Hardware support for tessellation has been around for a long time

As you know, the ability to tessellate at the hardware level was added by ATI a very long time ago, but without standardization it could not gain serious support. Therefore, the era of tessellation began only with its introduction in the corresponding DirectX 11 API. AMD's marketing slides mention the 7th generation tessellator, but it makes more sense to talk about the 2nd generation. AMD does not hide the fact that the geometric performance of its new products is lower than that of Fermi solutions. However, it is claimed that the achieved level of performance is optimal for modern games.

The fact is that the operation of the tessellation mechanism is determined by the game engine, which can set the tessellation factor depending on the specified graphics settings, the in-game distance to the object and other parameters. The tessellation factor determines how many polygons will be used when splitting the primitive. Obviously, there is a certain “golden mean” here, a value of a factor whose further increase will no longer lead to a noticeable improvement in picture quality.

AMD claims that this factor value roughly corresponds to polygons "consisting" of 16 pixels. Moreover, when the polygon size is reduced to less than 16 pixels, overshading begins on modern video cards - a phenomenon associated with a sharp drop in performance. Accordingly, the updated tessellator has been optimized to work with the average tessellation factor, where the specified twofold advantage is achieved. We will return to this issue when we consider the test results.

MLAA - new anti-aliasing mechanism

Along with the release of new video cards, AMD also optimized the anisotropic filtering mechanism and introduced a new anti-aliasing mode. And if the optimization of anisotropic filtering is not very interesting to us, then we will dwell on the new anti-aliasing mode in more detail. Meet: Morphological Anti-Aliasing. In fact, it is only new for PCs; on consoles, this method (at least the principle itself) was used previously. And it is called anti-aliasing only conditionally.

Unlike traditional methods, MLAA is a pure post-processing that works exclusively with a 2D image. It analyzes the image for too sharp and contrasting transitions between pixels and softens them by adding halftones. The advantages of this approach are obvious: lower load on the video card compared to MSAA and complete versatility. Being forced in the drivers, MLAA should work absolutely correctly and equally in all games, since it practically does not interact with their engine. The disadvantages are also obvious - sometimes the method may not work correctly due to the lack of real information on the in-game geometry. However, overall this approach seems very promising to us. We will monitor developments. NVIDIA can adapt it in the near future, fortunately the adaptation should not be very difficult.

There is probably nothing easier than choosing a gaming video card on the computer market entry level. After all, for most buyers in this segment, the cost of a graphics accelerator is more important than performance. However, there are exceptions - when the manufacturer offers the user to purchase an inexpensive device that can surprise the future owner with decent potential in work and games.

The focus of this article is the HD 6850 video adapter from the famous American concern ATI. The reader is invited to take a closer look at the game solution: see full review and find out the technical characteristics of an interesting device. Reviews from owners and recommendations from experts will help a beginner make a choice on the market.

Elite Representative

Oddly enough, the characteristics of the 6850 hint to the potential buyer that this is a representative of the gaming class, and not some kind of budget device. After all, according to computer market trends, only powerful video adapters are equipped with a high-speed 256-bit data transfer bus. It's no secret that the throughput of a video card is directly proportional to the frequency of the graphics core and bus.

The second hint is the size printed circuit board and cooling system. Only equipped with a turbofan and a closed casing for proper distribution of cooled air. It remains only to understand for what purpose the manufacturer set a low price for such an interesting and attractive product.

The key to high productivity

It is better to start the review of the HIS Radeon HD 6850 video card with technical characteristics, because it is thanks to them that the graphics accelerator is able to demonstrate high performance in gaming applications. The Barts Pro graphics core clocks at only 775 MHz. As for memory, the video card is also not in perfect order; modern GDDR5 modules have too small a volume - 1024 MB. Apparently, it was this indicator that forced the manufacturer to place its product in the lower price segment gaming class.

The device chip contains about 1.7 billion transistors and 960 stream processors. 48 blocks were allocated for textures (the previous model had 72). It is worth noting that the chip itself is produced using a 40-nanometer technical process.

About technologies and software

The supported version of DirectX 11 may be outdated, but most gaming applications are created specifically for this set of libraries. And it is unlikely that the situation will change radically in the next few years. In the HD 6850 video adapter, the graphics accelerator supports 3D visualization at the hardware level. This solution allows the user to connect TVs and monitors to the computer that support three-dimensional image display on the screen.

Naturally, the manufacturer has equipped its product with a built-in processor for processing surround sound, as well as a special decoder that allows you to display images in FullHD format. The only thing that confuses the AMD Radeon HD 6850 video card is the characteristics of the energy saving system. The device, even when idle, consumes too much electricity and does not automatically switch to energy-saving mode.

Power system features

In general, more time should be devoted to the topic of energy consumption. After all, the uninterrupted operation of the video card is a priority for every user who decides to find out in more detail the specifications of the Radeon HD 6850 graphics accelerator. User reviews in the media are quite clear and understandable - the gaming video adapter is not as simple as it seems.

The declared power consumption of the video card in the factory version is 127 Watts, but this parameter can almost double when overclocking the graphics core and memory bus. The most interesting thing is that the performance of the video adapter during overclocking shows an increase of only 20-25%. Yes, from the outside it looks really strange. As for the requirements for the power supply, it is better for a game lover not to skimp here and provide the gaming platform with a powerful device (500 Watts and above).

PCB components

While reviewing the components of the HD 6850 video card, I would like to draw the user’s attention to one interesting factor. The manufacturer AMD took the liberty of placing the power elements around the entire perimeter of the printed circuit board, rather than placing them in a separate unit, as is the case with more expensive gaming video adapters. The graphics core is located in the middle of the board, the memory modules are spaced around the processor socket, and the chips and stabilizers are soldered near the video output board.

Icy air flow

The domestic buyer has already become acquainted with the product IceQ X Turbo HIS Radeon HD 6850, the characteristics of which allow the video adapter to legitimately position itself in the entry-level gaming segment. The peculiarity of this graphics accelerator is not high performance, but complete noiselessness in operation. The manufacturer managed to achieve such results thanks to a proprietary cooling system.

A huge aluminum radiator not only completely covers the graphics core, memory modules and batteries, it, like a sarcophagus, rises up to the protective casing, hiding a huge fan. A low-speed cooler with large blades operates almost silently, pumping a large volume of cold air onto the radiator.

As for the performance of the HD 6850, the company's technologists managed to slightly improve the performance. Thus, the graphics core in normal mode is capable of demonstrating 820 MHz, and the memory operates at a frequency of 4400 MHz. But things are bad with overclocking of the chip; the processor is limited to a frequency of 850 MHz. The reason for this is the cooling system, which is designed only for basic characteristics devices.

Serious player in the market

But the HIS Radeon HD 6850 Overlocker video card clearly claims to take a leadership position in the mid-range gaming class segment. By at least, it is this product that is compared in many test laboratories with more expensive solutions from competitor Nvidia. Again, a modified power management system on the printed circuit board and a decent cooling system allowed the graphics accelerator to take the lead.

Taking the psychological barrier, the GPU shows the user a frequency of 1122 MHz. But the manufacturer did not experiment with memory - 4500 MHz. It is worth noting that the board contains Hynix modules (a photo of the original chip can be seen on the manufacturer’s website), which have a hardware memory frequency limit at this mark, so the company’s technologists do not recommend the owner to continue overclocking GDDR5 modules.

A separate story on cooling

The more productive and expensive HIS Radeon HD 6850 solution, the characteristics of which were discussed above, is equipped with a very interesting cooling system. The manufacturer clearly did not skimp on non-ferrous metals, so it is not surprising that copper became the predominant material for the radiator. There is no need to explain that in terms of thermal conductivity, red metal is much more efficient than aluminum, so users will not have any questions about cooling the platform.

The owner will also be pleased with the power reserve of the installed fan. In their reviews, many users claim that the HIS video card cooling system has excellent overclocking potential. The characteristics of the Radeon HD 6850 increase by almost 10%. The graphics accelerator easily accelerates to 1244 MHz and demonstrates stable operation. Only the noise of the fan on the board slightly gets on my nerves.

Similar products from other manufacturers

HIS is not a monopolist in the computer market. Quite a lot of other manufacturers have paid attention to this interesting product. The Radeon HD 6850 video card, the characteristics of which are directly proportional to the cooling efficiency, has attracted interest from XFX, PowerColor, MSI, Gigabyte, Asus and other equally well-known brands. However, no one has been able to surpass HIS products in terms of performance and overclocking efficiency.

The secret is quite simple - none of the market representatives delved into the problem of the AMD printed circuit board or made changes to the hardware. Therefore, installing a more powerful cooler did not bring victory to the Taiwanese companies Asus and Gigabyte. But there was potential, considering that world-famous brands equipped the graphics accelerator with proprietary cooling systems. That's why the description of the characteristics of similar video adapters based on the Radeon HD 6850 does not impress users.

Finally

In the price category up to 6,000 rubles, there are quite a lot of interesting solutions that can cope with many toys, but experts recommend that the buyer pay attention to this video adapter. The fact is that it is one of the few representatives that is capable of operating on a 256-bit bus. And we shouldn’t forget about the supported technologies; after all, DirectX 11 has a future.

The Radeon HD 6850 video card, reviews of which are more positive in the media, will be of interest to game lovers who want to use resource-intensive applications at low and medium quality settings. This is the only thing that a buyer will have to sacrifice when paying a small fee for an attractive video card in the computer market.

Describing Radeon HD 6850/6870, which were previously coded Barts.

Therefore, today we are catching up and bringing to our readers two practical parts, where we will study new AMD products in detail.

As usual, in this second part, we will study the video cards themselves, as well as get acquainted with the results of synthetic tests.

