Radeon HD 6870 What games will pull. Computer resource at SM. Test results: performance comparison
"... The performance of the Radeon HD 6870 fully corresponds to the price of 8,500 rubles. The card was successfully fitted between HD 5870 and overclocked GTX 460 versions, and in some cases even works at the level of more expensive GTX 470. If you are ready to spend on video
Gambling https: //www.Syt/ https: //www.Syt/
Iron shop
In October 2009, we - AMD Radeon HD 5870. After half a year NVIDIAreleased the answer - GTX 480. and GTX 470.. Today, the struggle goes to a new level: a fresh series of video cards appeared on sale AMD. - Radeon HD 6000.. At the moment there are only two boards in the line - HD 6850. and HD 6870.And on the tests we have the older version.
Re-read
Rumors about the appearance of a new series appeared long before the official release. Naturally, everyone thought that AMD was preparing a chip that could overtake GTX 480. It turned out not quite so. "New" series can be called with a large stretch. Video cards are based on the finalized version of the old CYPRESS processors and are positioned as MID-END solutions. So no FPS records.
Radeon HD 6870 has a BARTS crystal, an exact copy of the acquaintance of Cypress with HD 5870 is installed. Oddly enough, but the processor turned out to be trimmed: compared with the predecessor, Barts "lack" 450 million transistors. This means that the number of SIMD clusters has decreased on a new crystal of their total of 14 pieces instead of 20. From here, the loss in the number of streaming processors (on the barts of their 1120 against 1600) and textural blocks (56 instead of 80).
AMD compensated for losses simply: the frequency of the processor has grown from 850 to 900 MHz. Plus, the card was equipped with an "adult" tire on 256 bits and 1 GB of GDDR5 memory with a frequency of 4200 MHz. No rasterization blocks were not affected - all 32 ROPs are in place.
Freshness
There are in Barts and a number of improvements. For a start about tessellation. Here the AMD has two news at once. First: Tessellation engines are twice as fast. Second: with the settings you need to be careful. As an AMD explains, with extreme levels of tessellation, the model beats on such a number of polygons that the triangle screen takes no more than one pixel. In this case, the rasterization blocks begin to buck. For them, the optimal mode of operation is 16 pixels per polygon, otherwise they are idle. Plus, such small particles have to render several times, which also reduces performance. Conclusion, however, we will not see a honest, better picture, new polygons do not see.
The following improvement concerns antiazing. AMD has increased its work speed, and also entered the new Morphological AA mode. Its essence is that the corners are smoothed after the frame passed rendering. At the same time, the landmark is not the edges of polygons or alpha textures, but harsh, contrasting transitions. Thus, the antiazing is applied not selectively, but to the whole picture immediately. Plus this approach - the shader does not work with geometry, but with the ready-made stage, in fact with a graphic file, so the performance compared to traditional multi-sembam (MSAA) increases at times. Minus - with low screen resolution, the picture is cooled.
Improved in AMD and anisotropic filtering: now its quality and performance do not depend on the angle of rotation of the object. Additionally updated the video decoding acceleration unit - UVD3. In addition to Vc-1 and H.264 formats, it was taught to work with MPEG-2 (DVD), MVC and MPEG-4 (DivX, XVID).
Pink glasses
Well, the most important thing for us, players: AMD video cards are now working with 3D stereo. AMD call this feature HD3D., and it is drastically different from what NVIDIA offers. Instead of developing your own technologies and launch a new iron on sale, AMD used the help of partners.
For a start, she found special software developers, which will translate photos, video and games in stereo image. Among the Favorites were firms iZ3D and DDD. - They opened access to the source code drivers Catalyst.. Then AMD signed an agreement with the manufacturers of monitors and glasses. So if you want to try out HD3D, then the display will need to buy from LG., Samsung, CMI or ViewSonicand glasses u Xpand., Reald. and Bit Cauldron.. And after all, the most funny, a set of the program and active points should be cheaper than the similar proposal from NVIDIA. And most importantly, the support for stereo image is not laid in the cost of the video card, it is not necessary to overpay for the extra functional.
On the table
Radeon HD 6870 for tests sent by AMD itself, so we will tell about the reference option. From the predecessor of the video card differs in size (3 cm in short) and the appearance. The cooling system casing has become more angular, the red strip and "exhaust holes" disappeared in the rear end of the video card.
The fee is cooled on the principle of the turbine: 75 mm The fan delays air under the casing, drives it on the map and throws out of the computer case. On the crystal itself there is an aluminum radiator scored with 30 plates and penetrated with three copper heat pipes. During load, the card consumes only 151 W (37 watts less than HD 5870). In 2D-mode, HD 6870 requires 19 W. Power is provided by two 6-pin forks. Due to the low heat generation for hot air emissions, half of the slot on the I / O panel, all the rest of the space occupy new video connections.
The default Radeon HD 6870 is equipped with two DVIs, one HDMI 1.4A and Mini-DisplayPort 1.2 pair. All these figures hides a lot of interesting technologies. In DisplayPort 1.2, for example, bandwidth increased. Now two monitors can be connected to one connector. To do this, you will need a special splitter or display with the input and output for DisplayPort. Plus, in version 1.2, Blu-ray 3D support appeared and sound output as Dolby Truehd and DTS-HD Master Audio.
The new version of HDMI also learned to play Blu-ray 3D and, moreover, output stereo circuit on 3D TVs. True, the last function is more for a tick. TV manufacturers still have not come to a single specification of stereo imagery, so no one gives one hundred percent compatibility guarantee.
| What's in a name? | |
|
Radeon HD 6000 is far from the first time when the old architecture is produced under the new name. Such a scheme at one time tried NVIDIA when released modified GeForce 8000. under the guise of a new series GeForce 9000., and older versions later renamed generally GTS 250.. Why do you need it? The answer is simple - to attract the attention of the press and buyers. Well, seriously, new video cards could be called Radeon HD 5860 and HD 5840. But with such a situation, AMD could not be attracted to the tenth stake of the attention, which is now paid to new cards. But it is not worth grieving about the deceased line. Very soon AMD officially announce data on Radeon HD 6950., 6970 and 6990 . This is the real top, on a completely new processor called Cayman.. |
Test
Fine competitors for a new video card turned out to be not easy. At the time of delivery of the magazine, the fee has not yet come on sale, so focus on the recommended AMD price of 8000-8500 rubles. This means that HD 6870 is cheaper than HD 5870, but more expensive than HD 5850. Plus it enters the price category GTX 460 1 GB and its overclocked versions. For the same money, GTX 465 is sold, and just more expensive, somewhere for 10,500 rubles, you can buy GTX 470. I think a little thought, we decided to take all the video cards at once and compare the results.
