With the release of the Radeon HD 6850 and Radeon HD 6870 graphics cards, AMD delivers fantastic value for the average PC gamer. If you want great game performance but don’t want to spend too much money, you look for a graphics card in the $150 to $250 range. It appears as though the Radeon HD 6800 cards are set to be the best performers at their respective prices by a considerable margin. The architecture is based on last year’s Radeon HD 5800 graphics cards, but the HD 6800 cards also incorporatea host of optimizations such as improved tessellation performance, better video decoding, and more-robust display output options. The leap in performance offered by these cards is often the difference between being able to play games at full settings and having to reduce resolution or detail levels to keep things running smoothly. They’re quite energy-efficient, too.
Before we dig into the new features that these cards offer, let’s take a look at the specs. The main chip architecture is a modified and enhanced version of what we saw in the Radeon HD 5850 and 5870, which used a GPU code-named Cypress (part of the Evergreen family). The new chips use a GPU code-named Barts (the first in the Northern Islands family). The Barts GPU is smaller than Cypress, and is manufactured by TSMC using its 40nm process technology, but it delivers excellent performance. For comparison, we’ll include the specs of the less expensive Radeon HD 5770, which is based on the Juniper GPU of the Evergreen family.
A quick glance at the numbers shows that the new cards use a GPU that is roughly 30 percent smaller than the Cypress GPU found in the Radeon HD 5850 and 5870, and 50 percent larger than the Juniper GPU found in the 5770. The GPU still has a full 256-bit memory interface and 32 render back-ends. Though there are fewer shader units and texture units, the clock speeds are a lot higher, so actual performance isn’t too negatively affected. If all that is Greek to you, here’s the takeaway: On paper, the new Radeon HD 6850 looks as though it’s going to be a lot faster than the 5770, and the 6870 will nearly equal the performance of somewhat more expensive 5850.
At first, the naming of these new Barts-based graphics cards seemed odd to us. The 6870 will be slower than the 5870, and the 6850 will be slower than the 5850. Usually, cards of a new generation in the same series are faster, not slower, than their predecessors. What’s more, you’ll still be able to buy 5800-series graphics cards for a while, as they occupy shelves alongside less-expensive 6800-series cards. So why the confusing 6800-series moniker, rather than calling these cards the 6750 and 6770? The naming makes sense when you consider three things. First, the Radeon 5700 series will continue to be produced and sold for some time, occupying the sub-$150 price bracket. Having both the 5700 and a new 6700 series on the market at the same time would be confusing. Second, the new cards are coming in at a price that AMD has historically put the x800 models at (the Radeon 3800 and 4800 models fell between $200 and $300 when they were introduced, while the 5800 series was a high-priced anomaly). Last, the 5800 series may not be long for this world: The next GPU in the 6000 series that AMD will release should displace those cards on the high end.
Be sure to read our individual reviews of the Radeon HD 6850 and Radeon HD 6870.
Next: What’s New in the 6800 Series
New Features
To make a long and quite technical story short: Barts is a lot like Cypress, but scaled back a bit, with some of the difference in performance made up for by higher clock speeds. These cards have a few other neat tricks, too, however.
The Barts GPU that powers the Radeon HD 6850 and 6870 is very similar in design to the Cypress GPU that powers the Radeon HD 5850 and 5870. There’s a command processor on the front end, connected to a set of graphics engine features (the tesselator, geometry setup, rasterizer, and so on). This has been tweaked in the case of the Barts GPU to offer somewhat improved tessellation performance, but it is otherwise extremely similar. The texture and shader engine is arranged in much the same way, with a thread dispatch processor and scheduler connected to an array of SIMD shader units. Each SIMD has 16 blocks of five-way vector units, for a total of 80 processing units. There are 14 of these SIMD units in the chip, for a total of 1120 processing units. There were 20 SIMDs in the Cypress chip, for a total of 1600 processing units. The Radeon HD 5850 had a scaled-back version with two of the SIMDs disabled, just as the new Radeon HD 6850 has two of the Barts GPU’s SIMD units disabled.
