When you're up on top, it's pretty hard to imagine getting knocked back down. Perhaps that's why, after a mind-blowing Radeon HD 5000-series launch, AMD seems to have engaged the cruise control for these first two examples of its Radeon HD 6000-series.
Not that we'd blame the company. It enjoyed a solid six months of selling the world's only DirectX 11-capable product stack at a time when DirectX 11 and, more important, DirectX 11 games were actually shipping. Nvidia's response was compelling. But the heat and power consumption associated with a 3 billion transistor GPU counterbalanced some of its brighter performance highlights.
Only when Nvidia started rolling out derivatives did AMD's position seem truly challenged. The GF104-based GeForce GTX 460 offered the price tag and performance to make us reconsider the Radeon HD 5830, and the GF106-based GeForce GTS 450 was at least good enough to lock horns with AMD's Radeon HD 5770, even if prior-generation cards still offered (and continue to offer) better performance for your dollar. Interesting side-note: one of Nvidia's board partners lets us know earlier this week that G92 is officially dead. Supplies of GeForce GTS 250 should start drying up soon, leaving you to pick and choose between the current crop of DirectX 11 cards.
We know both of these companies are engaged in a brutal battle. In fact, that battle made the decisions in today's review very hard to make. First, we hear that we should be comparing the 6000-series boards to factory-overclocked GeForce GTX 460s because "they outnumber the reference-clocked boards." Then it's, "...and prices on the GeForce GTX 470 and 460 are going to be dropping; we just can't tell you to what level yet." AMD knows Nvidia doesn't have a target to aim for yet, so it holds back on pricing details on its new cards. When it can wait no more, that email lands. Less than a day later, Nvidia announces its own official price restructuring. Hooolllyyy...talk about corporate espionage enabled by wannabe journalists who can't keep email to themselves!
And in the midst of all of that jockeying, there are new games launching that may or may not be under the influence of developers who selectively cooperate with one GPU vendor or the other. These are anticipated games. Games we've wanted to test for some time now. But we face the possibility that one hardware architecture might be highly-optimized, while the other company's driver team still hasn't seen the title running. Now there's a recipe for hard-to-explain benchmark results.
What's the point? Today's DirectX 11-class graphics market is more competitive than anything we could have imagined one year ago, when AMD was undisputed king of the hill and Nvidia's GeForce GTX 295 was still the flagship. Naturally, then, when you hear that AMD is launching its Radeon HD 6870 and 6850 cards, you expect the next generation of high-end--a follow-up capable of knocking GeForce GTX 480 off of its perch, perhaps.
Not today. The potential for such an evolution will have to wait until next month. The Radeon HD 6870 is slower than Radeon HD 5870. Radeon HD 6850 is slower than Radeon HD 5850. It's confusing, we know, but AMD has what it considers a good explanation for the naming scheme.
And while raw performance is down, overall, the purpose behind AMD's Radeon HD 6800-series is purportedly an optimization of the architecture. The "Barts" GPUs realize a re-balancing of the Cypress design that performed so well already. A handful of features are being added, and price points are coming down. The idea here is to engage Nvidia's GeForce GTX 460 1 GB and 768 MB beyond performance.
Before we dig-in to the Radeon HD 6800-series, let's take a closer look at the targeted price points.
What's With That Name?
Now, if you're like us, that Radeon HD 6800-series moniker will strike you as disingenuous. Even after hearing the official party line, we still don't like the fact that the branding requires an explanation from us in order to make sense. What about the folks who don't get the memo? We can only hope that price insinuates performance. Barts is designed to fill the $150 to $250 range, far below today’s Radeon HD 5870. This is more like Radeon HD 5830 and 5850 territory. The high-end Radeon HD 5870 and 5970 will be replaced by the “Cayman” and “Antilles” Radeon HD 6900-series before the end of Q4 2010.
I’m sure we aren’t the first to be surprised by the new naming scheme—to us, it’s a cinch that Barts should file in as the Radeon HD 6700-series. AMD claims that 6800 was chosen because the Radeon HD 5700s will remain in production for some time to cover the sub-$150 market. We honestly don’t think this is a very good justification, as product generations have overlapped time and time again without too much of a problem. The biggest issue for us is that the ill-informed Radeon HD 5870 owner will assume that the Radeon HD 6870 is an upgrade, when in fact the new card wields less performance.
