had the ability to adjust the voltage, I would most likely be able to raise the frequency by a couple of tens of MHz without a significant deterioration in temperature and noise characteristics.

Video memory chips manufactured by Samsung were overclocked to 4660 MHz (overclocking is ~16.3% relative to the nominal value). This is not the best, but not a failure either. The increase in frequency led to an increase in bandwidth to almost 150 GB/s.

Based on the results of this section, I got the impression that the GeForce GTX 560 overclocks approximately the same as the older GTX 560 Ti model. The difference may be due to the slightly lower heat dissipation of the GF114 processor with disabled blocks (this should play into the hands of the younger model). On the other hand, it can be assumed that less successful GPU samples will be used to produce such video cards, then in the case of overclocking individual GeForce GTX 560 samples, problems may arise.

And yet, it’s difficult to imagine a sample of such a video card that will not reach a frequency of about 950 MHz with an increase in voltage (this time they did without it, and the Palit designers probably showed caution by setting not the most high value). This means that this accelerator unconditionally surpasses the GeForce GTX 460 in operating frequencies, which is usually overclocked to 800-850 MHz, depending on the success of the instance. An extra 100-150 MHz is what can really make a difference in the performance of these video cards that have the same core configuration.

Temperature and noise level

To warm up the video card, we traditionally used the Furmark tests (the well-known “donut” that provides extreme load) and Heaven Benchmark (in this case, milder conditions, closer to everyday ones, are simulated).

First, the video card was tested at standard frequencies of 810/4008 MHz. In general, this test does not make much sense, since a factory overclocked accelerator is used, and the supply voltage of the GPU is unknown. The maximum that can be found out is how much the core temperature will increase during overclocking without increasing the voltage.

Furmark, as always, poses a difficult challenge for the cooling system.

The temperature is not the highest for a donut, but the video card is clearly audible - the noise level is ~37.9 dB.

After overclocking, the temperature increases by only two degrees:

Fan speed increases by 5%. In this mode, being near a running video card is no longer comfortable - the noise level increases to 41.1 dB.

Now the data from the “game” test.

Palit designers have very accurately adjusted the automatic fan speed control. The fact is that 45% of the maximum speed is exactly the limit after which the video card becomes audible. And here the result is very good - about 33.4 dB, the accelerator hisses barely audibly.

After overclocking, the temperature increases slightly, and the fan speed increases by only 2%. Subjectively, this leads to a deterioration in noise characteristics - hissing begins to turn into a mid-frequency hum. The sound level meter, however, recorded a figure of 34.1 dB, which is not much.

In addition, I present the results of the same tests at the “native” frequencies set by Palit specialists (900/4200 MHz).

Heaven Benchmark:

The indicators differ little from those obtained on the maximally overclocked card; not surprising, since overclocking was carried out without raising the voltage, and the 60 MHz increase is not that much.

So, the Palit GeForce GTX 560 video card is not silent - even in the gaming test the accelerator can be “heard by the naked ear,” although it cannot be called particularly loud. On the other hand, Sonic Platinum is a product with serious factory overclocking, and the cooling system copes with its responsibilities adequately.

GeForce GTX 560 Ti | Introduction

Back in July last year, THG editors suggested that Nvidia had turned off one of the streaming processors (Streaming Multiprocessor) on the GeForce GTX 460 with the GF104 GPU to avoid pushing the GeForce GTX 465 graphics card out of the market. Wouldn't it be interesting to see a version of this chip with all eight stream processors included? We meant that the GeForce GTX 460 has already taken its rightful place among video cards with 336 CUDA cores and 56 texture units. And, of course, a full-featured version of this video chip can challenge more expensive (and less attractive) GeForce video cards. This could also be a threat to Radeon HD 6800 series video cards, which were released three months later.

It is logical that nVidia did just that. The GeForce GTX 465 has enough competitors even without the GF104. However, these video cards are no longer produced. It is also worth recalling that nVidia wants to completely stop producing video cards based on the GF100 processor.

The GeForce GTX 470 is the only video card left by the company that uses the original Fermi architecture. She filled the gap in the lineup nVidia video cards with a price of $259 right between the GTX 460 at $200 and GTX 570 at $349. This is considering the starting price of $350. It's amazing how healthy competition has a positive effect on us gamers, isn't it?