UVD3: The new models will offer hardware acceleration of a greater variety of video formats. AMD calls the video decoding engine in its GPUs “UVD,” and the Radeon 5000 series featured the second-generation UVD2 engine. The 6800 series introduces the third generation, UVD3. New to this generation of the engine is enhanced quality, the ability to decode the Multi-View Codec used by Blu-ray 3D discs, better support for MPEG-2, and support for MPEG-4 Part 2 (used in the DivX and Xvid codecs). All this is above and beyond the VC-1 and H.264 decoding features found in UVD2.
DiplayPort 1.2 and HDMI 1.4a: The Radeon HD 6850 and 6870 are the first graphics cards to offer a couple of new display output standards. The DisplayPort outputs have been upgraded to the new DisplayPort 1.2 specification, which supports higher resolutions and allows users to daisy-chain multiple monitors from a single output. This capability renders six-port Eyefinity 6 editions all but obsolete, especially when monitors supporting DisplayPort 1.2 become common. The HDMI output has been upgraded to version 1.4a, which supports higher resolutions and color depths in addition to full 3D support in hardware.
New Display Output Arrangement: The standard array of display outputs have been upgraded over previous Radeon cards. The reference for the Radeon HD 6850 and 6870 is two Mini DisplayPort outputs, a single HDMI port, and two DVI ports (one single-link and one dual-link). Individual card vendors may stray from this standard, but most of them probably won’t.
Hardware Per-Display Color Correction: A small but important feature for multiple-monitor users, the new display controller allows color gamut correction per-display in hardware, so you can color-match different displays with different resolutions and timings without a performance penalty.
HD3D: This is AMD’s brand for its support of the Open Stereo 3D Initiative, which it announced earlier this year. The idea is to enable 3D support through a wide array of methods (active-shutter, polarized glasses, glasses-free, and more) with various monitor and TV manufacturers. Support is relatively slim at this point, but iZ3D and DDD support the technology today and can enable 3D in dozens of games and other applications, with more on the way. The Radeon HD 5000 series will get some of this support through a driver update, but only the new 6000 series cards have fully hardware-accelerated 3D Blu-ray capability and the support for 3D features of HDMI 1.4a.
Next: Let’s Talk Performance
Performance: Synthetic Benchmarks
Enough of the technical talk–let’s look at the new cards’ performance. The cards will be priced very attractively, with Radeon HD 6850 cards selling for around $179 and Radeon HD 6870 cards selling for about $239, depending on the exact vendor and model. This fits them neatly between the $149 Radeon HD 5770 and the $279 Radeon HD 5850 (prices on the Radeon HD 5850 only recently came down to this level).
We didn’t have the proper nVidia-based cards on hand in time in time to compare them with AMD’s new cards, but we should still be able to get a good idea of the relative performance. We performed all of our benchmark testing on a system equipped with an Intel Core i7 980X CPU and 6GB of RAM running 64-bit Windows 7.
Let’s start with the Unigine Heaven benchmark, a synthetic test of a real DirectX 11 game engine, currently licensed by a number of smaller games. The test is rather strenuous and forward-looking, featuring high detail levels, dynamic lighting and shadows, and lots of tessellation. (Tessellation is a DirectX 11 feature that involves breaking up blocky, low-polygon 3D models into smooth, high-polygon models.) We ran the test at the middle “Normal” mode. Apparently, the high clock speeds and improved tessellation engine are paying off, because the sub-$200 Radeon HD 6850 beat the $279 Radeon HD 5850 at all resolutions, and the 6870 did even better.
FurMark is a synthetic OpenGL-based test that renders a torus (a geometric shape that resembles a doughnut) covered in fur. It’s rather simple, but no test we’re aware of stresses out a GPU more thoroughly. It’s a great way to see just how hot your graphics card will get, and how much power it will use. The 6850 was dramatically faster than the 5770 here, and surprisingly, it even matched the 5850 when antialiasing was enabled. The new 6870 managed slightly faster performance than the 5850, even though it costs about $40 less.