To do this, we collected a powerful test stand based on the motherboard Foxconn Renaissance., installed on it Intel Core i7-920 and added 6 GB of RAM, recruited three straps Kingston Hyperx DDR3-1666.. Unfortunately, the test list turned out short. Since we tested the Radeon HD 6870 with a beta version of the drivers, we failed to run benchmark Colin Mcrae.DIRT 2.. So the performance of the DX11 we evaluated only in Aliens VS. Predator and synthetics Unigine Heaven Benchmark 2.0. But with projects on DX10 there were no problems. Standard Resident Evil 5. and Devil May Cry 4 - On the spot. The overall performance assessment was carried out using 3dmark vantage.
Balance
Synthetic test OT Futuremark. The position of HD 6870 very accurately determined: it does not reach the level of HD 5870, it works slightly slower GTX 470, overtakes GTX 460 and GTX 465. In Unigine Heaven Benchmark 2.0, evaluating performance in DX11, the work done by AMD above the tessellation engines. In the synthetic test, the Radeon HD 6870 overtakes more expensive HD 5870, but still lagging behind all competitors from NVIDIA.
In real gaming applications, the picture is preserved. Radeon HD 6870 confidently overtakes not only HD 5850 and GTX 460, 465, but also GTX 470. With DX11 such superiority does not work: GTX 470 faster by 19%, which is quite logical.
* * *
The performance of the Radeon HD 6870 fully corresponds to the price of 8,500 rubles. The card was successfully fitted between the HD 5870 and the overclocked versions of the GTX 460, and in some cases it even works at the level of more expensive GTX 470. If you are ready to spend 8,500 rubles on the video card, then the Radeon HD 6870 is the right choice, it does not have competitors.
| Table 1 | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| table 2 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Table 3. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Get acquainted with Radeon HD 6850 and 6870 closer
Many rumors were about the characteristics of cards based on the BARTS graphics core, which only insane versions did not put forward. Now, it's time to know the truth. Let's look at the plate, corrades, so neatly compiled by Gilbert:
| Specifications | Radeon HD 5770. | Radeon HD 5850. | Radeon HD 6850. | Radeon HD 6870. |
| Graphic kernel |
Juniper XT. | Cypress pro. | BARTS PRO. | Barts Xt. |
| Technical processing |
40nm | 40nm | 40nm | 40nm |
| Frequency of graphic nucleus |
850 MHz | 725 MHz | 775 MHz | 900 MHz |
| Number of streaming processors |
800 | 1440 | 960 | 1120 |
| Memory frequency |
4800 MHz | 4000 MHz | 4000 MHz | 4200 MHz |
| Memory bit |
128-bit | 256-bit | 256-bit | 256-bit |
| Number of typical memory |
1 GB GDDR5 | 1 GB GDDR5 | 1 GB GDDR5 | 1 GB GDDR5 |
| Energy consumption |
108 Watt. | 151 Watt. | ~ 127 watts. | \u003e 150 watt |
| Performance |
- | - | Faster 5830. | Faster 5850. |
Both cards support all the "feathers" DX11 (and how otherwise?), Their graphic chips consist of 1.7 billion transistors. From the important characteristics not specified in the table, it is worth noting for 6850 (competitor 768MB version of the GTX460): 12 SIMD blocks, 48 \u200b\u200btextural units, 32 ROPA and energy consumption in a simple ~ 19 watt. Solution performance - 1.5 TFLOPS. As can be seen - 5770 far behind, as it should be.
HD 6870, in turn, has 14 SIMD blocks, 56 textural units, 32 ROPs. His task is to show what GTX460 is from 1GB of memory. Energy consumption in a simple is ~ 19 watts, performance (according to which the novelty takes place between 5850 and 5870) - 2 TFLOPS.
Many "feathers" cards inherited from predecessors, but something new, such as DisplayPort video output versions 1.2 and HDMI 1.4A appeared. Added support for HD3D and UVD3 technologies and a new anti-aliasing method called Morphological AA. Add to this improved anisotropic filtering and increased (~ 2 times, compared with cards 5000 series) with tessellation.
And now you can admire the core of the novelties:
What is HD3D and UVD3?
AMD HD3D - Analogue of 3D Vision technology, allows you to play 3D games, watch 3D Blurei video, etc. All this with the help of special glasses. According to Guru3D, this technology is still realized - to view the video in 3D compulsory requirements is the presence of those most specials, a graphics card, 3D TV or monitor and HDMI 1.4A cable + software, without which all this is useless. In games, it's even worse - there are no kit "s (sets for developers, to implement 3D in games) and support this" feature "in the Catalyst driver settings, in other words, you will have to use third-party solutions, and not from AMD, and this is extra spending and troubles. As an example, the author of the original statement, tells about the software from DDD (Dynamic Digital Depth), which costs 50 bucks (but so far - 25, apparently discounts) and allows you to play 3D games, but for elegance Performance and quality, this decision is inferior to NVIDIA. In other words, the situation is such - the software developers will be glad to be able to be welded on the unstoppableness of the amdschnikov, the users of the same - on the contrary (because there are few people who do not have additional expenses).
Universal Video Decoder 3.0 - video decoding technology by graphics cards. The useful thing is undoubtedly. Everyone knows that although the modern PC is coping with the playback of Full HD video in any formats, it still creates a certain load on the central processor, but the UVD 3.0 technology is capable of significantly reduced this load. And what's new here, because we saw the same in UVD 2.0 and older versions? Here is a list of changes / innovations:
- Supported hardware decoding of two 1080P signals at the same time
- Compatibility with Windows Aero - Playing HD content during activated aero (i.e., the latter does not need to be turned off)
- Independent color scheme control from Windows desktop
- Brighter white color due to the use of stretching blue (?!)
- Control the shade of black and white colors during video playback
Also supported automatic color correction, contrast enhancing and support for Bitstream video in MPEG2 and MPEG4 DivX / XVID formats and audio, of course.
A new family of graphic adapters from AMD leads, if this did not happen, we would have shown the slide
Eyefinity.
About "iPhiniti", I think all you have heard. This technology allows you to connect to one video card from three to 6 monitors, and with a ruler 6000 it becomes even easier, because Connect six monitors to two ports to connect to two ports (displayport version 1.2 help), using an additional rectangular device. Here is a slide:
The most common will be the connection of three monitors, for this, no devices will not have to arrive.