Nvidia is discontinuing the GF100 less than a year after it was first announced. It is being replaced by the GeForce GTX 560 Ti video card. Instead of using graphics core version of the GF104, the GTX 560 uses a more advanced chip with improvements at the level of basic elements, first used in the GeForce GTX 580.

The result is a GPU with just under two billion transistors (1.95 billion according to Nvidia) that matches or exceeds the performance of the three billion transistor GF100 found in the GeForce GTX 470.

GeForce GTX 560 Ti | How to Create a Faster Gaming GPU

As we all know, the GeForce 500 series is almost no different in architecture from the 400 series.

GeForce GTX 580 and GTX 570 video cards are built on the GF110 - a redesigned GF100 with improved texture filtering, effective Z-clipping, and optimization at the transistor level, which helps to increase performance while maintaining equal power consumption.

Likewise, the Geforce GTX 560 Ti is based on the GF114 processor, a redesigned version of the GF104. Remember that GF104 already includes texture filtering improvements that were not included in GF100. That is, the throughput of 64-bit FP16 texels is doubled from two per clock to four per clock in each texture unit. The GF104 and GF110 have these capabilities, the GF100 does not. Moreover, nVidia decided not to carry over the Z-clipping improvements to the GF114, but instead leave the raster engine unchanged.

Ultimately, GF114 is functionally identical to GF104. In fact, nVidia even lists the same number of transistors (1.95 billion) and the GPU is manufactured using the same 40 nm TSMC process.

The modified processor operates at higher clock speeds with lower power consumption, which gives more high performance, but it's still an improved version of the GF104. Of course, the main difference is that to create the GTX 460 video card, nVidia disables one of the stream processors in the GF104, and in the GeForce GTX 560 Ti it uses an uncut GF114. Compared to the GTX 460, this means more clocks, more CUDA cores, and theoretically better geometry performance thanks to eight polymorph engines and eight additional texture units. All these factors combine to create a video card that not only completely replaces the GeForce GTX 460, but is also fast enough to outshine the GeForce GTX 470.

GeForce GTX 560 Ti | These characteristics look familiar

If you're already familiar with GF104, this chapter will be a bit like last year's GPU101 tutorial. The GF114 consists of two GPCs (Graphics Processing Clusters), each with four stream processors. As you already know, all eight stream processors are fully included in the GeForce GTX 560 Ti.

Below we present a short excerpt from GeForce review GTX 460, changing it a little:

"The GF114 uses 48 CUDA cores per SM, instead of the GF100's 32 cores per SM. Providing these more complex SMs with information requires higher instruction throughput, so we see another improvement: an increase in dispatch units per SM from two in the GF100 up to four in GF114 In addition, each SM now has eight texture units (instead of four).

Simply put, this is a “wider” GPU than the GF100 and GF110. The result is improved performance over the GF100 in the most popular gaming applications.

The chip inside is also slightly different. The full GF100 offers six GPC-independent ROP sections, each capable of outputting eight 32-bit pixels per clock (48 total). All six sections are connected to a 64-bit memory bus, giving a total bus of 384 bits. The GF114 receives a maximum of four partitions, giving up to 32 pixels per clock and a 256-bit bus."

GeForce GTX 560 Ti | Old suffixes, new cards

Since nVidia has fully included the GF114 GPU, the GeForce GTX 560 Ti's specifications are fully in line with the GPU. Thanks to 384 CUDA cores, the result is a GPU with great shader performance. 64 texture blocks appear, thanks to 8 wider SMs. There are the same number of them as in the GF110, however, to maintain the same functionality on the flagship GeForce GTX 580, 16 SM are required. nVidia is able to provide a GPU clock of 822 MHz, while the CUDA cores run at 1644 MHz (1:2 ratio is obvious). Just like the GeForce GTX 460 1 GB, the GeForce GTX 560 video card has four included ROP sections, delivering up to 32 pixels per clock. Four 64-bit buses add up to 256 bits. The video card has 1 GB GDDR 5 video memory with a frequency of 1002 MHz, which gives a bandwidth of 128.3 GB/s.

Externally, the GeForce GTX 560 Ti looks the same as the GeForce GTX 460. Both cards use axial fans that direct air flow to massive aluminum radiators with a copper base. Both cards are dual-slot with the same number of outputs, including two dual-link DVI connectors and one mini-HDMI port. Additionally, both cards require external power supply and are equipped with a pair of auxiliary six-channel inputs.