GPU: Radeon HD 6850 (Barts) Interface: PCI-Express x16 : 775/775 MHz (nominal - 775/775 MHz) : 1000 (4000) MHz (nominal - 1000 (4000) MHz) Memory bus width: 256 bit Number of vertex processors: − − : 960 Number of texture processors: 48 (BLF/TLF) Number of ROPs: 32 Dimensions: 250×100×33 mm (the last value is the maximum thickness of the video card) PCB color: black RAMDACs/TMDS: GPU integrated Output jacks VIVO: No TV-out: not displayed : CrossFire (Hardware)
GPU: Radeon HD 6870 (Barts) Interface: PCI-Express x16 GPU operating frequencies (ROPs/Shaders): 900/900 MHz (nominal - 900/900 MHz) Memory operating frequencies (physical (effective)): 1050 (4200) MHz (nominal - 1050 (4200) MHz) Memory bus width: 256 bit Number of vertex processors: − Number of pixel processors: − Number of universal processors: 1120 Number of texture processors: 56 (BLF/TLF) Number of ROPs: 32 Dimensions: 270×100×33 mm (the last value is the maximum thickness of the video card) PCB color: black RAMDACs/TMDS: GPU integrated Output jacks: 2×DVI (Dual-Link/HDMI), 2×mini-Display Port, 1×HDMI VIVO: No TV-out: not displayed Multiprocessing support: CrossFire (Hardware)