Though it’s getting a little long in the tooth, 3DMark Vantage is a standard still commonly used in synthetic graphics benchmarks. The engine uses DirectX 10 only, though a new version of 3DMark geared for DirectX 11 should be coming soon. We present the 3DMark scores with standard settings for the “High” and “Extreme” profiles. While the 6850 ran about 30 percent faster than the 5770, it couldn’t catch up to the 5850. No matter; the 6870, which is less expensive than the 5850, beat it handily.
Next: Real Game Performance
Performance: Games
Synthetic tests can be useful for evaluating features that will be common in tomorrow’s games, but performance in real games is far more important. We tested five modern games that can push a modern graphics card to the limit.
Codemasters’ rally racer Dirt 2, one of the first DirectX 11 games, features an excellent built-in benchmark. We use the demo version (whose benchmark track differs from the track in the retail game), so you can run the game at home and compare your results. We enabled DirectX 11 and turned all of the detail levels up to full. The Radeon HD 6850 was dramatically faster than the 5770 despite costing only about $30 more; the Radeon HD 6870, meanwhile, managed to run a few frames per second faster than the more expensive 5850.
Tom Clancy’s H.A.W.X. is a graphically rich arcade flight game that uses DirectX 10.1 to enable features such as Screen Space Ambient Occlusion (SSAO), God Rays, and Soft Particles. Again, we turned all of the detail levels up to the max for our testing. The results: This was the first test where the Radeon HD 5850 actually outperformed its less-expensive successor, the Radeon HD 6870. Still, the performance numbers were very close, and both of the new cards delivered more than 60 frames per second even at 1920 by 1200 with antialiasing enabled.
World in Conflict is aging a bit, but it’s still a beautiful real-time strategy game with a DirectX 10-based graphics engine that can stress all but the most powerful graphics cards when you maximize the detail levels, as we did. The new cards performed exceptionally well here: The 6850 performed on a par with the 5850, which costs about $100 more; and the 6870 cruised along about 14 percent faster.
The S.T.A.L.K.E.R. series has always been on the leading edge of graphics technology. We used the demo benchmark for the Call of Pripyat sequel with DirectX 11 lighting enabled and all detail settings maximized. The scores shown here are for the “Day” benchmark run.Again the new Radeon HD 6850 blew away the 5770 for only about $30 more; and though the Radeon HD 5850 outran the 6850, the 6870 was the fastest of all by narrow margin, even though it costs less than the 5850.
Last but not least, we used the excellent benchmark built in to Just Cause 2. We maximized graphics settings and ran the “Concrete Jungle” test, which is the most strenuous of benchmarks. Once more, the Radeon HD 6870 came out on top, followed by the more expensive 5850. The affordable Radeon HD 6850 finished just behind the 5850, but ahead of the 5770 by a wide margin.
Next: Value and Efficiency
Value and Efficiency
How do these two new cards compare to AMD’s nearest kin–the less expensive Radeon HD 5770 (roughly $149) and the more expensive Radeon HD 5850 (around $279)? To find out exactly how much bang for the buck these new cards offer, we averaged the benchmark results for all of our real-world game tests and then divided by the price to arrive at a metric we call Dollars per Frames per Second. On this chart, lower numbers are better: The less you have to spend, the better. For this computation we used AMD’s suggested prices of $179 for the Radeon HD 6850 and $239 for the Radeon HD 6870. Prices from individual brands may vary a bit, especially for factory-overclocked models, but they should fall roughly in line.