However, there are a few differences. The length of the GeForce GTX 560 Ti is 23cm, and the GTX 460 is 21cm. Under the plastic casing, the GTX 560 can be seen in one heat pipe more, allowing more heat to dissipate. There is also a metal plate covering the memory chips and the video card's power circuit. Previously, these components were simply left open.

Of course, now we are talking about reference design. Several have already appeared in the laboratory GTX video cards 560 from third party manufacturers, not corresponding to the reference version from nVidia. For example, Gigabyte video card has a length of 24 cm, two fans, 6+1 power stabilizer phases (versus 3+1 in the reference version) and a completely different printed circuit board.

nVidia equips its reference model the same power control circuit as on the GeForce GTX 580 video card. This circuit prevents overload of the voltage regulation circuit. The company says that whether or not to include them in the GeForce GTX 560 Ti is up to the video card manufacturers.

GeForce GTX 560 Ti | Oh yes, about this Ti...

Depending on your age, the Ti suffix on the NVidia GeForce GTX 560 Ti may or may not make sense. The story is that back in 2002, Nvidia limited "performance-oriented" video cards to the Ti suffix and "budget" video cards to the MX suffix. The GeForce Ti line had several various models, most were based on the NV25 GPU, but we clearly remember that the GeForce 4 Ti 4200 was the leader. If you're good at searching the internet, you might even be able to find a few of our reviews of this card from 2002.

Anyway, nVidia is bringing back the Ti suffix. When the company asked us what we thought about the name of this card, we were a little taken aback at first. "Nice, guys." But if you look more closely, the name makes sense. GF114 is a chip that can control any video card. And instead of starting to fool the consumer with digital indices (GTX 555, GTX 550, and so on), we assume that we will soon be able to see the GeForce GTX SE if nVidia introduces a less powerful copy of the GF114.

Differentiation by suffix is ​​not as expressive as the already accepted number system. If you remember the days of GeForce 4, then the less expensive MX series video cards were discontinued due to the fact that they trailed behind the slowest GeForce 3 video cards in terms of performance. These days, performance isn't everything, there are a lot of basic features to consider. But even now, it is possible to confuse an unsavvy person who pays more attention to games than to the hardware he buys.

Fortunately, we don't think Nvidia will resort to these tricks, but the GeForce GTX 460 and GF104 are still fresh in our memory. Perhaps, just like in July, partner video cards will begin to leak onto the market with more and more aggressive frequencies (and more high prices). By the time AMD launched its Barts-based GPU, NVidia was having a hard time getting anyone to test overclocked graphics cards from partners to compare them to AMD's reference graphics cards. Perhaps a more thorough segmentation of the GeForce GTX 560 line will not cause a repeat of this situation, which caused a lot of negativity within Nvidia.

GeForce GTX 560 Ti | Tessellation Performance

Given AMD's initial focus on tessellation in DirectX 11, as well as Nvidia's focus on tessellation in DirectX 11 now, we wanted to put both manufacturers' graphics cards through their toughest test yet to evaluate the architecture's impact on geometry performance. This test was supposed to be the Unigine Heaven Demo, but now that the HAWX 2 gaming test is available, we decided to switch from the artificial test to it.

When testing nVidia video cards, it was interesting to observe how the frame rate changed depending on the number of polymorphic engines in the samples. We found that the Fermi architecture's capabilities do not necessarily scale linearly as nVidia might assume. This is evident in the fact that the GTX 570 with the GF110 processor uses fifteen polymorphic engines, the GTX 560 uses eight engines, and the difference when turning tessellation on/off in HAWX 2 is only one percent. Obviously, geometry is not a critical parameter.

The situation with AMD is not much clearer. Cayman gets a second tessellation unit, while Barts only gets one, but still scales better. The only thing we can say is that the Cypress architecture offers more with tessellation enabled. An excellent option for today is the Radeon HD 5870, especially if you can find it at a deep discount. But if more developers follow Ubisoft's example with the geometry-intensive game HAWX 2, then more performance will be needed than on the geometry-optimized Radeon HD 6000 series graphics cards.

GeForce GTX 560 Ti | Hardware and tests

GeForce GTX 560 Ti | Test results

Not long ago we discussed the 3DMark11 test with nVidia over the phone. The company doesn't seem to like this test because it doesn't reflect future developments in gaming, namely the greater use of geometry to improve realism.