Otherwise, the eyefinity technology itself has not changed much since the 5000th series. The author of the article with GURU3D gives an example that you can easily connect three 30 "monitor to seeing, each with a resolution of 2560x1600 and all this business will work correctly. The configuration of six monitors shows the image in the resolution of 7680x3200, which is quite cool. There are video:
Crossfirex
Little performance of one card? You can create from 2 or more seeas, CrossFirex tandem, which is certainly quenched thirst for speed (measured, in our case, by the number of frames per second).
What is needed to build a crossfayer? Just a couple of video cards, bridge for their connection and any motherboard with two (or more) PCIE X8 and PCIE X16 slots.
On Gur3D in tests, a reference design design of the HIS was used. 6850 is powered by only one 6-pin connector from BP, so that on CrossFire from such cards you need only two above-mentioned connector, for 6870 - four.
Here is what the bridge, connecting the cards in CrossFirex:
After installing the connecting bridge on both seats (pre-installed in the system in the corresponding PCIE slots), it remains only to turn on the PC and in the Catalyst driver settings, put a tick activating CFX mode:
That's all - now you can enjoy the dual power of new products!
Not so long ago, we published the material about the new lineup of the Radeon HD 6800 series video cards. It addressed all aspects of new technologies used in accelerators of this line, as well as improvements to the video cards on the BARTS kernel. It was noted that the expected change in the process to 32-nm did not occur. The graphic chips of the new generation of AMD Radeon HD 6800 series will still be made according to the old 40-nm technical process, the beginning of the use of which can be considered the presentation of the Radeon HD 4770, which passed in April 2009. The entire Radeon HD 5000 line also used 40-nM technology. All of the fault, as Eric Demerers said, who is the technical director of the GPU division in AMD, "the commitment of the developed technical process is currently related to economic reasons. Drawing attention to the percentage of workers crystals received today, the company collided with a choice, to launch two 32-nm or three 40-nm GPU at the same price. " Given this fact, in the near future AMD, most likely, will refuse 32 nm and will begin to develop production using 28 nm technology.
Without changing the technical process, AMD will be problematic positioning BARTS as a significant step forward. Mainly, the new GPU BARTS differs from its predecessor to serious changes in the field of architecture optimization, an increase in performance for each watt of consumed energy and a millimeter of the area, which improves overall efficiency. It also revised the main disadvantages of the previous ATI Radeon HD 5800 series associated with the low productivity of tessellation and geometry processing. As mentioned earlier in our reviews, it was these two drawbacks that became the most serious problem of AMD video cards compared to competing graphic accelerators of NVIDIA.
Initially, it was expected that the AMD Radeon 6800 rule should come to a change of AMD Radeon 5800, but this did not happen. As can be seen from the above scheme, the differences between the BARTS and CYPRESS kernels are practically no. However, a simple count of the number of SIMD engines, we are eloquently talking about significantly lower performance of the updated GPU Barts XT with respect to the CYPRESS kernel (Radeon HD 5870).
|
AMD Radeon HD 6850 |
ATI Radeon HD 5870 |
ATI Radeon HD 5850 |
ATI Radeon HD 5830 |
ATI Radeon HD 4870 |
||
|
Code name |
Barts Xt. |
|||||
|
Number of transistors |
1,7 billion |
|||||
|
Number of streaming processors |
||||||
|
Productivity, TFLOPS. |
||||||
|
Textural blocks |
||||||
|
Texture filtering, GTEXELS / S |
||||||
|
Number of ROPS. |
||||||
|
Pixel filtration, GPIXELS / S |
||||||
|
Z / Stencil, GSAMPLES / S |
||||||
|
Core frequency, MHz |
||||||
|
Memory frequency, GHz |
1.05 (4.2 GHz Effect - I) GDDR5 |
1.0 (4.0 GHz Effect - I) GDDR5 |
1,2 (4.8 GHz Effect - I) GDDR5 |
1.0 (4.0 GHz Effect - I) GDDR5 |
1.0 (4.0 GHz Effect - I) GDDR5 |
900 MHz (3.6 GHz Effect - I) GDDR5 |
|
Big box size |
256 bits |
|||||
|
Memory bandwidth, GB / C |
||||||
|
Personnel buffer |
||||||
|
TechProcess |
TSMC 40 nm |
|||||
|
Maximum / minimum power consumption, W |
||||||
In the Radeon HD 6870 graphics processor, fourteen SIMD engines are involved, each of them consists of four textural blocks and sixteen streaming processors. One streaming processor has five blocks of Allu (streaming nuclei). This total gives GPU 1120 streaming nuclei with 56 textural blocks. The result of the GPU processing is output through four clusters of the output part of the rendering conveyor. Each cluster includes eight ROP conveyors (raster operations), which make up the total thirty-two ROP conveyors. The 256-bit memory interface is provided by four 64-bit memory controllers.
The numerical characteristics of the architecture and the specification of the RADEON HD 6870 video cards are very similar to those of the Radeon HD 5830, but with a doubled end part of the rendering conveyor. However, if we consider the opposite direction, the Radeon HD 6870 uses the same end part of the rendering conveyor as Radeon HD 5870, but less than the shader nuclei. For the Radeon HD 5830 video card, the main disadvantage was just reduced twice the number of ROP blocks, which caused delays in performance. Radeon HD 6870 can be called finalized and improved Radeon HD 5830 graphics processor reincarnation. In its equipment Barts XT core, there are as many two UTDP control (Ultra-Thread Dispatch Processor), while Cypress has been only one, and has also been improved part GPU, which is responsible for processing tessellation algorithms. Such changes should provide almost two-time superiority of the new solution in the processing of tescole streams over the graphics accelerators of the previous series.
Another significant improvement in which the Radeon HD 6870 has an optimized crystal structure has become the possibility of its work at higher frequencies, which significantly affects the final performance. The Radeon HD 6870 frequency is 900 MHz, which means the simplified GPU BARTS core works with a significantly higher clock frequency than Radeon HD 5830 or even Radeon HD 5850. Such an increase in frequency leads to an increase in GPU overall performance by almost 25% compared to Radeon HD 5850, operating with a frequency of 725 MHz. As a result, a smaller amount of alu and textural blocks at the Radeon HD 6870, compared with the Radeon HD 5850, is fully compensated by the increased frequency of the GPU kernel, which practically equalizes their performance.