But we're not here to review the card's name. We'll voice our dissent and move on. The Radeon HD 6870 promises Radeon HD 5850-class performance at roughly $240. The Radeon HD 6850 should slide in ahead of the Radeon HD 5830 for $180 or so. Both new cards also do a handful of things the 5000-series couldn't do, including Blu-ray 3D acceleration and playback, stereoscopic 3D gaming, a new level of anti-aliasing, faster tessellation, and a beefed up version of Eyefinity that lets you connect six displays, just as soon as the DisplayPort 1.2 ecosystem fills out sometime in 2011.
In the past, we've seen a refresh to the Radeon lineup on a yearly basis, accompanied by a process improvement. But the Radeon HD 6800-series is being delivered using the same TSMC-based 40 nm node first used in April of 2009 to manufacture the Radeon HD 4770. Forty nanometer technology was also deployed across the Radeon HD 5000-series. And now it's being recycled for the Radeon HD 6800s. According to Eric Demers, CTO of AMD's GPU division, sticking with the established process made more fiscal sense this time around. Based on yields, the company essentially had to decide between two 32 nm processors or three 40 nm GPUs at a similar cost. From here, though, we're expecting AMD to skip 32 nm and transition straight to 28 nm manufacturing.
With no process improvement of which to speak, making Barts a worthwhile step forward becomes a more challenging task for the GPU architects.
We expect that you'll have a tough time telling the difference between AMD's Radeon 5800- and 6800-series GPUs on this schematic. In fact, if you squint and count the SIMD engines, the new Barts GPU looks notably weaker than the Radeon HD 5870's Cypress. Of course, performance is all in the numbers:
|Radeon HD 5850||Radeon HD 5830||Radeon HD 6870||Radeon HD 6850|
|Core Clock:||725 MHz||800 MHz||900 MHz||775 MHz|
|GDDR5 Memory Clock:||1000 MHz||1000 MHz||1050 MHz||1000 MHz|
|Memory Bandwidth (GB/s):||128||128||134.4||128|
|Die Size:||334 mm2||334 mm2||255 mm2||255 mm2|
|Maximum Power:||151 W||175 W||151 W||127 W|
|Idle Power:||27 W||25 W||19 W||19 W|
The Radeon HD 6870 employs 14 SIMD engines, each hosting four texture units and 16 stream processors. Each stream processor is armed with five ALUs (AMD calls these stream cores). As a result, this GPU has a total 1120 stream cores and 56 texture units. The GPU channels output through four render back-ends, each containing eight color ROP units, resulting in an effective total of 32 ROPs. Four 64-bit memory controllers yield a 256-bit aggregate memory interface.
If any of this sounds familiar, it’s probably because, from a raw specifications standpoint, the new Radeon HD 6870 is essentially a Radeon HD 5830 with twice the render back-ends. Or, viewed differently, the Radeon HD 6870 has the same back-end as the 5870, just fewer shader cores. When the Radeon HD 5830 was released, we lamented the merciless halving of ROPs, and in a way, the Radeon HD 6870 shows us what the 5830 could have been.
But there’s one more critical component to performance, and that’s clock rate. Despite the re-use of 40 nm manufacturing, at 900 MHz, the less-complex Barts GPU runs much faster than the 5830, or even the 5850. This means geometry throughput is almost 25% higher than the 725 MHz Radeon HD 5850 because it’s limited to one primitive and one vertex per clock cycle. At the same time, the 6870’s texture unit and ALU deficit (compared to the 5850) is offset by a higher core clock, resulting in roughly the same overall performance.
At the end of the day, this means the Radeon HD 6870 performs a little better than the Radeon HD 5850, but does it with about 25% less silicon, purportedly reducing idle power and overall cost--two definite benefits for the end user. The Radeon HD 6870's memory runs at 1050 MHz, slightly faster than the 5850's 1000 MHz. The MSRP of the Radeon HD 6870 is $240, a bit less than the Radeon HD 5850, which still sells online for $260 and up (though rebates take it lower).