AMD Radeon HD 6850 / 6870 1024MB 256-bit GDDR5, PCI-E

Each card has 1024 MB of GDDR5 SDRAM memory, located in eight chips on the front side of the PCB. It makes sense to say that both cards require additional power, with the 6870 requiring two 6-pin connectors, and the 6850 requiring one connector. About cooling systems. AMD Radeon HD 6850 1024MB 256-bit GDDR5, PCI-E It is clearly visible that the CO consists of two parts - a central cooler and radiators for cooling the memory, which work as if on their own, and the central device only cools the core. The device is of a cylindrical type, when a cylindrical fan is attached to one end, driving air through a radiator installed above the core. Despite the copper base, the radiator itself is small. In general, the device is quite quiet, and clearly indicates that the core heating is not so great. AMD Radeon HD 6870 1024MB 256-bit GDDR5, PCI-E The device is similar in principle, but the difference is that the central cooler already cools both the core and memory chips, so the radiator is reinforced (increased in size). And the cylindrical fan is more powerful. However, the overall device is still low noise. We conducted a temperature study using the EVGA Precision utility (author A. Nikolaychuk AKA Unwinder) and obtained the following results: AMD Radeon HD 6850 1024MB 256-bit GDDR5, PCI-E AMD Radeon HD 6870 1024MB 256-bit GDDR5, PCI-E As we can see, both COs work equally efficiently, and heating does not exceed 80-81 degrees, which is very good for this kind of modern accelerators. Maximum power consumption of cards under load: 6850 - 150 W, and 6870 - 180 W. Equipment. Considering that reference samples never have configurations, we will omit this question. Installation and drivers Test bench configuration: Computer based on Intel Core I7 CPU 975 (Socket 1366) Intel processor Core I7 CPU 975 (3340 MHz); Asus P6T Deluxe motherboard on Intel chipset X58; RAM 6 GB DDR3 SDRAM Corsair 1600MHz; hard drive WD Caviar SE WD1600JD 160 GB SATA; power supply Tagan TG900-BZ 900W. operating system Windows 7 64bit; DirectX 11; Dell 3007WFP monitor (30″); ATI drivers version Catalyst 10.10; Nvidia versions 262.99/260.99. 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. 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 brief description: , . Synthetic tests were carried out on the following video cards: Radeon HD 6870 HD 6870) Radeon HD 6850 with standard parameters (further HD 6850) Radeon HD 5830 with standard parameters (further HD 5830) Radeon HD 5770 with standard parameters (further HD 5770) GeForce GTX 470 with standard parameters (further GTX 470) GeForce GTX 460 with standard parameters, model with 1 GB of memory (hereinafter GTX 460) To compare the results of new models of Radeon HD 6800 series video cards, these solutions were chosen for the following reasons: Radeon HD 5830 is the closest in price and least productive solution based on the Cypress chip, HD 5770 is the company’s previous solution for the mid-price range (the same for which the new models are designed for), based on the Juniper video chip. Namely, these Nvidia solutions were taken because the Geforce GTX 470 is one of the cheapest cards on the previous top-end GPU, which has now dropped in price and become a competitor to the HD 6870 (there is simply no point in considering the GTX 465 as discontinued). Well, the GTX 460 with a gigabyte of video memory was taken as a direct competitor to the junior model of the HD line - 6850. 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: Let us repeat once again that in this RGB8 texture filtering test, most video cards show numbers that are far from theoretically possible. And then, in the test from the 3DMark Vantage package, there are more vital numbers. The results of our texture synthetics in the case of HD 6800 video cards fall far short of the peak values; it turns out that the new chip selects only up to 42 texels per clock cycle from 32-bit textures during bilinear filtering in this test, which is a third less than the theoretical figure in 56 filtered texels. It is not surprising that in heavy modes, the HD 6800 family cards show such high performance that they are significantly ahead of their rivals from Nvidia. The difference between the HD 6000 and HD 5000 families in different conditions turned out to be interesting. If in cases with a large number of textures, where the number of TMUs and their frequency have the greatest impact, options based on new GPUs win, then with a small number of textures per pixel, the HD 5000 family is already ahead. It’s also funny that we already noted a similar approach in the review of the Geforce GTX 580 - apparently, AMD also slightly changed the balance in the new GPUs and/or drivers and sacrificed easier conditions for more difficult ones. Let's look at the same results in the fill rate test: Well, these numbers show the fill rate, and in them we see everything the same, except perhaps taking into account the number of pixels recorded in the frame buffer. The maximum result remains with new solutions from AMD, which have a larger number of TMUs and are more efficient in this synthetic test. In cases with 0-3 overlaid textures, the solutions considered today are slightly inferior to the previous generation of AMD video cards, and in difficult conditions they are ahead of them. Direct3D 9: Pixel Shaders tests The first group of pixel shaders that we are considering is very simple for modern video chips, it includes different versions pixel programs of relatively low complexity: 1.1, 1.4 and 2.0, found in old games. The tests are very simple for modern GPUs and do not show all the capabilities of modern video chips, but they are still interesting for assessing the balance between texture fetches and mathematical calculations, and especially when comparing GPUs that differ in architecture. But in this case, there are no special differences between HD 5000 and HD 6000, so the results shown are similar, taking into account frequencies, of course. Performance in these tests is limited mostly by fillrate and texture unit speed, but takes into account block efficiency and texture data caching. The new Radeon models are slightly faster than the previous ones in pairs: the HD 6870 is faster than the HD 5830, and the HD 6850 is faster than the HD 5770. Well, they are all ahead of the two GeForce models - the GTX 470 in these tests shows results only at the level of the HD 5770, and even the GTX 460 is clearly visible lack of texturing speed. Let's look at the results of more complex intermediate pixel programs: Oddly enough, it turned out about the same. The Cook-Torrance test is more computationally intensive, and the difference in it roughly corresponds to the difference in the number of ALUs and their frequency. And because of this, this test is better suited for the AMD architecture, which has a larger number of mathematical units, and even the Radeon HD 5770 shows results at the level of a GF100-based video card. The “Water” procedural water rendering test, which is highly dependent on texturing speed, uses dependent sampling from textures of large levels of nesting, and the maps in it are arranged by texturing speed, adjusted for different efficiency of TMU use. There are two clear groups in this test: the HD 6870 and HD 5830, plus everyone else. The new Radeon models are again slightly faster than the older pairs - a good result. Direct3D 9: pixel shader tests Pixel Shaders 2.0 These DirectX 9 pixel shader tests are more complex than the previous ones, they are close to what we now see in multi-platform games, and are divided into two categories. Let's start with the simpler version 2.0 shaders: Parallax Mapping- a method of texture mapping familiar to most modern games, described in detail in the article. Frozen Glass- complex procedural frozen glass texture 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: Are these universal tests that also depend on the speed of ALU units? and the texturing speed, the overall balance of the chip is important in them. The performance of video cards in the Frozen Glass test is very similar to what we saw above in Cook-Torrance. The HD 6870 is again faster than the HD 5830, and the HD 6850 is faster than the HD 5770. Well, in general, AMD's solutions turned out to be faster than cards Nvidia this time too. In the second “Parallax Mapping” test, Nvidia’s solutions perform little better, and the HD 5770 competes with the GTX 460, and the GTX 470 is close to the HD 6850. The speed in the test is probably limited largely by mathematical performance. Let's consider these same tests in a modification with a preference for samples from textures over mathematical calculations: But with texturing speed, the latest modifications of AMD graphics architecture chips are doing very well, and therefore they are only increasing their advantage. And even the top-of-the-range GTX 470 is inferior even to the HD 5770 in these texturing-focused tests. Well, the new heroes from the HD 6800 family are far ahead. HD 6870 and HD 6850 are still faster than their predecessors, which is understandable theoretically. But these were somewhat outdated tasks, mainly focusing on texturing or fillrate, and then we will look at the results of two more pixel shader tests - but this time version 3.0, the most complex of our pixel shader tests for the Direct3D 9 API, which are much more revealing in terms of modern games on PC. The tests differ in that they place a greater load on both the ALU and 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. As usual, in our most difficult DX9 tests, Nvidia video cards are already stronger than AMD solutions. And it seems that with tests of complex pixel shaders version 3.0 for AMD solutions, everything is not as cloudless as it might have seemed previously. At the same time, both PS 3.0 tests are quite complex, the speed in them depends little on memory bandwidth and texturing, but the code has a large number of branches, which the new Nvidia architecture copes with very well. And in these tests, even the HD 6870 is difficult to keep up with the GTX 460, not to mention the GTX 470, which is the undisputed leader in this pair of test tasks. However, not everything is so bad, and at least the new solutions have confidently surpassed their predecessors from the HD 5000 series. It’s just that Nvidia’s position is traditionally stronger in these tasks. 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 most low 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 both on the number and efficiency of TMU blocks and on the fill rate with bandwidth, but to a lesser extent. The results in “High” are approximately one and a half times lower than in “Low”, as it should be according to theory. In Direct3D 10 tests of procedural fur rendering with a large number of texture samples, Nvidia solutions are usually strong, but the latest AMD architecture has caught up with them, and how! As a result, the HD 6870 is even slightly ahead of the GTX 470 in this test, and the HD 6850 performs at the level of the HD 5830 and better than the GTX 460. The effect of effective fill rate and bandwidth is clearly visible in how far behind the HD 5770 with a 128-bit bus memory. 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 impact: Enabling supersampling increases the theoretical load by four times, and this time the comparative results of Nvidia solutions drop even lower. Now the HD 5770 is at the level of the GTX 460, and the HD 6870 is one and a half times faster than the GTX 470. The difference between the cards from the HD 6000 and HD 5000 lines remains approximately the same. The second DX10 shader test measures the performance of complex pixel shaders with loops with a large number of texture samples and is called Steep Parallax Mapping. At low settings it uses 10 to 50 texture samples from the height map and three samples from the main textures. Enabling heavy mode with self-shadowing doubles the number of samples, and supersampling quadruples this number. The most complex test mode with supersampling and self-shadowing selects from 80 to 400 texture values, that is, eight times more than the simple mode. Let's 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 in many ways similar to the previous ones. In the updated D3D10 version of the test without supersampling, the HD 6870 becomes the leader among the selected video cards, and the HD 6850 competes with the HD 5830 with varying success. Nvidia video cards are slightly below the AMD solutions, and the GTX 460 again showed results at the level of the cheaper HD 5770. We'll see. What will change the inclusion of supersampling, it should cause an even greater drop in speed on Nvidia cards. When supersampling and self-shadowing are enabled, the task becomes even 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 the tested 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 5770 is already ahead of the GTX 460, and the HD 6850 provides rendering performance similar to the speed of the GTX 470. Comparative figures in pairs HD 6870 and HD 5830, as well as HD 6850 and HD 5770 were repeated again, the difference in favor of the latest models is about the same . Based on these tests, we can conclude that both cards of the HD 6800 line coped with shader tasks perfectly, which is not surprising, since the new GPU has a fairly large number of ALU units. Direct3D 10: PS 4.0 Pixel Shader Tests (Compute) The next couple of pixel shader tests contain a minimum number of texture fetches to reduce the performance impact of the TMU units. They use a large number of arithmetic operations and 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 usually correspond to the difference in frequencies and the number of ALUs. And this explains the fact that AMD solutions are clearly significantly more productive in these tests. Modern AMD architecture in such cases has a great advantage over competing video cards from Nvidia. This has been confirmed once again, even the HD 5770 is faster than both Nvidia cards, not to mention the new HD 6870 and HD 6850. As for comparing the new and old families of AMD video cards, the HD 6870 is the clear leader in the test, outperforming the most weak card comparison - GTX 460. And the HD 6850 showed results at the level of the HD 5830, which does not correspond slightly to the theoretical difference - in this case, the new GPU worked more efficiently than the old one. But all other solutions are located approximately according to theory, this applies to both Nvidia and AMD cards. Let's look at the second shader calculation test, which is called Fire. It is heavier for an ALU, and there is only one texture fetch, and the number of sin and cos instructions has been doubled, to 130. Let's see what has changed with increasing load: And this time all GPUs remained in approximately the same positions; we can only note the fact that the HD 5830 in this test is still ahead of the HD 6850. And, unlike the previous test, this is already fully consistent with the theory, since the HD 5830 and should be a little faster. Otherwise, everything is the same, since the rendering speed is limited solely by the performance of the shader units, so AMD cards are far ahead of Nvidia solutions - the usual defeat is evident. Direct3D 10: geometry shader tests The RightMark3D 2.0 package has two speed tests for geometry shaders, the first option is called “Galaxy”, the technique is similar to “point sprites” from previous versions 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. The task for modern video cards is not particularly difficult; performance in general is limited not only by the speed of geometry processing, but also by memory bandwidth to a certain extent. And here we see for the first time the result of architectural changes in the form of improved geometric performance of the Barts video chip. Both video cards of the new Radeon HD 6800 family showed results that were noticeably faster than the solutions in the HD 5000 line. Moreover, they both overtook the GTX 460, but the new HD 6870 was just a little short of defeating the GTX 470. In any case, the HD 6800's execution of geometry shaders has become noticeably more efficient, and the new chip is faster than all previous AMD chips in this test. Let's see how the situation changes when we transfer part of the calculations to the geometry shader: When the load in this test changed, the numbers for both Nvidia and AMD solutions remained almost unchanged. New video cards of the HD 6800 family in this test almost do not respond to changes 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. And, interestingly, they behave more like Nvidia video cards, rather than HD 5830 and HD 5770. The latter just slightly improved their performance in this case. Well, let's see what changes in the next test, which assumes a large 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 Direct3D 10 feature - 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: Relative results in different modes again correspond to the load: in all cases, performance scales well and is close to theoretical parameters, according to which each subsequent level of “Polygon count” should be less than twice as slow. In this test, rendering speed is most limited by geometric performance. New AMD video cards show significantly stronger results compared to older models, which is explained by architectural changes in the GPU. And although the GeForce GTX 470 remains the leader of the test, it is closely followed by the HD 6870. And in the pair of HD 6850 and GTX 460, the AMD solution wins altogether. This clearly indicates the presence of serious optimizations for processing geometric data in Barts. But the numbers should change in the next diagram, in a test with more active use of geometry shaders. It will also be interesting to compare the results obtained in the “Balanced” and “Heavy” modes with each other. But in this test we still see a clear difference between chips with a traditional graphics pipeline (all Radeons, including new solutions based on Barts) and chips with Fermi architecture. Yes, the GF104 lags behind in the speed of execution of geometry shaders in this test, showing a worse result than both Barts, but this is easily explained by the reduced capabilities of geometry processing in a mid-price range chip. But look at the result of the GTX 470, which is based on the GF100 chip - it is significantly higher than all other video cards tested today. The capabilities of top-end Nvidia chips in processing geometry and the execution speed of geometry shaders greatly exceed their mid-price solutions, as well as all competing AMD solutions. But still, the new Barts chip, used in the HD 6800 line, allowed it to overtake the GF104 in these tests and significantly reduce the gap even with the recent top-end Nvidia chip. Excellent result! Direct3D 10: texture fetching 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. And this is clearly visible in the results of the Radeon HD 5770, which has lower bandwidth and is far behind other test participants. The difference between the other solutions is not that big, although it is interesting that the GTX 470 turns out to be a leader in two difficult modes, and the HD 6870 is in the easiest one. But what's important is that both cards in the HD 6800 family are ahead of the previous generation HD 5830. 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 hardly changed, but for some reason both Nvidia cards lost even more performance in the lightest mode. In this case, the GTX 460 and GTX 470 remain out of reach of their rivals, but only in two difficult test modes. Both cards in the HD 6800 line are still ahead of the old ones. The influence of bandwidth is noticeable here too - the result of the HD 5770 is quite low. 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 similarly to the previous test. But the results in the “Waves” test are not at all similar to what we saw in the previous diagrams. AMD products do not have an overwhelming advantage here, but in this test it was the two new cards that became the leaders, with the GTX 470 and HD 5830 slightly behind them. The GTX 460 shows even lower performance, and the Radeon HD 5770 is usually and deservedly the slowest. Apparently, the test is still affected by bandwidth. Let's consider the second version of the same test: There are almost no changes, although Nvidia cards have lost ground a little and now the GTX 470 matches the HD 5830 in speed, except for the most difficult mode. Again we see that Nvidia video cards have become stronger in heavy mode, but lose a lot in easy mode. In any case, the results of the new GPU Barts, as well as video cards based on it, are very good in the second vertex sample test, and the new GPU even became the fastest in this test. 3DMark Vantage: Feature tests Synthetic tests from 3DMark Vantage can show us something we previously missed. The feature tests of this test package 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 texture fetch speed test. Feature Test 1: Texture Fill 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. As you can see, the Futuremark test also does not show the theoretically possible level of texture fetch speed, although the efficiency of new AMD cards in it is slightly higher than in ours. Nvidia cards They also make more efficient use of the available texture units, and this texture test produces a different ratio of results compared to ours. And we believe that these numbers are more similar to the real state of affairs. The two new video cards of the Radeon HD 6800 family showed results slightly better than their paired rivals: HD 5830 for the HD 6870 and HD 5770 for the HD 6850. It can be seen that Barts has increased mainly mathematical performance. Both Nvidia video cards still continue to show not very high results, but they have already come closer to AMD solutions. The GTX 470 was roughly on par with the HD 5770, while the GTX 460, which has more TMUs, was almost as good as the HD 6850. Feature Test 2: Color Fill This is a 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. In this test we see two groups of video cards, arranged in accordance with theoretical fill rate figures, but without taking into account the influence of video memory bandwidth. Vantage's numbers show the performance of the ROP units and only that, but not the amount of throughput. Therefore, the results of the HD 5830, HD 5770 and GTX 460 are very close, as are the numbers of both new cards and the GTX 470. However, the HD 6870 shows the best result, 10 percent ahead of its rival from Nvidia, and the HD 6850 is not only ahead of its direct competitors, but also takes precedence over the GTX 470. So, we note the high fill rate of the new video card models, corresponding to the level of the recent top from a competitor. 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. This 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, the correct balance of GPU and video memory blocks is important. Significantly affects the speed and efficiency of branching in shaders. The comparison results of AMD graphics cards in the chart are quite similar to what we saw in the 3DMark Vantage texture performance test. But for Nvidia this is not the case - in this case, the GTX 470 received a clear acceleration, apparently due to the different efficiency of executing shader programs with branches. And in general, it’s a little surprising that it was the GTX 460 that became the outsider in this test, losing even to the HD 5770. But the new heroes from AMD are again, in pairs, albeit a little, but still faster than their predecessors in the form of the HD 5830 and HD 5770. 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. The rendering speed in this test depends on several parameters at once, the main of which are geometry processing performance and the efficiency of geometry shaders. And therefore, video cards produced by Nvidia feel like fish in water, significantly ahead of competitors from AMD. The difference between Nvidia solutions from different price ranges is also clearly visible. Specifically, the recently introduced video cards of the new Radeon HD 6800 series have a higher rendering speed in this test than the previous line, since Barts has increased the speed of geometry processing and execution of geometry shaders. And although the HD 6870 still does not reach the GTX 460, it is significantly ahead of other tested solutions from the company, and the HD 6850 is somewhere close. 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 most of all loads shader units with vertex calculations; stream out is also tested. The results of the next test are very similar to those we saw in the previous diagram, but here the speed of geometry processing is even more important than in the previous test. That is why the old generation in the form of Radeon HD 5830 and HD 5770 cards lagged behind both GeForces, which are leaders in comparison, and the new line of video cards reviewed today. And both models based on Barts showed good results, not losing too much to the GTX 460. In general, in synthetic tests simulating fabrics and particles from the 3DMark Vantage test package, where geometry shaders are actively used, the new Barts chip performed just fine, since it had accelerated geometry processing. And although both solutions of the HD 6800 line continue to lag behind their competing video cards, the difference between them has noticeably decreased - Barts has done a good job on this improvement. But we still expect even greater architectural changes from AMD’s next top solution. 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 put more stress on the video chip. Perlin noise is a standard algorithm often used in procedural texturing and involves a lot of math. In a purely mathematical test from the Futuremark package, showing the peak performance of video chips in extreme tasks, we see a picture that is already familiar to us. The performance of the solutions shown in the diagram approximately corresponds to what should be obtained according to theory, and to what we saw earlier in our mathematical tests from the RightMark 2.0 package. Since the new HD 6870 and HD 6850 cards have seriously strengthened their position in mathematics, it is not surprising that the older model is the leader in the comparison, and the younger one is ahead of the previous mid-price board - HD 5770. Geforce video cards do not show very high results, losing to all boards AMD, which is completely consistent with the theory. After all, simple but intensive mathematics is performed much faster on Radeon video cards. Conclusions on synthetic tests Based on the results of synthetic tests of video cards from the new Radeon HD 6800 family based on the Barts GPU, as well as the results of other video card models produced by both manufacturers of discrete video chips, we conclude that this is a very suitable replacement for mid-price solutions on previous generation chips . Although the Barts GPU is not very different from previous chips architecturally, the number of execution units and their frequency have increased so much that the performance has come close to the top series of the previous generation - HD 5800. The new GPU also features some architectural improvements aimed at eliminating the one of the most important shortcomings compared to competitor products - and from synthetic tests we see that geometry processing performance has increased. Thanks to all the changes, the results of the new series of video cards in many synthetic tests are the maximum for solutions in this price sector. This is especially clearly seen in the parallelized, but not too complex in algorithm, computational tests from the RightMark and Vantage packages. And in all other applications, the speed of the HD 6800 is very good - noticeably higher than that of the corresponding solutions from the previous line. We can assume that the very good results of the Radeon HD 6870 and HD 6850 in our synthetic tests will be confirmed by similar results in the next part of our material, where you will get acquainted with the gaming tests from our set. Accordingly, in gaming tests the HD 6870 should outperform the HD 5830, and the HD 6850 will be faster than the HD 5770. But what will happen in comparison with Geforce video cards is not so easy to predict, since both of them have their own strengths and weak sides. It is likely that AMD's recently released solutions will excel in some games, while their competitors from Nvidia will prevail in others. It will be even more interesting to look at the results!