There's no doubt that 3DMark11 has an emphasis on lighting, particularly one of the DirectX 10 shaders. However, we had a candid conversation with Futuremark, who said that they designed their test with proper tessellation optimizations. Excessive scene geometry greatly affects the quality visual display, but can potentially significantly reduce the performance of entry-level and mid-range graphics cards. While 3DMark's first graphics test doesn't involve geometry at all, the second, third, and fourth tests include what we consider to be a reasonable amount of tessellation - meaning there aren't any scenes with triangles that big where we'd think: "Oh! We could add a little geometry here."

Of course, Nvidia claims that such a synthetic test is focused on future games, and not on those that are released today. However, Futuremark claims that game developers will not increase the difficulty of scenes for the sake of difficulty.

In the end, what's more important is how accurately 3DMark11 represents most real-world games. When you look at the benchmark results based on DirectX 11 games (everything we tested today, with the exception of World of Warcraft, is based on DirectX11), you'll notice that 3DMark11 actually doesn't perform well for some reason copes. The most noticeable difference is between the GeForce GTX 560 Ti and the Radeon HD 6870. In this test, the AMD video card outperforms nVidia, but in all cases with real games (with the exception of F1 2010 - more on that later), nVidia takes the lead. The GTX 470 also gives disproportionately low results compared to the Radeon HD 6870. It is interesting to note that even in its own line AMD Radeon The HD 5870 beats the Radeon HD 6950 in many tests, although 3DMark11, on the contrary, shows a significant lead.

Perhaps 3DMark11 is more representative of the future than nVidia thinks, if the dual tessellation engine helped the HD 6950 get so far ahead of the HD 5870. Or maybe AMD's confidence in DirectX 10 lighting shaders only highlights the strong performance of DX10, while nVidia, perhaps offering a more forward-thinking, future-proof architecture. In any case, do not base your purchasing decision solely on the results of the synthetic test. The people at Futurmark have gone to great lengths to eliminate the influence of AMD/nVidia's policies when participating in game development, but this in turn leads to different results in synthetic tests and in real games.

Metro 2033 (DX11)

At 1680x1050 and 1920x1080 resolutions, the GeForce GTX 570 and Radeon HD 6950 2GB stand out, especially with anti-aliasing enabled. On lower-end graphics cards there is little need to enable anti-aliasing, as Metro 2033 is one of the more demanding games we used in our testing.

Surprisingly, the Radeon HD 5870 holds a solid third place in all three tests, proving that AMD's previous generation of cards is still competitive more than a year after its release.

Nvidia's GeForce GTX 560 Ti sits in the middle, beating the Radeon HD 6870 without anti-aliasing, losing slightly at two resolutions with anti-aliasing, and then beating AMD's card at 2560x1600 (though not at playable frame rates). Perhaps even more remarkable is the fact that the 560 Ti beat out the GeForce GTX 470, which uses a more complex GPU, dissipates more heat, and consumes more power.

LostPlanet 2 (DX11)

It's really no surprise that GeForce cards dominate Lost Planet 2, since it sports the TWIMTBP (The WayIt's Meant To Be Played) logo. What's surprising is that the Radeon HD 6950 came in second place at 2560x1600.

The GeForce GTX 560 Ti is being put on par with the GeForce GTX 470 today, showing us that CUDA cores, by themselves, don't translate into increased performance - texture bandwidth is also important, and the GF114 architecture is better suited for mainstream gaming than some more powerful ones GPUs from NVidia.

Interestingly, the AMD Radeon HD 5870 lags behind in last place, but at 2560x1600 it moves up one spot.

Aliens Vs. Predator (DX11)

The Radeon HD 5870 regains its leadership among mid-range AMD video cards, even beating the Radeon HD 6950 2 GB in all three resolutions.

Meanwhile, the nVidia GeForce GTX 560 Ti falls in the middle, beating out the Radeon HD 6870 and GeForce GTX 470. It's actually quite difficult to get good frame rates with 4xAA enabled, so we recommend paying more attention to tests with anti-aliasing turned off.

Battlefield: Bad Company 2 (DX11)

Many of the games that were the first to support DirectX 11 were labeled as "DX11 verified games". They, of course, supported some DirectX 11 features, but this is not enough to show off the full capabilities of the API. Battlefield: Bad Company 2 was one of the first DX11-class games and it also received the "tested" title, as did AvP and DIRT2.