The second product based on AMD's RV870 GPU is the Radeon HD 6850. The only difference here is that two of the GPU’s 14 SIMDs are disabled, resulting in 12 SIMDs with a total 960 stream cores and 48 texture units. The core clock drops to 775 MHz, but all of the render back-ends are fortunately left intact. The resulting performance should be notably faster than Radeon HD 5830. AMD’s suggested retail price of the Radeon HD 6850 is $180, and while Radeon HD 5830 cards can sometimes be found for that price, the majority of them are in the $200 range. More threatening to the Radeon HD 6850 will be an onslaught from Nvidia, which we'll also dissect in more depth.
On the surface, both Barts-based SKUs appear to be (at least architecturally) derived wholly from Cypress. But AMD's engineers assure us that notable improvements have been made, and features added. Perhaps the highest-profile performance-adder is an enhanced tessellation unit. According to the company, improved thread management and buffering yields between 1.5 and 2x performance in what it considers the most important (real-world) range, below 14 tessellation factors. We can see the result of this in the Unigine Heaven benchmark:
The first chart shows the benchmark results with no tessellation applied, and the second demonstrates results with the normal tessellation level enabled. When tessellation is used, we can see the Radeon HD 6850 reach Radeon HD 5850-class performance, and the Radeon HD 6870 leaves the other Radeons behind.
This is a particularly sensitive subject for AMD. When Nvidia launched GF100, it went straight for AMD's throat, claiming the single fixed-function tessellation unit would not scale well. Instead, Nvidia advocated its Polymorph Engine, present on each of its architecture's Shader Multiprocessors. Parallelizing geometry, it claimed, was the key to adding the next level of realism. AMD counters that geometry is actually better-addressed through its single unit, and that its approach is much more efficient.
At the end of the day, Nvidia still outperforms AMD in synthetic measures of tessellation. But we still haven't seen a game capable of coming anywhere close to giving Nvidia an advantage due to its geometric processing potential. From what we hear, Nvidia took Ubisoft to bed, and the result, HAWX 2, employs what amounts to a worst-case geometry scenario for AMD. That might turn out to be the first example of Nvidia's advantage, even if it was sponsored. It remains to be seen whether AMD can get the developer to add a slider for geometry detail. At least, that's the current plan, according to company representatives.
After watching Nvidia build its patented and proprietary 3D Vision empire (the company recently announced that 1000 products are in its ecosystem), AMD is finally engaging the stereoscopic market with its Radeon HD 6000-series GPUs. Up until now, the company has downplayed the significance of and market readiness for this technology. So it's a little surprising to see how much thought has gone into stereo 3D in preparation for the Radeon HD 6800-series.
HD3D is the brand under which AMD is filing its stereoscopic technologies. The company takes a very different approach from Nvidia in that there are no proprietary glasses or specific display technologies associated with HD3D. This presents some interesting advantages and disadvantages compared to Nvidia’s approach.
Unlike 3D Vision, AMD’s HD3D technology doesn't require a specific display with an AMD-licensed technology built in. Instead, you only need a consumer 3D display that supports HDMI 1.4—a 3D television for example—and plug it into your Radeon HD 6800 series graphics card. From there, it’s up to the software. If you want to play a Blu-ray 3D disc, you’ll need to buy Blu-ray 3D playback software, such as CyberLink’s PowerDVD 10 Ultra or Arcsoft’s TotalMedia Theatre 3. If you want to play games in stereoscopic 3D, you’ll need DDD’s Tridef driver or iZ3D’s driver. With the software installed, your display handles whatever 3D method it’s designed for, whether it uses passive polarized glasses, active shutter glasses, or even glasses-free solutions that might emerge in the future.
AMD demonstrated a number of games and Blu-ray 3D movies at the launch event, using both active and passive glasses technology, but I was curious to see whether or not HD3D is ready for the consumer. A friend of mine has a 120 Hz 3D television, so I paid him a visit to try it out. Using an updated version of PowerDVD 10 Ultra Mark II, I was able to play the Blu-ray 3D movie Cloudy With A Chance Of Meatballs with no fuss whatsoever—it just worked. After this, I installed DDD’s TriDef driver for 3D gaming. Once again, it worked. I expected a lot more of the hassle that generally accompanies an open initiative.