Based on data collected by Steam, multi-GPU graphics configurations are not popular among gaming enthusiasts, with less than one percent of gamers using them. The most interesting thing is that of this very small number of systems, 96% belong to NVIDIA SLI technology, and only 4% belong to ATI(AMD) CrossFireX technology. Of course, Steam only collects information on games and applications that are distributed through or support the service, and this may include many well-known and widely used games, including latest news, such statistics, as you understand, cannot claim to be absolute truth. But be that as it may, there is simply no other statistical data on the network.

At the same time, you don’t need to be a prophet in your own country to understand that multiprocessor graphics are still very, very poorly widespread. There are only a few owners of pairs of NVIDIA or ATI (AMD) video cards in hardware conferences, and graphics systems, consisting of three or even four video cards, you won’t find them at all during the day. The reason for this is not only the at least doubled cost of such bundles (we mean without taking into account more expensive motherboards and powerful power supplies), but also instability, as well as not always high efficiency of SLI and CrossFireX technologies, depending on drivers and graphics applications .

However, in the last year and a half there has been a rather optimistic trend towards improving the performance of multiprocessor technologies. First of all, this concerns ATI(AMD) CrossFireX, which until now has had much more problems with stability and efficiency than NVIDIA SLI. The latter, by the way, still has a unique ability to select the rendering mode in the drivers, due to which it is operational in more games. But, it should be noted that AMD does not stand still. Let's remember the recent introduction of CrossFireX profiles installed with Catalyst drivers and their regular updates. In addition, as today's tests will show, with the advent of AMD Radeon HD 6850 and Radeon HD 6870 video cards, changes have occurred at the hardware level...

How exactly all this affected the stability and efficiency of CrossFireX technology - we will find out today.

Technical characteristics of video cards

In the table of characteristics, as well as in the diagrams with test results below, video cards are arranged from left to right in descending order of recommended cost:

AMD Radeon HD 6850 and HD 6870 1 GB

Since almost all the information about new video cards has already been presented to your attention in the article “ ATI Radeon HD 6800: Next generation?", in this subsection of today's material we will fill in the missing points regarding the reference video cards AMD Radeon HD 6850 (pictured here and below - on the left) and AMD Radeon HD 6870 (hereinafter - on the right):

So, on the AMD Radeon HD 6850 the length of the PCB PCB is 229 mm, and the cooling system does not extend beyond the edge of the PCB. The thickness of the card is 36 mm. The GPU is shifted to the panel with the video card outputs, and the distance from the nearest mounting hole of the GPU radiator to this panel is 74 mm, which will certainly make impossible to install many alternative cooling systems:

In turn, on the AMD Radeon HD 6870 the PCB length is 13 mm longer and amounts to 242 mm. In addition, the cooling system extends 7 mm beyond the edge of the PCB. The thickness of the card is the same (36 mm). The GPU is also offset towards the output panel, but the distance from the nearest mounting hole to the output panel is 87 mm (that is, slightly further from the outputs than on the HD 6850). On both video cards, the diagonal distance between the GPU radiator mounting holes is 75 mm, the same as on the HD 5870 and HD 5850.

The area of ​​the crystals of the graphics processors of the Radeon HD 6870 and HD 6850 video cards is the same and amounts to 255 mm². Externally, they differ only in markings: they are dated 37 weeks of 2010 on the HD 6850 and 36 weeks on the HD 6870:

Let us recall that the graphics processor of the AMD Radeon HD 6850 video card has 160 unified shader processors and 8 texture units less than the HD 6870, and also operates at a 775 MHz clock frequency, versus 900 MHz for the HD 6870. In energy saving mode, the frequencies of the GPUs are reduced up to 100 MHz, and voltages - from 1.094V to 0.95V on the HD 6850, and from 1.172V to 0.95V on the HD 6870.

Both video cards have the same memory capacity of 1024 MB of GDDR5 standard manufactured by Hynix Semiconductor Inc. (marking H5GQ1H24AFR T2C) with a nominal voltage of 1.5V and a theoretical effective frequency of 5000 MHz:

All memory chips are located on the front side of the printed circuit boards. The width of the memory exchange bus is 256 bits. As you can already see from the characteristics table, the Radeon HD 6850 has memory operating at an effective frequency of 4000 MHz, and the Radeon HD 6870 - at 4200 MHz.

Thus, the specifications of the reference video cards are as follows:

Cooling systems for video cards are strikingly different:

The GPUs have small aluminum heatsinks with copper bases. The differences are that the Radeon HD 6850 has an evaporation chamber at the base of the radiator, and the Radeon HD 6870 has a copper plate with grooves for three heat pipes, two of which are 6 mm in diameter, and the diameter of one central pipe is 8 mm:

The turbines of the cooling systems of the AMD Radeon HD 6850 and HD 6870 reference video cards differ not only structurally (the turbine on the HD 6870 is wider), but also electrically. The HD 6850 is equipped with a FirstD turbine labeled FD9238M12D and a maximum power of 8.4 Watts. In turn, the Radeon HD 6870 has a turbine manufactured by NTK Ltd. with marking FD9238Р12S:

The rotation speed of the turbines is controlled automatically using pulse-width modulation. It should be added that the first turbine is mounted in a plastic casing of the cooling system, and the second in a metal frame. These differences could not but affect the noise level, which we will tell you about in one of the following sections of the article, but for now let’s check the temperature conditions of these reference video cards.

Temperature tests of all video cards in today's article were carried out in a closed system case at a room temperature of 24 °C. You will find the configuration of the system unit in the next section of the article. Used as a load FurMark test version 1.8.2, launched from a renamed exe file with the “Xtreme Burning Mode” option in a resolution of 2560x1600 with 16x anisotropic filtering activated in the Catalyst and GeForce/ION drivers. Monitoring was carried out using programs GPU-Z versions 0.4.7 and MSI Afterburner version 2.0.1 beta 1 (and later beta 2, 3). All checks were carried out before removing the reference cooling systems for video cards with a standard thermal interface on the graphics processor.

Here are the test results for the first two video cards tested: automatic mode turbine operation...

AMD Radeon HD 6850AMD Radeon HD 6870
(auto mode)(auto mode)

...and at maximum power:

AMD Radeon HD 6850AMD Radeon HD 6870
(turbine maximum speed)

In automatic turbine mode, the graphics processor of the Radeon HD 6850 video card warmed up to 93 degrees Celsius, and the power elements to 74 degrees. At the same time, the turbine rotation speed increased from 1000 to 2435 rpm. On the Radeon HD 6870, despite the higher-frequency graphics processor, its temperature did not exceed 87 degrees Celsius, and the power elements did not exceed 82 degrees at 2430 turbine rpm. The difference in the efficiency of the standard Radeon HD 6850 and HD 6870 coolers at maximum turbine speeds becomes even more obvious, when the temperature of the “Barts Pro”, even at a turbine speed of 4000 rpm, could not drop below 78 degrees Celsius, but the 900-MHz “Barts XT” "could only warm up to 62 degrees Celsius at 4420 rpm. However, all this is little consolation, since already at 1900 rpm the noise of both turbines begins to cause obvious discomfort, and at speeds from 2400 rpm it is already very high.

In CrossFireX modes, both pairs of video cards operate at even higher temperatures and at even higher temperatures. high speeds their turbines:

AMD Radeon HD 6850 CrossFireXAMD Radeon HD 6870 CrossFireX
(auto mode)(auto mode)

Due to the changed layout of the printed circuit boards of video cards and the shift of graphics processors to the outputs, the installation of alternative cooling systems turned out to be very, very problematic. For example, the new Thermalright Shaman cooler cannot be installed on either the reference Radeon HD 6850 or the reference Radeon HD 6870. Highly efficient cooler Arctic Cooling Accelero XTREME 5870 does not fit on the Radeon HD 6850, but fits perfectly on the Radeon HD 6870:

True, as we see, a good 50 mm of the length of its radiator simply turned out to be unnecessary. It’s quite possible to limit yourself to Arctic Cooling Twin Turbo Pro on such a card, especially since this cooler has already received official support new AMD Radeon. Well, with the Accelero XTREME 5870, the Radeon HD 6870 video card, even without radiators on the power elements and at a quiet 1100 rpm of three 92 mm cooler fans, turned into a card with a very modest temperature regime:

AMD Radeon HD 6870AMD Radeon HD 6870
(Accelero XTREME 5870 3x1100 rpm) (Accelero XTREME 5870 3x1970 rpm)

At the maximum speed of three fans, the GPU temperature reached only 57 degrees Celsius.

When testing the overclocking of video cards, both Radeon HD 6850 showed approximately the same results for the graphics processor, without any loss in stability and image quality, reaching 890 and 910 MHz at standard voltages. When voltages were increased, the frequency potential of video card GPUs was not tested, since it was impossible to provide them with good cooling. As for overclocking the video memory of the Radeon HD 6850, the first video card was limited to a modest 4520 MHz, but the second was more fortunate - the maximum frequency of its video memory was 4880 MHz. The result of overclocking the best of the two video cards is shown in the following screenshot:

Overclocking of two Radeon HD 6870 with standard cooling systems and at the nominal voltage of the GPUs turned out to be different: on the first video card the core overclocked better, and on the second - the video memory:

At the end of the subsection, we provide links to the BIOS of the video cards reviewed: AMD Radeon HD 6850 1 GB And AMD Radeon HD 6870 1 GB.