All these improvements allowed the Radeon HD 6870 to be more productive than the Radeon HD 5850, with a lesser 25% of the crystal area. Such a change explicitly reduces power consumption in the mode of inactivity and the cost of production of the nucleus itself as a whole. These two explicit advantages should have a positive effect on the popularity of the final product. Radeon HD 6870 memory runs on the "native" frequency of 1050 MHz (effective 4200 MHz), a little faster than the Radeon HD 5850 (1000 MHz). According to the manufacturers, the RADEON HD 6870 retail price will be $ 240, and it is a little less than $ 260 on Radeon HD 5850.
Theoretical information is very encouraging, but always want to get concrete results in test applications. For these purposes in our laboratory and hit a video screen. The model provided by MSI is a complete copy of the reference sample produced at AMD factories. It is for this reason that our review and testing, in fact, will describe the reference Radeon HD 6870 solution from AMD.
Specification
There is nothing unusual in the specification, all the characteristics of the reference solution have already been listed above by us above, now only worth repeating them for a specific instance sold under the MSI brand name.
|
MSI Radeon HD 6870 (R6870-2PM2D1GD5) |
|
|
Graphic kernel |
AMD Radeon HD 6870 (Barts XT) |
|
Conveyer |
1120 Unified streaming |
|
Supported APIs |
DirectX 11 (Shader Model 5.0) |
|
Core frequency, MHz |
|
|
Volume (type) of memory, MB |
|
|
Memory frequency (efficient), MHz |
|
|
Memory bus, bit |
|
|
Tire Standard |
PCI Express X16 2.1 |
|
Maximum resolution |
Up to 4 displays of 1920x1200 each |
|
2x DVI-I (VGA through adapters) |
|
|
HDCP support |
there is |
|
Drivers |
Fresh drivers can be downloaded from: |
|
Products webpage |
A few innovations are visible from the table, which began to have "reference" Radeon HD 6870 video cards. The DisplayPort port 1.2 allows you to display the image immediately into several monitors using a special hub, and different permission can be installed for each monitor. Supported up to four displays with a resolution to 1920x1200 each. However, the total accelerator can be connected to six monitors using two hubs and two DVI-I connectors, but already with a resolution up to 1600x900.
The presence of HDMI 1.4 will allow the user to take advantage of the 3D stereo transmission technology, which is necessary for connecting 3D monitors. Support for new technologies OpenGL 4.1 and AMD HD3D, as well as already well-known to us on the 5800th series of accelerators of AMD Eyefinity and AMD EYESPEED. Such a variety of new capabilities of using the accelerator will be a worthy answer to AMD to the proposed NVIDIA in the 400th series of partially similar technological solutions.
The video screen is presented in a dense box. In her design, the possibility of working with the branded overclocking of the MSI Afterburner is emphasized. The company's logo is visible, and on the right side, the manufacturer of GPU AMD Radeon Graphics will be mentioned, which indicates the release of this line already by AMD, and not ATI (AMD) as it was before. The R6870 graphical core itself is indicated below, the presence of 1 GB GDDR5 accelerator, the DIRECTX 11 support and the presence on the rear panel of two Mini DisplayPort connectors. You can connect up to four screens through a special hub to MINI DisplayPort connectors using AMD Eyefinity technology. In the very bottom there are logos of support by the product of MSI branded technologies for overclocking and use on textolite boards of capacitors of increased reliability and durability.
The last two features are described in more detail on the back of the box.
Here at the top of the R6870-2PM2D1GD5 graphics accelerator itself, and the support of the MSI AfterBurner utility is described in detail, which allows you to disperse both one accelerator and somewhat installed in CrossFirex mode. There is the possibility of conservation of acceleration profiles, as well as profiles with reduced frequencies of the video card, which leads to economy of electricity and reduced heat generation, and therefore a decrease in the total noise of the system. Below is the use of solid-state capacitors in this model, which guarantee 10 years of trouble-free operation. The right marked the main advantages of this accelerator in 30 languages. Packaging leaves the impression of a high-quality completed product and has full information about internal content.
MSI Radeon HD 6870 Supply Set (R6870-2PM2D1GD5) is quite worthy:
- Two adapters from two peripheral power connectors to one 6-pin video card power connector;
- Crossfire bridge;
- Drive with drivers and utilities;
- Brief instructions for installing a video card;
- MINI DisplayPort adapter on DisplayPort;
- Adapter with DVI on VGA.
Fully acquitted kit, providing 80% of customer requests. But to connect a group of displays here, it is clearly necessary to purchase the necessary adapters or a special hub, with simultaneous support to four monitors.
Compared to the flagship from NVIDIA, a video card on the GeForce GTX 480, a new solution from AMD looks more elegant. This is not only the first impression, but also judging by the declared innovations of AMD, the accelerators on the Radeon HD 6870 should be two times less power consumption with comparable in some performance applications. The Radeon HD 6870 video card differs from the "top" solutions of both lesser manufacturers, it has, as well as the Radeon HD 5870, two 6-pin additional power connections located on the top end of the video card. The connection of the additional power will not cause inconvenience, and the wires themselves will not interfere with the installation of the video card, increasing its length by the connectors themselves and the wires extending from them.
What is surprising, the reference sample is only one CrossFire connector. Most likely, subsequently, decisions will appear from another manufacturer with two CrossFire connections, which are very useful to increase the overall performance of the video system when this accelerator is connected in the CrossFirex modes together with other similar models.
The back of the AMD Radeon HD 6870 video card has no technological holes, everything is closed with a plastic casing, harmoniously complementing the overall accelerator design. It's nice that power connectors are rotated up - this allows the accelerator to be compatible with a large number of housings.
The reverse side of the accelerator did not receive a single meaningful element.
Thanks to the two-plate design of the cooling system on the rear panel of the accelerator, five displays connectors were placed (2x DVI-I, HDMI 1.4, 2x Mini DisplayPort 1.2). Here there is a sufficiently large grille for the removal of hot air from the housing. Such a variety of connection connectors will provide AMD Radeon HD 6870 full compatibility with all the latest offers of the image output devices.
On the front of the printed circuit board, you can select a graphics processor and the eight video memory chips surrounding it. Here are all the elements of the GPU power supply system and the accelerator memory.
The frequency of the graphics processor is 900 MHz. The BARTS XT chip itself is made in 40 nm in the process and contains 1120 universal sheer conveyors and thirty-two rounding blocks, and the exchange of data between the graphics core and memory is carried out through a 256-bit bus.