NVidia's GeForce GTX 560 Ti performed quite well at 1680x1050, slightly ahead of the GTX 470 and even the more expensive 2GB Radeon HD 6950. Further, at a resolution of 1920x1080, the GTX 560 Ti lost to the HD 6950, and at 2560x1600 it was even lower than the Radeon HD 6870. At this resolution, we see three video cards, such as the HD 6870, GTX 560 and GTX 470, which have almost the same number of frames in give me a sec.

The real news is that AMD's Radeon HD 5870 outperformed almost every card (except the GeForce GTX 570) at all three resolutions.

GeForce GTX 560 Ti | Test results. Continuation

F1 2010 (DX11)

We've heard that F1 2010 is also listed as a "tested" DX11-grade game, but it's actually inferior to DiRT 2 by removing tessellation support and limiting API support to shader mask and blur effects only. However, the game looks pretty good.

At 1680x1050 and 1920x1080 resolutions, the GeForce GTX 570 takes first place, but the Radeon HD 5000 and 6000 series are close behind. At a resolution of 2560x1600, AMD's Radeon HD 5870 even took first place.

Meanwhile, the GeForce GTX 560 Ti lags slightly behind. It was able to catch up with the GeForce GTX 470, which it should replace, but still loses to cheaper cards from AMD.

Just Cause 2

Although Just Cause 2 also has the TWIMTBP logo, the Radeon HD 5870 clearly shows us that the logo does not help nVidia graphics cards much, but this does not seem to be the case for the HD 6000 series cards.

Radeon HD 6950 2 GB was able to beat the GeForce GTX 560 Ti only at a resolution of 2560x1600. If the Radeon HD 5870 didn't deliver better numbers for less money, one might suspect it was because of the nVidia sponsorship logo.

World Of Warcraft: Cataclysm

Recently, we have been receiving many requests to include World of Warcraft in our constant reviews video cards, which is what we did. In addition, we wanted to conduct additional tests on scaling the performance of configurations with multiple video cards, since the game initially did not support such configurations. We will talk about this in more detail below.

As the performance level of single graphics cards continues to rise, the results here are pretty close to what we saw before this part. The difference is that we're running the game as shipped - in DirectX 9 mode (rather than experimental code to run DX11, which improves performance). As before, video cards from nVidia performed very well, GeForce GTX 570 and 560 Ti took the first two places in all resolutions.

The Radeon HD 6950 is fighting with the GeForce GTX 470 for third place. The remaining video cards are located below. It's worth noting that the Radeon HD 5870 ranks fourth at 2560x1600, just behind the Radeon HD 6950. Once again, the Cypress architecture demonstrates its superiority.

We ran a little ahead and added the results of the Radeon HD 4870 X2 for those who use older DirectX 10 class video cards. As you can see, the old flagship is still capable of competing with modern models, even at high settings at a resolution of 2560x1600. However, the HD4870 X2 has a slight problem with high quality when using the latest beta driver from AMD.

Scaling in SLI and CrossFire

In Lost Planet 2, the GeForce GTX 560 Ti dominates the Radeon HD 5870 and 6870 in single-player mode, so the fact that a couple of these cards do the same in SLI isn't really surprising.

Perhaps more interesting is that the Radeon HD 5870 and 6870 showed almost identical performance. We saw the Radeon HD 6870 produce more frames at 1680x1050 and 1920x1080, only outperforming the HD 5870 at 2560x1600. However, in CrossFire, the two HD 6870s showed a slight lead at 2560x1600, demonstrating the better scaling of Barts-based cards.

With Aliens Vs. Predator is a completely different story. In single-card mode, the Radeon HD 5870 finished second at 2560x1600, and it uses that advantage to take first place in the dual-card test. GeForce GTX 560 Ti and Radeon HD 6870, which previously showed almost identical results, are now very close in SLI and CrossFire.

In F1, in dual mode, the GeForce GTX 560 Ti, as in single mode, performed poorly, and two video cards from AMD again took the lead. The cheaper Radeon HD 6870 even achieved higher frame rates with 8x MSAA enabled than the GTX 560 Ti with it disabled.

The Radeon HD 6870's better scaling allows it to outperform two GeForce GTX 560 Tis in SLI mode, even though Nvidia's card was faster at 2560x1600 in single-card mode. What's more remarkable is that the Radeon HD 5870 outperforms other pairs of graphics cards in CrossFire and finishes in first place.