Palit GeForce GTX 460 Sonic Platinum 1 GB

In addition to new AMD video cards, two serial video cards will take part in today's testing Palit GeForce GTX 460 Sonic Platinum 1 GB, which are delivered in small but very brightly decorated boxes:

On the front and back sides of the packages you can find all the comprehensive information about the video card. Inside, in the central compartment, there is the video card itself, and next to it are the components:

Frankly speaking, the set of accessories supplied with video cards is very poor. It contains only a cable for connecting additional power, a CD with drivers and brief instructions on installation. It’s strange what prevented Palit from equipping its video cards with SLI bridges, adapters and some old game? The savings are logical, but, in the case of the Palit GeForce GTX 460 Sonic Platinum, they are not justified, in our opinion.

The first impression of the Palit GeForce GTX 460 Sonic Platinum, after studying the Radeon HD 6850 and HD 6870, is made by its dimensions, namely a length of only 188 mm. And even though the card also has a dual-slot cooling system design, compared to the new “Barts” it looks somehow childish:

Despite this, the “baby” is equipped with one analog, one DVI-I, one DVI-D and one HDTV outputs:

That is, almost a complete set, with the exception of DisplayPort. The remaining space on the video card panel from the connectors is occupied by two grilles for partial exhaust of air heated by the video card from the system unit case.

The casing of the video card cooling system, in which the fan impeller is installed, is attached to the printed circuit board separately from the graphics processor radiator and the radiator on the power elements:

As you can see, the first one has a copper base in contact with the GPU heat spreader through a thick layer of unevenly applied thermal paste, two copper heat pipes with a diameter of 6 mm and thin aluminum plates. The second, also made of aluminum, is a simple comb, painted black and equipped with a thermal pad.

Palit GeForce GTX 460 Sonic Platinum is made according to its own unique design and has a four-phase power supply:

Two six-pin connectors for connecting additional power are directed upwards, so the video card in practice turns out to be very short and easy to use. But, we would like to immediately draw your attention to the proximity of the graphics processor to the panel with outputs - because of this feature, installing alternative cooling systems on this video card will be a big question.

The GF104 graphics processor, produced in Taiwan using a 40 nm process technology, is covered with a heat spreader, which is marked with the marking and week of release (week 23 of 2010):

The frequency formula of the chip is 800/1600 MHz, which is 18.5% higher than the reference GeForce GTX 460. Not a bad factory overclock, it should be noted. All other characteristics of the GPU remained identical to the reference ones.

Palit GeForce GTX 460 Sonic Platinum is equipped with 1 GB of GDDR5 video memory installed on the front side of the printed circuit board. The chips were released by the Samsung Semiconductor division and are marked K4G10325FE-HC05:

The nominal access time of the memory chips is 5 ns, and the theoretical effective operating frequency is 4000 MHz. It is at this frequency that the memory of the Palit GeForce GTX 460 Sonic Platinum video card operates (+11%). Regular GeForce GTX 460 have a memory frequency of 3600 MHz. The width of the exchange bus with the video card memory is 256 bits.

Thus, we can say that the Palit video cards reviewed today are among the fastest commercially available GeForce GTX 460:

Only the video card has higher frequencies Gigabyte GeForce GTX 460 GV-N460SO-1GI- 815/4000 MHz. Similar video cards from Zotac and Leadtek have the same increased frequencies as the Palit GeForce GTX 460 Sonic Platinum.

One 92-mm fan manufactured by Power Logic is responsible for cooling the radiators of the video card:

The rotation speed is adjusted automatically using pulse width modulation in the range from 1200 to 3900 rpm. Let's see how the Palit GeForce GTX 460 Sonic Platinum cooling system copes with cooling an overclocked video card:

Auto modeMaximum power

In automatic fan mode, the GPU temperature reached 83 degrees Celsius, and at maximum fan power it reached 80 degrees. It seems that, taking into account the increased frequencies at the factory, this is a completely normal temperature regime, but note that even in the first case, the fan spun up to 3480 rpm, which is very noisy, not to mention the maximum speed mode. It is clear that, like FurMark, not a single application or game warms up the video card, but even in games the fan rotation speed reached almost 2900 rpm, which cannot be called quiet. Looking ahead, we note that the high noise level of the cooling system is the only drawback of the Palit GeForce GTX 460 Sonic Platinum.

As for overclocking video cards, both of them, unlike the AMD reference products discussed above, demonstrated almost synchronous frequencies, overclocking the core to 830(840)/1660 MHz, and the video memory to 4220 MHz:

This is quite a bit compared to the standard frequencies of the Palit GeForce GTX 460 Sonic Platinum, but quite decently compared to the standard frequencies of the reference GeForce GTX 460.

It remains to check what the temperature conditions of the video cards will be when combining them into an SLI tandem:

Unfortunately, CrossFireX and SLI configurations, in terms of noise levels, remain the domain of liquid cooling systems, since with such close proximity of video cards, it is simply impossible to organize effective heat removal with air cooling systems while maintaining a low noise level.

Let's add a link to BIOS Palit GeForce GTX 460 Sonic Platinum and the fact that the cost of such a video card is less than 250 US dollars.

Test configuration, tools and testing methodology

All performance tests of video cards were carried out in a closed system case with the following configuration:

Motherboard: ASUS P6T Deluxe (Intel X58 Express, LGA 1366, BIOS 2101);
CPU: Intel Core i7 Extreme Edition i7-980X 3.33 GHz(Gulftown, B1, 1.225 V, 6x256 KB L2, 12 MB L3);
Thermalright Silver Arrow (one Thermalright TY-140 700-1280 rpm PWM);
Thermal interface: Arctic Cooling MX-2;
RAM: DDR3 3x2 GB OCZ Platinum Low-Voltage Triple Channel (1600 MHz / 7-7-7-24 / 1.65 V);
Sound card: Auzen X-Fi HomeTheater HD;
System disk: RAID-0 2xSSD Kingston V-series SNV425S2128GB(SATA-II, 128 GB, MLC, Toshiba TC58NCF618G3T);
Disk for programs and game distributions: Western Digital VelociRaptor WD3000HLFS (SATA-II, 300 GB, 10000 rpm, 16 MB, NCQ) in a Scythe Quiet Drive 3.5" box;
Archive disk: Western Digital Caviar Green WD10EADS (SATA-II, 1000 GB, 5400 rpm, 32 MB, NCQ);
Case: Antec Twelve Hundred (front wall - three Noiseblocker NB-Multiframe S-Series MF12-S2 at 840 rpm; rear - two Thermalright X-Silent 120 at 840 rpm; top - standard 200 mm fan at 400 rpm /min);
Control and monitoring panel: Zalman ZM-MFC2;
Power supply: Zalman ZM1000-HP 1000 W, 140 mm fan.
Monitor: 30" Samsung 305T Plus.

The 32nm six-core processor was overclocked with a multiplier of 24 and the Load-Line Calibration function activated to 4.512 GHz with a voltage increase of Motherboard BIOS boards up to 1.475 V:

Moreover, 6 GB random access memory DDR-3 operated at a frequency of 1.5 GHz with timings of 7-7-7-14_1T at a voltage of 1.64 V. Turbo Boost and Hyper-Threading technologies were disabled during testing.

Testing, which began on October 29, 2010, was carried out under the operating system Microsoft Windows 7 Ultimate x64 with all critical updates as of the specified date, with the installation of the following drivers:

chipset motherboard Intel boards Chipset Drivers - 9.1.2.1008 WHQL ;
DirectX End-User Runtimes libraries, released June 2010;
drivers for video cards on ATI Catalyst 10.10c (26.10.2010) graphics processors with CrossFireX profiles;
video card drivers for GPUs NVIDIA GeForce/ION 260.99 WHQL(25.10.2010) including PhysX drivers version 9.10.0514.

Testing of video cards in games was carried out in two resolutions: 1920x1080 and 2560x1600. In our opinion, at the current cost of monitors with a screen resolution of 1920x1080 in the area 150-170 US dollars testing at lower resolutions is gradually losing its relevance.

For the tests, two graphics quality modes were used: “High Quality + AF16x” - maximum texture quality in the drivers with 16x level anisotropic filtering enabled, and “High Quality + AF16x + AA 4(8)x” with 16x level anisotropic filtering enabled and full screen anti-aliasing (MSAA) of 4x or 8x, if the average number of frames per second remained high enough for comfortable gameplay. Anisotropic filtering and full-screen anti-aliasing were enabled directly in the game settings. If these settings were not available in games, then the parameters were changed in the control panels Catalyst drivers and GeForce/ION. Vertical sync is forcibly disabled in driver control panels.

Following tradition, the list of test applications and games has again been supplemented and updated. In addition to updating games with the latest patches, three new games are included in the tests: Civilization V, F1 2010 and NBA 2K11. In addition, the newest campaign in the game Left 4 Dead 2: “The Sacrifice” has been added. As a result, the test list consisted of two semi-synthetic packages, one techno demo and 19 games. Here's what it looks like with a brief description of the techniques (hereinafter the games are arranged in the order of their release):