In the absence of a load, the frequency of the chip is reduced to 100 MHz, and the video memory slows up to 81 MHz (efficient frequency of 324 MHz). This reduces the supply voltage, which ensures a noticeable reduction in power consumption and temperature.
The video memory of 1 GB is typed by eight microcircuits of the GDDR5 standard produced by Hynix H5GQ1H24AFR T2C with 0.8 ns access time, which allows them to operate at an effective frequency up to 5000 MHz. Since the resulting frequency of memory chips on a video card is slightly lower and is 4,200 MHz, then a good frequency corridor remains about the reserve, which we hope can be used when overclocking.
The accelerator has a 4 + 2-phase power scheme. Here, four phases go to the GPU and implemented using the CHIL CHL8214 controller having support for the I2C protocol. This protocol must provide software using the new version of the MSI AfterBurner utility. The MSI AfterBurner 2.0.0 utility comes with a video card with a video card, it did not allow the voltage to the GPU.
The cooling system itself consists of two parts. The metal casing through the thermal interface comes into contact with all the hot board elements: video memory chips and power subsystem elements. For cooling GPU, a separate design is used - a massive aluminum radiator with three copper thermal tubes of different diameters, which are soldered to the copper plate in contact with the GPU crystal. The central copper tube of increased diameter should contribute to a more efficient heat removal from the GPU crystal. The thermal interface is quite viscous, which will also be positively affected on the overall efficiency of the cooling system. A turbine type fan having standard sizes, with excessive enough to cool the graphics processor. In general, the cooling system can be called improved and more technological compared to the reference turbine used on the hot models of the 5800th series.
To assess the efficiency of the cooling system, we used the utility Furmark. And detailed monitoring was carried out using GPU-Z and MSI Afterburner.
Working on standard frequencies and controlling the speed of rotation of the turbine automatically, the graphics processor warmed up to 82 ° C. Considering that at the same time, the speed of rotation of the turbine was only 35% of the maximum and the entire cooler worked quietly, we can talk about the weak heat dissipation of the accelerator and the excellent work of the reference cooling system in automatic mode.
After we manually installed the speed of rotation of the turbine to a maximum, without removing the load, the cooler began to make noticeably, but the temperature of the graphics core fell to 67 degrees.
In general, the cooling system used on a reference solution AMD Radeon HD 6870 showed high efficiency, while it remained quite quiet in the work.
When testing, a stand for testing video cards No. 2 was used
| CPU | Intel Core 2 Quad Q9550 (LGA775, 2.83 GHz, L2 12 MB) @ 3.8 GHz |
| motherboards | Zotac NForce 790i-Supreme (LGA775, NForce 790i Ultra SLI, DDR3, ATX) Gigabyte GA-EP45T-DS3R (LGA775, Intel P45, DDR3, ATX) |
| Coolers | Noctua NH-U12P (LGA775, 54.33 CFM, 12,6-19.8 dB) THERMALRIGHT SI-128 (LGA775) + Vizo Starlet UVLED120 (62.7 CFM, 31.1 dB) |
| Additional cooling | VIZO PROPELLER PCL-201 (+1 Slot, 16.0-28.3 CFM, 20 dB) |
| RAM | 2x DDR3-1333 1024 MB Kingston PC3-10600 (KVR1333D3N9 / 1G) |
| Hard drives | Hitachi Deskstar HDS721616PLA380 (160 GB, 16 MB, SATA-300) |
| Power supplies | SEASONIC M12D-850 (850 W, 120 mm, 20 dB) Seasonic SS-650JT (650 W, 120 mm, 39.1 dB) |
| Housing | SPIRE SWORDFIN SP9007B (Full Tower) + Coolink SWIF 1202 (120x120x25, 53 CFM, 24 dB) |
| Monitor | Samsung Syncmaster 757MB (DynaFlat, [Email Protected] Hz, MPR II, TCO "99) |
Select what you want to compare Radeon HD6870 1GB DDR5 MSI
Evaluating the resulting test results do not need to forget that the performance of the graphics accelerator among other things also depends on the optimization of the drivers. Therefore, the results came out not entirely unequivocal. In half the tested games, the MSI Radeon HD 6870 video card performance is located between the Radeon HD 5870 and Radeon HD 5850. At the same time, the Radeon HD 6870 is on a par with Radeon HD 5870. At the same time there are a pair of three games where the Radeon HD 6870 video accelerator is even a RADEON HD 5850. Thus, nevertheless, in terms of performance Radeon HD 6870, you can approximately be considered equal to Radeon HD 5850. But it is in games without tessellation, that is, with the support only DirectX 9 and DirectX 10. Comparing MSI Radeon HD 6870 with NVIDIA video cards , You can note superiority over GeForce GTX 460 with 1 GB of memory, which threatens 20%. The video screen on the GeForce GTX 470 in many tests is on a par with MSI Radeon HD 6870, but there are games in which the first has noticeably better performance.
According to the results of the Suniengine Heaven 1.0 synthetic application, we can state the explicit progress of the AMD Radeon HD 6870 video accelerator when building a 3D scenes with tessellation. Of course, a two-time increase in productivity in tests with tessellation at the BARTS graphics core relative to Cypress, but the increase is clearly noticeable. As a result, the AMD Radeon HD 6870 video screen showed very similar with the GeForce GTX 470 result that in itself is significant.
Acceleration MSI Radeon HD 6870 (R6870-2PM2D1GD5)
The video card when cooled by its standard turbine operating in the maximum efficiency mode, was able to function consistently at frequencies: 975 MHz for the graphics nucleus (+ 8.3%) and 1160 MHz (+ 10.4%) for video memory, which ensured its operation to efficiently Frequency 4640 MHz. This is not the best result for the GPU, but the situation should dramatically change when the utility is released that allows you to raise the supply voltage on the graphics core. Memory chips also could not exhaust their potential fully, but the increase in 10% can also be considered quite normal. Now it is worth evaluating the performance gains from overclocking in real tests.