We literally banged our heads against the wall for days trying to figure out the problem. launch World of Warcraft: Cataclysm from Blizzard on SLI/CrossFire. AMD says it released a CrossFire profile update shortly after we finished testing. Nvidia still has bug reports indicating that the scaling is not as consistent as it should be. However, after a number of minor fixes and device driver updates, we are pleased to announce that you can now see zooming in the same Crushblow when flying through The Krazzworks. Nvidia says it doesn't see perfect scaling, but the performance of two GeForce GTX 560 Ti (with 8x anti-aliasing) is enough to match two Radeon HD 5870 (without anti-aliasing). The Radeon HD 6870 lags behind in the same way as in the results of tests of single video cards.

GeForce GTX 560 Ti | Overclocking

Gigabyte sent us their GV-N560SO-1GI video card, specially overclocked to 1 GHz (compared to the stock frequency of 822 MHz) for the GPU and up to 1145 MHz (compared to 1 GHz) for the video memory. Gigabyte says the card will retail for $269, up $20 from the previously announced price of $249.

As you can see, the GeForce GTX 560 Ti, running at 1 GHz, is significantly faster than the reference video cards from nVidia sent earlier. It's not as fast as the GeForce GTX 570 as Gigabyte promises, but the fact is that the results are pretty similar and the price is $80 less. Now, Gigabyte is using its own process to select GF114s that can sustainably operate at 1 GHz. Not every GPU is capable of performing at this level. However, it is fair to assume that most GeForce GTX 560 will run at average speeds. Nvidia confirms that frequencies around 900 MHz are quite typical for video chips tested previously.

GeForce GTX 560 Ti | Noise and power consumption

Power

Having recently changed the way we measure power, we wanted to make sure we were getting the right results. Having run several games on DirectX 9, 10 and 11, we couldn't get any more power consumption than Metro 2033 (although older games can apparently use higher power levels). However, we noticed that low resolutions are more power hungry than high resolutions. This makes some sense - unless you put artificially high loads on the GPU, the CPU is forced to work harder. Therefore, our power consumption test runs at 1680x1050 using AAA and 4x AF.

Obviously, the Radeon HD 4870 X2 was a real beast in its time. Video cards with two GPUs can still provide decent performance in modern games, but they literally suck energy while running. The absolute speed may not be that high, but the performance per watt ratio leaves much to be desired.

Next in power consumption is the GeForce GTX 570, followed closely by the GeForce GTX 470. The difference is that the GTX 570 is much more powerful, and the 4W difference is still small compared to the speed you'll get.

The GeForce GTX 560 Ti consumes less than the two graphics cards, averaging 25 Watts, but still more than AMD's Radeon HD 6950 2GB, which is faster at higher resolutions. The total difference of 10 watts in power consumption of the entire system puts the performance per watt ratio in AMD's favor.

Radeon HD 5870, GeForce GTX 460 and Radeon HD 6870 are close to each other and close the list. However, as you can see from the benchmark analysis, the Radeon HD 5870 is by far the performance favorite, more than a year after its introduction.

Noise

We couldn't include the Radeon HD 4870 X2 in our DirectX 11 tests because it only supports DirectX 10. But we were able to test it for noise, power consumption, and performance in World of Warcraft. This card serves as a good starting point for our acoustic test. Under load, the card is quite noisy.

However, at idle the Radeon HD 4870 X2 is no more noisy than other video cards reviewed here. The most buzzy cards tend to be flagships rather than mid-range and price cards. It must be said that nVidia really solves the noise problem of GF100-based video cards. The GeForce GTX 560 Ti is the quietest card of all tested. After the 10th launch of Metro 2033, it produces as much noise as a GeForce GTX 460 at idle.

GeForce GTX 560 Ti | Conclusion

We weren't impressed with top-end graphics until the launch of AMD's Antilles-based Radeon HD 6990. Radeon HD 5970 costs around $600 or more. But this video card was released more than a year ago. GeForce GTX 580 costs $500. If you buy several flagship cards in order to connect them in SLI or Crossfire, it will cost you an exorbitant amount of about $1000 or more. Now there are options in the middle class.

GeForce GTX 460, Radeon HD 6870, GeForce GTX 570, Radeon HD 6950 - any of these video cards in SLI or Crossfire will please you.

Now the GeForce GTX 560 Ti appears, occupying a place right in the center of this four video cards. In terms of performance, this card is almost exactly a replacement for the GeForce GTX 470. In terms of price, it will save you $10. In terms of noise, the 560 Ti is definitely the winner, although the GTX 470 is also not as noisy as the GeForce GTX 480. The GTX 560 Ti also wins in terms of power consumption, requiring an average of 25 W less than the GTX 470.