3DMark 2006(DirectX 9/10) - build 1.2.0, default settings and 1920x1080 with AF16x and AA8x;
3DMark Vantage(DirectX 10) - version 1.0.2.1, “Performance” and “Extreme” settings profiles (only basic tests were tested);
Unigine Heaven Demo(DirectX 11) - version 2.1, maximum quality settings, tessellation at the “extreme” level;
Crysis(DirectX 10) - version 1.2.1, “Very High” settings profile, double cycle of the “Assault Harbor” demo recording from Crysis Benchmark Tool version 1.0.0.5;
Far Cry 2(DirectX 10) - version 1.03, “Ultra High” settings profile, double test cycle “Ranch Small” from Far Cry 2 Benchmark Tool v1.0.0.1;
BattleForge: Lost Souls(DirectX 11) - version 1.2 (09/14/2010), maximum graphics quality settings, shadows enabled, SSAO technology enabled, double run of the test built into the game;
Resident Evil 5(DirectX 10.1) - version 1.2, testing the variable test with maximum graphics settings without motion blur, the result was taken as the average value of the third scene of the test, as the most resource-intensive;
(DirectX 11) - version 1.6.02, settings profile “Improved dynamic lighting DX11” with additional manual setting of all parameters to the maximum, tested our own demo recording “cop03” at the “Backwater” level;
Borderlands(DirectX 9) - game version 1.2.1, testing “timedemo1_p” with maximum quality settings;
Grand Theft Auto IV - Episodes From Liberty City(DirectX 9) - version 1.1.2.0, test from the part “The Ballad of Gay Tony”, settings “Very High”, “View Distance” = 23%;
Left 4 Dead 2: The Sacrifice(DirectX 9) - game version 2.0.4.5, maximum quality, tested our own demo recording “d45” (two cycles) on the map “1. Docks”, stage “Victim”;
Colin McRae: DiRT 2(DirectX 9/11) - game version 1.2, built-in test, consisting of two laps along the London circuit with maximum graphics quality settings;
Metro 2033: The Last Refuge(DirectX 10/11) - version 1.2, official test was used, quality settings “High”, tessellation, DOF and MSAA4x disabled, AAA anti-aliasing was used, double sequential pass of the “Frontline” scene;
Just Cause 2(DirectX 11) - version 1.0.0.2, maximum quality settings, “Background Blur” and GPU Water Simulation techniques activated, double sequential pass of the “Dark Tower” demo;
Aliens vs. Predator (2010)(DirectX 11) - “Texture Quality” Very High, “Shadow Quality” High, SSAO On, two test cycles in each resolution;
Lost Planet 2(DirectX 11) - game version 1.0, maximum graphics quality settings, motion blur enabled, performance test “A” was used (average of all three scenes);
StarCraft 2: Wings of Liberty(DirectX 9) - game version 1.0, all graphics settings at the “Ultra” level, “Ultra” physics, reflections included, double two-minute test of our own demo “jt1”;
Mafia 2(DirectX 11) - game version 1.0.0.1, maximum graphics quality settings, double run of the test built into the game;
Sid Meier's Civilization V(DirectX 11) - game version 1.0, maximum graphics quality settings, double run of the “diplomatic” test of the five most difficult scenes;
F1 2010(DirectX 11) - game version 1.01, built-in Ultra-quality test, consisting of one lap along the “Silverstone” track;
NBA 2K11(DirectX 11) - game version 1.0, built-in test at maximum graphics quality settings, one run;
Tom Clancy's H.A.W.X. 2(DirectX 11) - version 1.04, maximum graphics quality settings, shadows activated, tessellation enabled, double run of the test scene.

More detailed description You can find methods for testing video cards and graphic settings in some of the listed games in a specially created thread of our conference, as well as participate in the discussion and improvement of these techniques.

If games implemented the ability to record a minimum number of frames per second, then this was also reflected in the diagrams. Each test was carried out twice; the best of the two values ​​obtained was taken as the final result, but only if the difference between them did not exceed 1%. If the deviations of the test runs exceeded 1%, then the testing was repeated at least once more to obtain the correct result.

Video card performance test results and their analysis

As already mentioned in the introduction of today's article, through comprehensive and extensive testing we will try to find out how effectively CrossFireX technology works on the new AMD Radeon HD 6870 and HD 6850 video cards in comparison with NVIDIA SLI technology using the example of two GeForce GTX 460 video cards. Based on positioning of mid-class video cards by the GPU manufacturers themselves, it would be logical in today’s article to also see an SLI tandem from a pair of GeForce GTX 470, but since I didn’t have such cards at my disposal, we will try to make up for their absence with a good overclocking of the GeForce GTX 460 with frequencies 675/1350/3600 MHz to 830/1660/4220 MHz. Since overclocking within pairs of Radeon HD 6870 and HD 6850 video cards turned out to be significantly different, these video cards were tested in CrossFireX mode only in nominal mode. However, in our opinion, in addition to all of the above, it will be interesting to compare the Radeon HD 6850 at frequencies equal to the HD 6870 - with the HD 6870 itself. Thanks to this, we will be able to evaluate how much the hardware disabled 160 unified shader processors and 8 texture blocks.

In the diagrams, the test results for AMD Radeon HD 6870 1 GB video cards are highlighted in purple, AMD Radeon HD 6850 video cards and CrossFireX configurations of them are marked in red, and GeForce GTX 460 in green (Palit card frequencies are referenced). Video cards and CrossFireX bundles in the diagrams are arranged in descending order of their retail cost. Go.

3DMark 2006

Since the 3DMark 2006 test package on modern mid- and top-class video cards is very dependent on the speed of the platform, the diagram shows the results of the least processor-dependent test “HDR/SM3.0”, and not the total number of “parrots” (which you can still find in the table at the end of this section of the article):

Even despite overclocking the six-core Intel Core i7 processor to 4.5 GHz, with the default 3DMark 2006 settings, CrossFireX and SLI configurations were not able to reveal their full potential. Only when testing at a resolution of 1920x1080 using anisotropic filtering and full-screen anti-aliasing does the difference in the performance of video cards become clearly visible. If we talk about tests in single mode, the lead belongs to the Radeon HD 6870, which is quite predictable. However, the Radeon HD 6850, overclocked to the frequencies of the HD 6870, lags behind the leader by only 2%, although at nominal frequencies it loses 11%. The slowest is the GeForce GTX 460.

Both multiprocessor technologies demonstrate very high efficiency in this semi-synthetic test. So, for a pair of Radeon HD 6870 the performance increase was 93%, for the Radeon HD 6850 - 94%, and a combination of two GeForce GTX 460 works as much as 98% faster than one such video card. And these are also “flowers”! In addition, we can add here that well-overclocked GeForce GTX 460 are still inferior in performance to a pair of Radeon HD 6870 operating in nominal mode.

3DMark Vantage

In 3DMark Vantage, the diagram shows the results of the “GPU” test, as the least dependent on the speed of the platform:

The more resource-intensive 3DMark Vantage test allows all video cards and tandems to shine in the “Performance” settings profile. Here the GeForce GTX 460 feels more confident compared to its competitors, and the efficiency of CrossFireX and SLI exceeds 90%, which is also a very good result. The gap between the Radeon HD 6850 and the HD 6870 is greater here: at nominal frequencies it is 24-25%, and when overclocked it decreases to 9-11%.

Unigine Heaven Demo

The first thing that should be noted from the results of testing video cards in the Unigine Heaven Demo is the impressive efficiency of CrossFireX technology. For example, the most “modest” performance increase for tandems from Radeon HD 6870 and HD 6850 is 95%, and in one case even exceeds 100%! Against this background, 83% and 89% on SLI from the GeForce GTX 460 do not look so impressive, although the performance increase itself is very good. In addition to this, the GeForce GTX 460 manages to lead in this test, and overclocking video cards in SLI mode puts these video cards beyond the reach of today's competitors. The overclocked Radeon HD 6850 is inferior to the nominal Radeon HD 6870 by about 4-5%, which is quite a bit, considering the difference in the cost of these video cards.

Crysis

Not impressed with CrossFireX's performance in synthetic tests? Here you go - Crysis with a 100% increase in performance on both combinations of AMD video cards, compared to single video cards. GeForce GTX 460 SLI also works great here, but still its 83-89% is not as surprising as 100%, and even slightly higher, on the Radeon HD 6870 and HD 6850. Only in the heaviest graphic mode and at a resolution of 2560x1600, all multiprocessor tandems moderated their ardor, limiting the increase to 80% on AMD and only 24% on NVIDIA. In general, the GeForce GTX 460 is slower than its competitors, and the overclocked Radeon HD 6850, lagging behind the HD 6870 in nominal operating mode by 18-21%, fully compensates for this lag by overclocking to frequencies of 900/4200 MHz.

Far Cry 2

The Radeon HD 6870 and HD 6850 pairs also shoot at 100% and above in the game Far Cry 2, but this does not in the least bother the GeForce GTX 460 SLI with its 83-93% performance, since these video cards fight with freshly baked competitors on an equal footing ( especially during overclocking and in anti-aliasing mode). The overclocked Radeon HD 6850 reduces the gap to the Radeon HD 6870 to a minimum.

BattleForge: Lost Souls

And in the third game - BattleForge: Lost Souls - both pairs of Radeons demonstrate impressive CrossFireX performance. GeForce GTX 460 SLI work less efficiently, but are seriously inferior in performance only in mode without anti-aliasing. When MSAA8x is enabled, the results are very close, although AMD is still slightly ahead.

Resident Evil 5

Despite the fact that out of the four test scenes of Resident Evil 5, the most resource-intensive third scene was chosen, the efficiency of the pair Radeon HD 6870 (84-95%) turned out to be limited by the performance of the platform and, in particular, the central processor, since compared to them two are slower and as a result, the less speed-sensitive Radeon HD 6850 platforms still boast a 97-101% increase in average frames per second compared to a single graphics card. Two GeForce GTX 460 in SLI mode also work excellently in this test - the performance increase ranges from 90 to 95%. Overclocked to the frequencies of the older video card, the Radeon HD 6850 lags behind it by only 2-6%.

S.T.A.L.K.E.R.: Call of Pripyat

In the game S.T.A.L.K.E.R.: Call of Pripyat the situation is very interesting. Both pairs of AMD Radeon HD 68xx video cards were able to demonstrate only 56-68% of performance here compared to a single video card, being completely inferior in terms of operating efficiency to the GeForce GTX 460 pair with its 96-97% of performance. Probably the reason for such an unconvincing performance of the new Radeons is in the drivers, since in this game ATI/AMD video cards always looked no worse than their competitors, including in multi-processor configurations. Looking ahead, we note that this is not the only such case in today’s article.

Borderlands

But in the Borderlands game, CrossFireX technology did not work very effectively before, so in today’s tests we can’t expect miracles from it, which was confirmed (40-50%). In SLI, on the contrary, the performance increase reaches 93%, which allows the GeForce GTX 460 pair to win a convincing victory. The overclocked Radeon HD 6850 lags behind the Radeon HD 6870 running at nominal frequencies by only 3-5%.

Grand Theft Auto IV: Episodes From Liberty City

Grand Theft Auto IV: Episodes From Liberty City is the third testing game in which CrossFireX does not provide a 95-100% performance increase. However, this does not prevent AMD from outperforming NVIDIA in this game, and overclocking a pair of GeForce GTX 460s does not correct this situation in NVIDIA's favor.

Left 4 Dead 2: The Sacrifice

Surprisingly, in the game Left 4 Dead 2, or rather in its most recent part - “The Sacrifice”, CrossFireX technology turned out to be ineffective, which led to a complete fiasco of both pairs of AMD video cards. At the same time, in single modes, even the younger Radeon HD 6850 turns out to be faster than the more expensive GeForce GTX 460, not to mention the HD 6870. By the way, the advantage of the latter card over the Radeon HD 6850 overclocked to its frequencies is 7-10%.