|
Test package |
Standard frequencies |
Overclocked video card |
Performance growth,% |
|||||||||||||||||||||||||||||||||||||
|
Far Cry 2, Maximum Quality, NO AA / AF, FPS |
||||||||||||||||||||||||||||||||||||||||
|
MSI GeForce GTS 450 1GB GDDR5 CYCLONE |
||
|
SAPPHIRE RADEON HD 5770 1 GB GDDR5 FLEX |
||
|
Zotac GeForce GTX 460 1GB GDDR5 |
||
|
Gigabyte Radeon HD 5850 1GB GDDR5 |
||
|
MSI Radeon HD 6870 1GB GDDR5 |
||
|
Gigabyte Radeon HD 5870 1GB GDDR5 |
||
|
ZOTAC GeForce GTX 470 AMP! 1280 MB GDDR5 |
||
|
ZOTAC GeForce GTX 480 1.5 GB GDDR5 |
||
|
MSI Radeon HD 5970 2GB GDDR5 |
MSI Radeon HD 6870 1GB GDDR5 Video Card Energy Care Conditionally can be equated to Radeon HD 5850 1GB GDDR5 power consumption. A slightly smaller level of energy consumption has a competitor solution at the GeForce GTX 460 C 1GB GDDR5 memory accelerator. The video card is the Gigabyte Radeon HD 5870 1GB GDDR5 in the "consumption" of energy exceeds MSI Radeon HD 6870 1GB GDDR5, actually, as well as in terms of performance. As you can see everything in our places, it is so hard to talk about some kind of essential superiority of the graphic kernel BARTS over Cypress in terms of better performance ratio on watts spent energy. Perhaps the power consumption of a new chip and decreased, but not very much.
The individual "mansion" is two other "top" NVIDIA accelerator: GeForce GTX 470 and GeForce GTX 480. For power consumption GeForce GTX 470 is inferior to Radeon HD 6870 approximately 60 W in load mode, and the difference between the energy consumption of GeForce GTX 480 and Radeon HD 6870 is integers W It is noticeable that NVIDIA engineers tried to squeeze a maximum of their graphic accelerator on the GF100. As a result, the energy consumption of GeForce GTX 480 1.5 GB GDDR5 turned out to be almost the same as the MSI Radeon HD 5970 2GB GDDR5 2GB video card.
conclusions
The firstborn of the new generation of AMD video cards was not so productive as expected, but given the recommended price for solutions on the BARTS kernel, one can safely talk about their high competitiveness by the criterion of price / performance. Although the price criterion of choice still can be challenged in mind the temporary absence of Radeon HD 6870 models in a free sale. A total of a total of the 6800th series of AMD accelerators for introducing a large number of new features in a new lineup of a large number of new features, such as: Using DisplayPort 1.2 with support for up to 4 displays through a hub; HDMI 1.4, allowing the user to use the 3D image transmission technology; Support for new technologies OpenGL 4.1 and AMD HD3D. AMD engineers left aside and already well-known technologies like AMD Eyefinity and AMD Eyespeed. To good qualities of accelerators on the Radeon HD 6870, they can be attributed to the heat release and increased efficiency for each millimeter of the crystal area and the very low level of the standard turbine, even at maximum load, it is lower than the average.
Directly tested by us video card MSI Radeon HD 6870 (R6870-2PM2D1GD5) Not much different from the reference sample, but MSI's engineers will soon provide the MSI Afterburner branded utility with the ability to change the voltage on the GPU video screen, which should significantly affect its ability to overclock. Video cards on the Radeon HD 6870 should level, then a slight advantage that NVIDIA has received with the release of the 400th series of video cards. For this, the AMD Barts core has all the prerequisites, and most importantly, a fairly low price for graphic accelerators on Radeon HD 6870.
Advantages:
- Low recommended price;
- Low noise level of standard turbine;
- Support for ATI Eyefinity technology with up to 6 displays;
- DIRECTX 11 support (Shader Model 5.0), OpenGL 4.1 and AMD HD3D.
- Availability of DisplayPort 1.2 with support for up to 4 displays through a hub;
- Support HDMI 1.4 for 3D stereo image transmission technology; Published 03-11-2010
The article read 32075 times (a)
| Subscribe to our channels | |||||
 |
 |
||||
It seems to me that over time, even at the most conservative person, consciousness flows towards changes. We, consumers of computer technologies, were taught from year to year and learned to the fact that the announcement of the new generation of video cards should be accompanied by the output of the top decision. But AMD broke all the stereotypes, released its Middle-End to the AMD Radeon HD 6850/6870 at the beginning of the month. A small acquaintance in this video, which we prepared for the new number "I", and more information and tests are looking for a cat.
Immediately I was recruited February 2008 and the launch of the ATI Radeon HD3870X2 video card. Many people in Eldorado, drawing newcomk video cards and extreme acceleration. And already almost after three years, we can see the video card for three generations older, and in terms of performance, perhaps, even higher than three times.
All as usual, we look at the specifications:
Probably, it's time and tell about the names and models of the new AMD series. The ATI brand can already be forgotten, we will no longer see it. Now only Radeon will remind us of red video cards.
Before the release of the 6000 series, AMD video card had the following division by classes depending on the name.
The first figure in the title meant a generation of video cards, that is, the current novelties is 6 generation (within the framework of this name), and the already familiar to us is the fifth. The second digit was responsible for the market segment, which is focused on this map. The top class was gained by the figure 8, and all that was gradually lowered down from 8 to 3, just meant the staircase from Hi-End to Low-end. The third digit meant the level of performance within this market segment. 7 - fast, 5 - medium, 3 - slowly. Well, the last digit was 0, which did not affect anything.
There was an exception. This is a Radeon HD5970, which did not fit into all this merarchia because of his two-indifference. Just figure 9 and meant that two graphics processors are planted on the board.
What happened now?
AMD Radeon HD6870 is not a single-core top, as many may seem from the name. The second digit in the title passed several positions. 8 is no longer Hi-End. Now there will be a full-fledged HD69xx card line.
AMD Radeon HD6990 is the very top with two GPUs.
AMD Radeon HD6970 - single-core top. (faster than GTX480 and Radeon HD5870)
AMD Radeon HD6870 - Middle-End, GTX460 1GB competitor, price up to 9,000 rubles in the Russian retail with recommended $ 238.
AMD Radeon HD6850 - Middle-End, GTX460 768 MB competitor.
AMD Radeon HD5770 - Already the famous old man remains in the game. Rename will not.
Quickly might at the interesting moments of the novelties:
Two DVI ports - do not surprise anyone, this is the standard, as well as the presence of one HDMI. But DisplayPort from the fifth generation turned into two minidisplayport in the sixth. And representatives of AMD even told us how to assemble the AMD Eyefinity system with six monitors and one such video card. Two monitors via DVI and one through HDMI is enough to work AMD eyefinity with three monitors. For "igromeaks" and owners of six monitors and video cards of the sixth generation AMD, the remaining three monitors can be connected in two ways. Buying a special adapter from one minidispayport for two, and then for three monitors there are free three connectors. Or wait for super new monitors that will allow you to consistently connect to each other through one MinidisPlayport.