The closest direct comparison from nVidia’s lineup would have been the 768MB and 1GB versions of the GeForce GTX 460, but AMD’s launch has prompted a last-second price reduction from nVidia on the 1GB GeForce GTX 460 (down from $269 to around $219) and the faster GeForce GTX 470 (down from over $300 to $259). These price drops can’t be good for nVidia’s bottom line: The GeForce GTX 460 is based on a GPU (the GF104) that is 44 percent larger than AMD’s Barts chip, and the GTX 470 is based on a GPU (the GF100) that is over twice as big. It certainly makes purchase decisions interesting, though.
It appears that the Radeon HD 6850 provides exceptional bang for the buck, even surpassing the value-oriented Radeon HD 5770. At $239, the Radeon HD 6870 is a far better deal than the Radeon HD 5850 (whose price just dropped), offering similar or superior performance at a lower price.
AMD says that it worked hard to optimize power utilization on these new cards, delivering even lower idle power than the already-excellent Radeon HD 5850 and 5870. The Radeon HD 6850 requires only a single six-pin PCIe power connector, while the 6870 requires two. This is an improvement over the 5800 series, but the cards are going to have a hard time matching the low power draw of the Radeon HD 5770. The numbers on the chart below represent total system power as measured at the plug, not just graphics board power. With a high-end CPU like our Core i7 980X, you can expect high overall power numbers.
It’s unreasonable to expect these new cards to match the power consumption of the Radeon HD 5770, which has a much smaller GPU and a 128-bit memory interface, but the Radeon 6800 series impressed us by using about the same amount of power at idle and 20 to 50 watts more under load.
Factoring in the performance of these cards, we can get an idea of how power-efficient they are, not just how much raw power they use. By dividing the average frames per second of all our game tests by the power use under load from the previous chart, we arrive at a measure of Watts per Frames per Second. Here again, lower numbers are better.
Though the Radeon HD 5770 uses less power than any of the other cards, it is a much slower performer. Thus, its performance per watt is much worse than any of the other three cards assessed below. The new Radeon HD 6850 and 6870 offer similar performance per watt to the Radeon HD 5850–but both also use about 25 watts less power at idle.
Next: A Slam Dunk for AMD
A Slam Dunk for AMD
It looks as though AMD has come up with a pair of winners with the Radeon HD 6850 and Radeon HD 6870. In its performance in modern games, the $239 HD 6870 model was slightly faster on average than the $279 Radeon HD 5850. Given that the 5850 cost about $300 until recently and was considered quite a good buy, the new card represents a fantastic value. Amazingly, the Radeon HD 6850 is an even better value. At a suggested price of $179, it was only about 10 percent slower on average in our benchmark tests than 5850, a card that costs $100 more. The HD 6580 is 30 to 40 percent faster than the Radeon HD 5770 in modern games and costs only about $30 more. That leap in performance may be the difference between being able to play a game at its maximum detail and having to settle for lower resolution or detail levels to keep things smooth.
These new cards offer more than just a great mix of price and performance, though. They’re both quite energy-efficient, sipping a very small amount of power when you’re simply sitting at the Windows desktop. (AMD claims that the cards use just 19 watts, and if our total system power measurements are anything to go by, we believe them.) Even under heavy 3D graphics load, the power utilization is quite good, considering the performance you get.
The cards’ other improvements include better video decoding and hardware 3D Blu-ray support from the new UVD3 video engine. DisplayPort 1.2 and HDMI 1.4a support highlight an impressive set of new standard display outputs. AMD has promised further image-quality improvements with these cards, including some welcome tweaks to its anisotropic filtering algorithm and new antialiasing modes, though we haven’t had the opportunity to single out those features for testing yet.
Given their attractive pricing, performance, power utilization, and new features of these cards, the Radeon HD 6850 is likely to be the card to beat at under $200, and the Radeon HD 6870 should be king of the hill at under $250. Though we don’t have any direct comparisons to show at this time, our test results suggest that these cards should generally surpass the GeForce GTX 460 cards that they’ll compete against.
Be sure to read our individual reviews of the Radeon HD 6850 and Radeon HD 6870.