Colin McRae: DiRT 2

After four games in a row that were unsuccessful for CrossFireX, the situation for the Radeon HD 6870 and HD 6850 pairs is starting to improve. The performance increase is about 90%, as is the case with the GeForce GTX 460 SLI. The performance of the latest video card and a combination of two such video cards, on average, is between the HD 6850 and HD 6870, and when overclocked, it is ahead of them. The gap between the Radeon HD 6850 and the HD 6870 after overclocking decreases from 15-30% to 6-8%.

Metro 2033: The Last Refuge

In this game, the efficiency of CrossFireX technology is also high - the performance increase is over 90%. But, in general, the test results of all video cards and multiprocessor tandems leave much to be desired. Still, Metro 2033: The Last Refuge requires even more powerful video cards or more modest ones graphic settings. Overclocking the Radeon HD 6850 allows the video card's performance to reach the level of the Radeon HD 6870. The same thing happens when overclocking the GeForce GTX 460 SLI against the nominal HD 6870 CrossFireX.

Just Cause 2

Just Cause 2 brings us back to the amazing performance gains of CrossFireX. Even if you find the minimum increase, it is equal to 95(!)%, and on average it does not fall below 97%. At the same time, on a pair of GeForce GTX 460, SLI operating efficiency varies from 88 to 94%, which, however, is also very good. Overall, the Radeon is faster than the GeForce in Just Cause 2.

Aliens vs. Predator (2010)

Alien vs. Predator follows the previous game and demonstrates a performance increase of 92 to 100% in Radeon CrossFireX mode compared to a single graphics card, with 86-92% on GeForce. At nominal frequencies, the performance of the GeForce GTX 460 is comparable to the Radeon HD 6850 (a little slower in places), and the Radeon HD 6870 is slightly ahead of them. In anti-aliasing modes, play Aliens vs. Predator is only available in resolutions up to 1920x1080, and then only on SLI or CrossFireX tandems from the video cards tested today.

Lost Planet 2

Are you still bored with the 100% efficiency of CrossFireX? Then here are similar results in the test of the game Lost Planet 2 :) SLI technology also works very confidently here (about 90%), but still the performance gains are lower than in the case of CrossFireX. After overclocking a pair of GeForce GTX 460, this combination is ahead of two Radeon HD 6870 operating in nominal mode.

StarCraft 2: Wings of Liberty

The maximum performance of video cards in the game StarCraft 2: Wings of Liberty is limited to 60 frames per second, so the Radeon HD 6850 and HD 6870 have nowhere to turn. But in anti-aliasing mode, the performance of these video cards is limited not by this parameter, but by the game engine itself, which is better suited for NVIDIA products. AMD knows about this and has already expressed its official opinion online about the tests in StarCraft 2: Wings of Liberty. We note that CrossFireX, apparently, does not work here, or works somehow selectively (in other words, in places).

Mafia 2

In the game Mafia 2, pairs of Radeon HD 6870 and HD 6850 demonstrate a performance increase of only 48 to 77%, and if previously this could have been considered a good achievement, then against the backdrop of the results obtained in today’s article, ATI (and now AMD) programmers have something to work on work. A pair of GeForce GTX 460 is ready to offer Mafia 2 players a performance increase of 73 to 86%, which is quite typical for NVIDIA SLI.

Sid Meier's Civilization V

But tests in the game Sid Meier's Civilization V return everything to its place - the performance of Radeon HD 6870 and HD 6850 video cards in CrossFireX mode is equal to the sum of the performance of two single video cards. NVIDIA SLI on two GeForce GTX 460 also works well, but is ready to offer players Civilization V is "only" 90% to one video card.

F1 2010

As a rule, in new games, multiprocessor technologies either do not work or work very inefficiently. One of these games is F1 2010, which was released recently, but even in it the efficiency of CrossFireX reaches 97% compared to a single video card, although this is more of a one-time result than a pattern, since in other modes the increase decreases to as much as 43%. In turn, the GeForce GTX 460 SLI will allow fans of Formula 1 and this game to increase performance compared to a single video card by 58-70%.

NBA 2K11

In the also new game NBA 2K11, everything is very simple - multi-processor configurations do not work here. Neither AMD CrossFireX nor NVIDIA SLI. That is, they don't work at all. Absolutely. In principle, it’s not necessary, because the average number of frames per second remains quite high even with maximum resolution and anti-aliasing mode. Simply put, fans of Michael Jordan, when choosing the number of video cards for a system unit, can be guided by the rule - no more than one ball on the court.

Tom Clancy's H.A.W.X. 2

In the new Tom Clancy's H.A.W.X. 2 test, two pairs of AMD Radeon video cards in CrossFireX mode finally achieved another (by the way, the tenth) 100% increase in performance compared to a single video card. NVIDIA SLI cannot boast of such impressive gains, however, even without this, the GeForce GTX 460 easily holds the lead in this test in both single and dual modes.Overclocking the Radeon HD 6850 to the frequencies of the Radeon HD 6870 reduces the gap from 13-17% to 3-5%.

In this subsection, all that remains is to attach the promised one, and you can move on to the summary diagrams.

Performance comparison summary charts

Using the first two pairs of summary diagrams, let's try to compare the performance efficiency of AMD CrossFireX and NVIDIA SLI technologies (as a percentage of the performance of the corresponding single video cards):

In many games and tests, CrossFireX on the new AMD Radeon HD 6870 and HD 6850 video cards works more efficiently than the SLI of the GeForce GTX 460 pair. In ten games, the performance increase of CrossFireX is at 100%, which in itself is an extraordinary result. At the same time, in games such as S.T.A.L.K.E.R.: Call of Pripyat, Borderlands, Left 4 Dead 2: The Sacrifice, StarCraft 2: Wings of Liberty and Mafia 2, the advantage is on the side of NVIDIA, and in the game NBA 2K11 both technologies do not work.

The following diagrams show the advantage of the Radeon HD 6870 over the Radeon HD 6850 in nominal operating mode and when overclocking the latter video card to HD 6870 frequencies (900/4200 MHz):

On average, for all games and tests, at nominal frequencies the Radeon HD 6850 loses to its “big sister” by 15-20%, but overclocking this video card to the standard frequencies of the Radeon HD 6870 reduces this gap to 3-6%, and in some games, resolutions and quality modes - the video cards turn out to be completely equal.

Finally, with the following diagrams we will try to compare the performance of a pair of AMD Radeon HD 6870 in CrossFireX mode at nominal frequencies with a pair of NVIDIA GeForce GTX 460 in SLI mode when overclocked to frequencies of 830/4220 MHz (a kind of attempt to emulate the GeForce GTX 470 SLI). The performance of the GeForce GTX 460 SLI is taken as the zero axis, and the performance of the Radeon HD 6870 CrossFireX is shown as a percentage deviation from it:

As we see, the struggle is being waged with varying degrees of success. Naturally, in those games where the efficiency of CrossFireX is still far from ideal, a pair of Radeon HD 6870 fails, but in others, as a rule, it turns out to be faster. However, you can see everything yourself in the diagrams.

Energy consumption and noise level

energy consumption measurement

The energy consumption of systems with different video cards was carried out using specially modified for these purposes the power supply. The maximum load was created by running one FurMark version 1.8.2 in stability test mode and a resolution of 2560x1600 (with AF16x), as well as FurMark together with Linpack x64 (LinX 0.6.4, 4750 MB, 5 threads). Considering that both specified programs generate maximum load to, respectively, the video system and CPU, in this way we can find out the peak power consumption of the entire system and determine the power supply required for it (taking into account efficiency).

The results obtained are shown in the diagram:

As we found out, the new Radeon HD 6850 and HD 6870 video cards are more energy efficient than the GeForce GTX 460. The consumption of systems with such video cards is lower both in single modes and in multiprocessor configurations. The difference is not critical, but, nevertheless, not in favor of NVIDIA. Separately, I would like to note a significant reduction in the consumption of systems with all video cards without exception in idle mode. Analyzing the results obtained in the diagram, we can say quite accurately that in idle mode one video card consumes only about 20 W of electricity.

noise level measurement

The noise level of video card cooling systems was measured using an electronic sound level meter CENTER-321 after one in the morning in a completely closed room of about 20 m² with double-glazed windows. The noise level of each cooler was measured outside the system case, when the only source of noise in the room was the cooler itself and its fan(s). The sound level meter, fixed on a tripod, was always located strictly at one point at a distance of exactly 150 mm from the fan rotor/cooler turbine. The motherboard, into which a video card was inserted with a cooling system installed on it, was placed at the very corner of the table on a polyurethane foam backing. The lower measurement limit of the sound level meter is 29.8 dBA, and the subjectively comfortable (not to be confused with low) noise level of coolers when measured from such a distance is around 36 dBA. The rotation speed of the cooler fan(s) was changed over the entire range of their operation using the controller by changing the supply voltage in steps of 0.5 V.

Based on the results of measuring the noise level of video cards from today's testing, the following graph was constructed:

Unfortunately, none of the video cards tested today can be called quiet, or at least comfortable in terms of noise level. They all make noise when running a 3D application. But most of all I didn’t like the sound of the reference cooler for Radeon HD 6850 video cards - it was harsh, with an unpleasant plastic sound, it was tolerable for no more than 5-10 minutes. The Radeon HD 6870 also disappointed in terms of noise level, as it turned out to be louder than the already noisy reference Radeon HD 5870 or HD 5830. Judging by the graphics, the Palit GeForce GTX 460 Sonic Platinum is the quietest video card tested, but in fact the term is “quiet” "is simply not applicable to her. So, if you like any of the video cards in today's testing, and the noise level for your computer is not the least important characteristic, then be prepared to replace the cooling system.

Conclusion

The main conclusion of today's testing is this: on the new AMD Radeon HD 6870 and HD 6850, CrossFireX works more efficiently than on the previous line of Cypress GPUs and video cards based on them. It's no joke, but in ten out of nineteen games in today's testing, the performance increase from adding a second video card to the system is at 100%, which was previously achievable only in one or two games on selected engines optimized for the ATI (AMD) architecture. Thus, from the point of view of arithmetic, it is enough to simply add the performance of two video cards to find out the final performance of the CrossFireX combination of them. At the same time, you should not think that CrossFireX technology has become simply flawless - it is not. And in confirmation of this, there are four or five games of today's testing, in which CrossFireX either does not work effectively enough or does not work at all. However, in general, the performance gain with this technology is currently higher than with SLI from NVIDIA. In addition, you need to take into account the newness of the Radeon HD 6870 and HD 6850 video cards, and the fact that drivers for them will still be optimized and optimized for many months. For example, I personally have no doubt that in Left 4 Dead 2 the bug with CrossFireX not working will be fixed in the near future.