All of these things, taken individually, are incremental improvements that Nvidia should have made anyway. Thanks to the powerful and hot GF100 GPU, the GeForce GTX 470 sits in the company's product lineup alongside sleeker cards like the GTX 570 and 460. So the GeForce GTX 560 Ti doesn't impress even with all these little extras thrown together. But there's nothing wrong with this graphics card, it's just an improved version of the GTX 470, but $10 cheaper.

We are not enthusiastic about video cards from nVidia partners built on the GF114. Gigabyte's GV-N560SO-1GI didn't beat the GeForce GTX 570 in any of our tests, but it came very close in a few of them. Gigabyte says the graphics card will retail for $269. This is a fairly low price for a card with a 1 GHz GPU clock speed. However, can the company maintain this price when the GeForce GTX 570 sells for $349?

What about the competitors? While the GeForce GTX 560 Ti is faster in almost every way than the Radeon HD 6870 and only $20-$40 more expensive, it would be unfair not to note that some HD 6870s sell for less than $200 after discounts. Even now, you can spend $100 less on a Crossfire configuration than on SLI with a GeForce 560 Ti. The Radeon HD 5870 also looks pretty good, but only if you can find it cheaper than the GeForce.

Surely now best time to update a video card that is one or two generations out of date. If you already bought a decent card last year, the GeForce 560 Ti hardly gives you a reason to spend the extra $250. The 560 Ti is a little faster, a little more efficient, a little quieter, and a little cheaper than the GeForce GTX 470, which you can kiss goodbye to.

Postscript: AMD appears with twenty Radeon HD 6950 1 GB

We love good competition, but bad marketing can ruin everything.

AMD has contacted us to say it will ship a 1GB version of the Radeon HD 6950 in mid-February for between $269 and $279. “Well, good,” we thought then. "The worry will be when she comes out... in three weeks. Until then, these are just empty words on paper."

After some time, the video cards appeared and, of course, we put them through our full set of tests.

In almost every case, the smaller framebuffer (and tighter memory timing) produces one or two frames per second more than the 2GB model. This is not worthy of its own diagram. In the game Metro 2033 with 4x MSAA and 16x AF enabled, the 1 GB video card did not have enough buffer capacity to maintain performance and FPS dropped to 10 frames per second versus 30 on the 2 GB version.

To be honest, perhaps AMD should have introduced the Radeon HD 6950 first, rather than a $300 2GB graphics card. However, they don't do this, perhaps to avoid competition with the $240-$260 Radeon HD 6870. Obviously, AMD released them together before the start of sales of the video card from nVidia, and they seemed to hold the option in reserve until they saw what the GTX 560 Ti could do. Losing sales of the HD 6870 to the cheaper Radeon HD 6950 sounds better for AMD than losing sales to Nvidia.

Now, in the face of the GeForce GTX 560 Ti, AMD says the 1GB Radeon HD 6950 will be priced at $259, while the Radeon HD 6870 will drop to $219. Whether these prices will remain or not remains to be seen. In any case, our conclusion regarding the GTX 560 Ti remains unchanged. It still doesn't give us a reason to upgrade. If we compare video cards in terms of performance, solutions from AMD look better.

Dirty gossip creates...

The story would have ended if the Radeon HD 6950 1 GB had not appeared. On the eve of the nVidia launch, one of the partners told us that AMD had the HD 6950 1GB up for sale on Newegg for $259. At the time of writing, there were 20 video cards available. We spoke to Newegg and they confirmed this. By the time you read this, they will no longer be on sale.

Obviously, 20 people will be very happy to get a graphics card that is faster than the GeForce GTX 560 Ti for exactly the same price. Everyone else will have to wait. Either way, this is one of the weakest marketing moves we've seen. AMD's graphics card is strong enough to stand on its own, even at the originally projected price of $269 - $279.

It's unlikely that you'll buy the HD 6950 1GB if you're gaming on a three-monitor Eyefinity configuration in CrossFire. To do this, you will need a graphics card with 2GB of memory to support 5760x1080 resolution with AA and AF active. However, the Radeon HD 6950 1 GB is more attractive at 1920x1080 with maximum detail or at 2560x1600, but with slightly less detail.

In essence, thanks to memory, the Radeon HD 6950 creates a worthy competitor to nVidia's GeForce GTX 560 Ti.