And how to make noise?
With this, AMD has never had problems))) noise approximately as well as the previous video cards of the fifth generation. After all, no changes in the design of the turbine did not happen. Loud - yes, but also cool well, why then change?
Two PCI-E sixpin connections feed Radeon HD6870. Here is a full order, because in my opinion, the power system 6 + 6 is the optimal combination.
Tests, results:
Test stand:
Intel Core i7 980x
ASUS RAMPAGE III EXTREME
2x2 GB Corsair Dominator GTX2
128GB SSD ADATA S596
1200W Antec TPQ.
Conclusions:
My opinion that AMD was once again (or rather the third time in a row) to keep leadership in the market. The network more and more information is manifested that today the day of removing NDA with GTX580, but there is no information about this. The conclusion suggests one - AMD and this time will release a new generation faster and for a couple of month will remove the most cream with fans of fast speeds!
The second half of the past and most of this year passed for the graphic division of AMD a very successful. Judge for yourself, the release of the world's first graphics accelerator with DIRECTX 11 support, which was distinguished by excellent performance and relatively cold moral, forced many to take a fresh look at the Radeon family. Often, even the most sable fans of the Californian NVIDIA products moved to the AMD side, because their favorite company, at that time, could not provide an alternative. As a result, NVIDIA lost a significant market share of the desktop chart, giving way to AMD leader.
Now for AMD, the main task is to strengthen its positions and further development of success, while NVIDIA will try to change their position and establish the release of competitive solutions based on the Fermi architecture. Some of these decisions that have already published on the market create a lot of trouble for representatives of the Red Camp. For example, GeForce GTX 460 sufficiently successfully competes with Radeon HD 5830, and GeForce GTX 470 together with the flagship in the face of GTX 480 looks quite confident. And if with the eldest accelerator on the basis of Fermi, you can deal with a decrease in the cost of Radeon HD 5870, then in the middle segment of the possibilities for maneuvers not so much. Considering that the manufacturers receive great profits from the "middling", and not from top solutions, the most logical step with the promotion of new products in this situation is "Attack" at the position of GeForce GTX 460 and GeForce GTX 470. For this, as a weapon in AMD chose Accelerators Radeon HD 6850 and Radeon HD 6870. Well, let's get acquainted with new items.
Let's start with the name. Despite the higher digital index, the AMD company itself positions the novels by no means as a full replacement for Radeon HD 5870. Radeon HD 6850 accelerators, like Radeon HD 6870, in most cases will be slower than the company's single-chip. And changes in the digital index are associated with the next update of the AMD GPU architecture, and this time, as already mentioned, "Red" decided to start expansion not from the top, but from the middle.
AMD's official slides say that the new Radeon HD 6870 in terms of performance should be a little faster than Radeon HD 5850, and Radeon HD 6850 will take place between the Radeon HD 5770 and Radeon HD 5850.
Senior accelerators based on new GPUs, which will receive marketing names Radeon HD 6970 and Radeon HD 6990. About the most senior representatives of the new family so far nothing is unknown, but about Radeon HD 6850 and 6870 we have something to tell.
AMD Radeon HD 6850 and Radeon HD 6870 accelerators are built on the basis of graphic processors with code name BARTS. From the point of view of architecture, new GPUs do not contain revolutionary changes. All modifications conducted by AMD engineers were aimed at improving the consumer characteristics of new products.
So, here are the main parameters, over the improvement of which AMD engineers worked:
- Higher performance per unit area of \u200b\u200bcrystal, thanks to internal GPU optimizations;
- Updated DX11 architecture, higher geometric performance, as well as advanced tessellation;
- Updated and improved 3D graphics quality improvement technologies. The appearance of a new smoothing mode and an improvement in the anisotropic filtering algorithm;
- Improved acceleration of work with multimedia content. Support UVD3, AMD App, Blu-ray 3D;
- Improved content output technologies. Support AMD Eyefinity, HDMI 1.4A, DP 1.2.
As for the internal GPU BARTS device, then it is, as we said earlier, was not substantially recycling since CYPRESS:
CYPRESS graphics processor flowchart
BARTS XT graphic processor flowchart
Compared to CYPRESS chip, the new GPU Barts XT received a smaller function block. Thus, the amount of SIMD nuclei, including 16 shader processors, decreased from 20 to 14 pieces, as a result, the total number of shader processors decreased from 1600 (320) to 1120 (224). For each SIMD core accounts for 4 textural blocks, as in the case of GPU CYPRESS, but the total number of TMU from Barts is less than - 56 (against 80 from Cypress).
For the most efficient control of the threads of vertex, geometric and pixel shaders in modern graphics processors, there is a so-called stream manager, called "Ultra-Threaded Dispatch Processor". In the CYPRESS chip, only one such dispatcher, but judging by the BARTS XT flowchart, this GPU has already two such processors.
Perhaps one of the main differences of the new graphics processor was the improved tessellation unit. According to AMD, its performance in the case of Radeon HD 6870 can be up to two times higher than the Radeon HD 5870, however, the extension depends on the tessellation level directly.
Despite the significant increase in the quality of 3D models with tessellization included, for modern GPUs, it is still not so easy to maintain a fairly high FPS at the maximum possible level of tessellation. This slide shows an example in which one object is shown, with different degrees of tessellation. If one triangle accounts for only one pixel, then in the rasterization process there are a number of problems, for example, insufficient disposal of the rasterization unit, the so-called overshading effect (when each pixel is painted more than 8 times), in addition, there are too many polygons in such a scene to It was possible without prejudice to activate multisampling.
Of course, this problem is not to solve the simple buildup of the "coarse" graphics processor. Much more efficiently use a number of optimizations. For the maximum efficient operation of the rasterization unit of modern GPU AMD, it is required that one triangle accounted for about 16 pixels. In addition, the high degree of tessellation is worth using only if:
- The object is as close as possible to the camera;
- Only on the contour of the object or in those parts where it is necessary to show a lot of small parts.
In turn, on remote objects, you can use a low degree of tessellation to increase productivity.