As for comparing the performance of AMD CrossFireX from a pair of Radeon HD 6870 and HD 6850 with NVIDIA SLI from two GeForce GTX 460, this combination of video cards in terms of performance is right between the Radeon HD 6870 and HD 6850, and in some games even outperforms them, for account of more mature drivers and the ability to manually select the SLI rendering mode in the driver control panel. That is, in fact, all three pairs of video cards ranked in terms of performance in accordance with their recommended cost. When overclocked, the GeForce GTX 460 SLI 1 GB becomes a very formidable weapon in the fight not only with the Radeon HD 6850, but also with the older HD 6870. However, to complete the picture, it should be added that the power consumption of today's NVIDIA representatives is slightly higher than that of systems with AMD video cards.

In conclusion, we can say about the only serial video cards in today's material - Palit GeForce GTX 460 Sonic Platinum. These video cards, in our opinion, have two drawbacks: very poor equipment and a high noise level in 3D mode. And if you can put up with the first drawback, or not consider it such at all, then the second one cannot be ignored. The compactness of the graphics card's printed circuit board led to the fact that developers simply did not have enough space to accommodate a larger GPU heatsink and install, for example, two fans instead of one, which would allow maintaining high cooling efficiency with a significantly lower noise level. Alas. Otherwise, Palit GeForce GTX 460 Sonic Platinum video cards are very good and attractive not only with significantly increased frequencies, but also with Palit’s affordable pricing policy. The choice, as always, is yours.

Thank you:
Russian representative office of AMD and personally Kirill Kochetkov,
Palit Microsystems Ltd. and personally Marina Pelepets
for video cards provided for testing.

Other materials on this topic

AMD Radeon HD 6800: Next generation?AMD Radeon HD 6800: Next generation?
"Goodbye ATI!" - the latest test of all gaming video cards Radeon HD 5xxx
Asus ENGTX460 DirectCU TOP: time for maturity

“...Radeon HD 6850 turned out to be surprisingly fast. It lags behind the older version by 20% and is confidently on par with the GTX 460 and GTX 465. But the whole impression is spoiled by the price in Russia. If in the West an AMD board will cost significantly less than its competitors, then here it is

Gambling addiction https://www.site/ https://www.site/

When AMD presented a line Radeon HD 6800, we are almost ready to print “Gaming” No. 12/2010 - and test both video cards, HD 6870 And HD 6850, we didn’t have time. Therefore, in the last issue we studied only the older version, and in this issue we talk about the younger one.

Scalpel

Like the older model, the HD 6850 is based on a modified version of the crystal Cypress - Barts. There are even fewer SIMD blocks left on it than on the HD 6870: 12 instead of 16. Accordingly, the number of stream processors has also decreased - from 1120 to 960 units. There are 8 fewer texture blocks - 48 pieces. Decreased and clock frequency processor: instead of 900 MHz, the crystal operates at 775 MHz. But the memory was almost untouched. The HD 6850 has an adult 256-bit bus and 1 GB of GDDR5, which is only 200 MHz slower than the HD 6870. The same applies to rasterization units: all 32 of them are in place.

There are no differences in the architecture of the junior and senior versions. The Radeon HD 6850 supports HD3D stereo imaging, Morphological AA filtering, advanced anisotropic filtering and anti-aliasing algorithms, and is also equipped with a video decoding module UVD3 supporting VC-1, H.264, MPEG-2 (DVD), MVC and MPEG-4 (DivX, Xvid) formats.

The card itself is 2 cm shorter than the HD 6870, but they have the same design - black bricks with red inserts. The cooling system is exactly the same: heat pipes, radiator, turbine. The only thing is that the number of power contacts has been reduced. The HD 6850 consumes 127 W, and it only needs one 6-pin plug. Video outputs are the same as on the HD 6870: two mini-DisplayPort 1.2, two DVI and one HDMI 1.4a. All this allows you to connect up to six monitors to one video card at once, watch Blu-ray 3D and display the image on TVs that support 3D Stereo.

Option

The company sent the Radeon HD 6850 for testing Sapphire. The card turned out to be non-standard - the developers changed the cooler. The branded plastic casing is somewhat reminiscent of a spaceship: glossy black color, silver inserts, intricate stamping. An aluminum radiator rests on the processor through two copper heat-conducting tubes, and there is a 75 mm fan on top.

The set of video outputs has also changed: one full-size DisplayPort 1.2, HDMI 1.4a and two DVI. Otherwise, nothing new, processor and memory frequencies are standard, no overclocking.

Price issue

To test the Radeon HD 6850, we took motherboard Foxconn Renaissance on a chipset Intel X58 Express, put a processor on it Intel Core i7-920, added three memory sticks Kingston HyperX DDR3-1666 2 GB each and installed operating system Windows 7 Ultimate64-bit. We chose a standard set of tests: we looked at how many points the card scores in synthetics 3DMark Vantage And Unigine Heaven Benchmark 2.0, measured the performance speed in DX10 and DX11 games.

With the selection of competitors for the HD 6850, an overlay came out. The official price of the card is 5500-6000 rubles. But our great retail has inflated it to 7300-8000 rubles. So it turns out that in Russia the video card does not have to compete with any GTX 460 768 MB and GTS 450, but with serious comrades like GTX 460 1 GB, GTX 465 And HD 5830 With HD 5850.

The amazing is nearby

The first test, 3DMark Vantage, put the Radeon HD 6850 in last place, even the weak HD 5830 was 3% faster. Unigine Heaven Benchmark 2.0 turned out to be a little more lenient and brought the new video card above HD 5850 and HD 5830. But, as it happens, synthetics are far from an indicator.

In games, the Radeon HD 6850 worked with a bang. IN Resident Evil 5 she beat all competitors from NVIDIA and lost a meager 2-3% only to HD 6870 and HD 5850. Devil May Cry 4 GeForce video cards caught up and performed on par with the competitor from AMD: the difference was 2-5%. The same thing happened with Aliens vs. Predator. Both GTX 460, GTX 465, and HD 6850 showed 28-29 frames per second in it - an excellent result.

* * *

The Radeon HD 6850 was surprisingly fast. It lags behind the older version by 20% and is confidently on par with the GTX 460 and GTX 465. But the whole impression is spoiled by the price in Russia. If in the West an AMD board costs significantly less than its competitors, then here it is 500-1000 rubles more expensive than the GTX 460. And this is a very shame, because the HD 6850 was really a success.

Table 1

Comparison table of technical characteristics Characteristic AMD Radeon HD 6850 AMD Radeon HD 6870 AMD Radeon HD 5850 AMD Radeon HD 5830 NVIDIA GeForce GTX 465 NVIDIA GeForce GTX 460 Core Barts Pro Barts XT Cypress Pro Cypress XT GF100 GF104 Number of transistors 1.7 billion 1.7 billion 2.15 billion 2.15 billion 3 billion 1.95 billion Technical process 40nm 40nm 40nm 40nm 40nm 40nm Number of stream processors 960 pcs. 1120 pcs. 1440 pcs. 1120 pcs. 352 pcs. 336 pcs. Graphics core frequency 775 MHz 900 MHz 725 MHz 800 MHz 607 MHz 675 MHz Stream processor frequency 775 MHz 900 MHz 725 MHz 800 MHz 1215 MHz 1350 MHz Type, memory capacity GDDR5, 1 GB GDDR5, 1 GB GDDR5, 1 GB GDDR5, 1 GB GDDR5, 1 GB GDDR5, 1 GB Memory frequency 4000 MHz 4200 MHz 4000 MHz 4000 MHz 3200 MHz 3600 MHz Data bus 256 bit 256 bit 256 bit 256 bit 256 bit 256 bit Number of texture blocks 48 pcs. 56 pcs. 72 pcs. 56 pcs. 44 pcs. 56 pcs. Number of rasterization blocks 32 pcs. 32 pcs. 32 pcs. 16 pcs. 32 pcs. 32 pcs. Interface PCIe 2.0 x16 PCIe 2.0 x16 PCIe 2.0 x16 PCIe 2.0 x16 PCIe 2.0 x16 PCIe 2.0 x16 Price as of December 2010 7500 rubles 9300 rubles 7800 rubles 6700 rubles 7500 rubles 7000 rubles

table 2

Synthetic tests 3DMark Vantage Video card model GPU CPU Overall AMD Radeon HD 6850 12 596 16 499 13 365 100% AMD Radeon HD 6870 15 131 16 493 15 450 116% AMD Radeon HD 5850 14 832 17 597 15 427 116% AMD Radeon HD 5830 12 781 17 594 13 720 103% NVIDIA GeForce GTX 465 11 674 42 636 14 264 107% NVIDIA GeForce GTX 460 12 556 40 963 15 188 114% Unigine Heaven Benchmark 2.0 Video card model FPS Overall Performance ratio AMD Radeon HD 6850 11,8 298 100% AMD Radeon HD 6870 13,8 348 117% AMD Radeon HD 5850 11,4 288 97% AMD Radeon HD 5830 10,5 266 90% NVIDIA GeForce GTX 465 16,7 421 141% NVIDIA GeForce GTX 460 16,9 426 143%

Table 1

Gaming tests (frames per second) Game name, settings AMD Radeon HD 6850 AMD Radeon HD 6870 AMD Radeon HD 5850 AMD Radeon HD 5830 NVIDIA GeForce GTX 465 NVIDIA GeForce GTX 460 Resident Evil 5 (DX10) High, 1680x1050, AF 16x, AA 8x 93,9 96,1 92,4 74,5 83,7 83 High, 1920x1080, AF 16x, AA 8x 86,8 89,6 90,5 67 76,4 76,2 Performance ratio 100% 103% 102% 79% 89% 88% Devil May Cry 4 (SC2, DX10) SuperHigh, 1680x1050, AF 16x, AA 8x - - - - 95,7 101,3 SuperHigh, 1920x1080, AF 16x, AA 8x 92,6 126,3 114,8 77,7 93,3 93 Performance ratio 100% 136% 124% 84% 102% 105% Aliens vs. Predator (Demo, DX11) VeryHigh, 1680x1050, AF 16x, AA 2x 32,5 39,6 32,2 24,4 33,1 32 VeryHigh, 1920x1080, AF 16x, AA 2x 29 35,4 33,3 21,7 29,3 28,5 Performance ratio 100% 122% 107% 75% 101% 98% Value for money 100% 124% 104% 89% 100% 93% Performance ratio 100% 120% 111% 80% 97% 97%