For lovers of maximum quality, AMD pictures prepared a pleasant surprise. Owners of Radeon HD 5XXX / 6XXX will be available new method of full-screen smoothing - Morphological AA
To implement MLAA algorithms, a special post-filter is used, whose work, in contrast to some other methods of full-screen smoothing, does not depend on the location of the boundaries of the polygons. The MLAA acceleration is carried out by directcompute. If the MLAA work is simplified simplified, it turns out about the following: first the algorithm detects high-contrast vertices, analyzes them and based on the data obtained produces colors for each pixel.
The data obtained can be used again, and to avoid redundant data sample and, as a result, the performance drops are used, a special LDS buffer (Local Data Share) is used.
We assume that the quality of the new algorithm must be close to 4x MSAA, with performance, according to AMD representatives will be close to a simpler CFAA mode. The new smoothing algorithm is compatible with any DirectX 9/10/11 by the application, it will be necessary to enable the appropriate option in the control panel of the drivers. We will definitely check how the appropriate opportunity is presented.
Not forgotten in AMD and about improving the quality of anisotropic filtering. New Radeon HD 6850/6870 accelerators support advanced AF operation algorithms, the quality of the implementation of which does not depend on the viewing angle, while practically does not affect performance.
For Radeon HD 6850/6870, new options for setting up Catalyst A.I are available in new AMD drivers. Changing these parameters affects the quality of texture filtering. Three positions are available:
- High Quality. In this mode, all optimization of working with textures is turned off;
- Quality.. In this mode, which, by the way, is enabled by default, trilinear optimization and optimization of anisotropic filtering works. As AMD states, the "sharpening" used in no way affect the quality of the picture;
- Performance. When this position is selected, the trilinear and anisotropic filtering optimization works in more aggressive mode, which allows you to raise speed, but you can notice the loss of quality picture.
After entering the AMD Radeon HD 5870 market, enough time passed, before we were able to independently check in the EyeFinity technology. Already at that time, this technology worked fairly stable enough, and most importantly, a positive effect was felt in games. During the announcement of the Radeon HD 6850/6870, AMD representatives did not forget to mention a number of improvements in Eyefinity, among which the ability to create multiple eyefinity groups, improved displays, color control algorithms, color correction support on each single monitor, etc.
Moreover, this time in AMD decided directly compare Eyefinity with similar NVIDIA technology, which, by the way, was announced after a while after Eyefinity. Some of the arguments given looks very convincing, but some can still argue with some.
Be that as it may, now AMD EyeFinity looks like a more completed solution that has passed the annual run in the market and judging from everything proved its viability.
But in terms of the implementation of technologies to view 3D content AMD was not the first. For a long time, NVIDIA has been promoting 3DVision technology to the market, the support of which is implemented in hundreds of games. However, not to be first, does not mean becoming the worst. And in AMD, this is perfectly understood. Therefore, along with the launch of Radeon HD 6850/6870 accelerators, the manufacturer announced AMD HD3D technology, which will allow owners of new AMD video cards to enjoy 3D stereo games in full. Unlike NVIDIA, which independently sells glasses for 3DVision, AMD allows third-party manufacturers to develop 3D glasses. The final consumer is from this only benefit, of course, if the proposed solutions will be quite high-quality.
Finally, it is impossible not to mention that AMD is trying to actively participate in the promotion of parallel computing by graphic processors, as well as use internal video card resources to speed up the playback of multimedia content, for example, Blu-ray 3D. In particular, the 3rd generation video processor is built in GPU BARTS.
To summarize the theoretical part of our material, we have compiled a table containing data on the technical features of modern solutions AMD and NVIDIA:
| Property | Title video cards | |||||||||
| NVIDIA GeForce GTS 450 1 GB | NVIDIA GeForce GTX 460 1 GB | NVIDIA GeForce GTX 470 1280 MB | AMD Radeon HD 5750 1 GB | AMD Radeon HD 5770 1 GB | AMD Radeon HD 5830 1 GB | AMD Radeon HD 5850 1 GB | AMD Radeon HD 5870 1 GB | AMD Radeon HD 6850 1 GB | AMD Radeon HD 6870 1 GB | |
| Code name nucleus | GF106. | GF104. | GF100 | Juniper | Juniper XT. | Cypres. | Cypres. | Cypress XT. | Barts. | Barts Xt. |
| TEHPROTSESS, NM | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 | 40 |
| Number of transistors, million pcs. | 1170 | 1950 | 3200 | 1040 | 1040 | 2150 | 2150 | 2150 | 1700 | 1700 |
| Maximum power consumption, W | 106 | 160 | 215 | 86 | 108 | 175 | 170 | 188 | 127 | 151 |
| GPU kernel frequency, MHz | 783 | 675 | 607 | 720 | 850 | 800 | 725 | 850 | 775 | 900 |
| Number of blocks ROP, pcs | 16 | 32 | 40 | 16 | 16 | 16 | 32 | 32 | 32 | 32 |
| Number of TMU, pcs | 32 | 64 | 56 | 36 | 40 | 56 | 72 | 80 | 48 | 56 |
| Number of CUDA blocks / Universal processors | 192 | 336 | 448 | 720 | 800 | 1120 | 1440 | 1600 | 960 | 1120 |
| Frequency of a shader domain, MHz | 1566 | 1350 | 1215 | |||||||
| Type of video memory | GDDR-5. | GDDR-5. | GDDR-5. | GDDR-5. | GDDR-5. | GDDR-5. | GDDR-5. | GDDR-5. | GDDR-5. | GDDR-5. |
| Bit Bit Bit | 128 | 256 | 320 | 128 | 128 | 256 | 256 | 256 | 256 | 256 |
| Effective video memory frequency, MHz | 3600 | 3600 | 3348 | 4600 | 4800 | 4000 | 4000 | 4800 | 4000 | 4200 |
| PSP video memory, GB / s | 57,7 | 115,2 | 133.9 | 73,6 | 76,8 | 128 | 128 | 153.6 | 128 | 134,4 |
| Approximate retail value according to Market 3DNews | 4600 | 6700 | 10200 | 3800 | 4700 | 7500 | 8700 | 12500 | ~179$ | ~239$ |
According to preliminary data, the cost of Radeon HD 6850 and Radeon HD 6870 will be about 179 and 239 dollars, respectively. Well, let's see later, we will see whether new items can justify their price tag to the level of performance in modern games, but for now go to an external inspection.
How to pay a domain name
Domain zone of tokelau islands
What is domain what problems may be
Yandex Wordstat: detailed instructions for using the service and grouping operators and a complicated request
Editing DBF files
Xenu Link Sleuth - What is this program how to use the Xenu program
Methods Copy and insert text from keyboard without using mouse