But it’s also important to recognize the limitations of HD3D. First, there are no dedicated 3D monitors in North America that employ the HDMI 1.4 standard—all of the available 3D monitors at the moment are Nvidia 3D Vision-exclusive, and will not work with the Radeon HD 6800-series cards.
The good news is that this will probably change now that AMD offers an alternative. But it will take some time. In Europe, the Viewsonic V3D241wm-LED is already available for £330 (about $530 USD) and these were out in force at the launch event. The harsh reality, in the near future at least, is that North Americans who plan to rock a Radeon HD 6800 card on a stereoscopic 3D monitor will have to wait for the displays to arrive. If, on the other hand, you own a commercial 3D television, you can jump right in, though.
In addition, while DDD and iZ3D’s stereoscopic gaming drivers can work perfectly in some situations, our experience has shown us that Nvidia’s 3D Vision driver works better and more consistently in the majority of games. Integration with the GeForce driver also allows for quicker adoption of newer features like DirectX 11, something the other driver developers have traditionally struggled with.
There's one more limitation to bear in mind. Because AMD utilizes the HDMI 1.4a specification, which boasts a maximum TMDS throughput of 10.2 Gb/s, you can either game in stereo at 720p maxing out at 60 frames per second per eye, or you can game at 1080p with up to 24 frames per second per eye. That's actually pretty severe, considering we've been playing around with 5760x1080 using 3D Vision Surround and dual-link DVI connectors (each display running at 1920x1080). AMD says it'll transcend the shackles of HDMI 1.4a next year sometime when monitor vendors begin incorporating DisplayPort 1.2. A peak effective bandwidth of 17.28 Gb/s is enough to enable 1080p at 60 frames per eye.
It’s important to mention that Nvidia can also handle stereoscopic 3D on commercial displays over HDMI 1.4a with the release of its 3DTV Play driver (perhaps not coincidentally timed for today's launch). This driver is now available as a free download for 3D Vision owners, and will be bundled in a number of products, such as the XPS laptops from Dell. If you want to enable 3D playback on your compatible television, but don't want to buy the 3D Vision kit, the 3DTV Play should be available for purchase from Nvidia.com by the end of November for $39.99. Keep in mind that this driver will work only on newer GeForce models that can handle the HDMI 1.4a standard, such as the GeForce GT 220, GT 240, and the GeForce 400-series. Also note that, if you're using 3DTV Play over HDMI, you'll suffer the same resolution/frame rate limitation as AMD's Radeon HD 6800-series boards. Only by switching to a dual-link DVI output can you overcome that.
With this information in hand, let’s consider costs. If you want to watch Blu-ray 3D, you’re going to have to pay for Blu-ray 3D playback software, regardless of the graphics card. With an Nvidia card, you need to purchase the $40 3DTV Play driver from Nvidia at least or a $200 3D Vision kit at most. The interesting part is that AMD has actually bypassed any proprietary expense, so Blu-ray 3D playback is that much cheaper on Radeon cards.
Having said that, Nvidia includes the stereoscopic game driver with 3DTV Play and 3D Vision, while AMD hardware requires a third party game driver from DDD or iZ3D. These typically cost in the neighborhood of $50. Aside from cost, Nvidia’s driver solution provides a much smoother overall experience than its competitors--at least right now. Hopefully the extra revenue generated from Radeon HD 6800 owners who invest in these third-party drivers will result in faster development and better results.
Bottom line: with the Radeon HD 6000-series and HD3D, AMD is now offering a viable alternative in the 3D stereoscopic race. For HTPCs attached to commercial 3D televisions, AMD and Nvidia offer surprisingly similar functionality for Blu-ray 3D playback. When it comes to desktop monitor availability and game driver compatibility, Nvidia has a definite advantage, although it no longer enjoys an unassailable position.
Morphological anti-aliasing (AA) is an all-new option for the Radeon HD 6000-series cards. It presents a different approach to the aliasing problem in that it needs no insight into the makeup of the scene’s geometry; morphological AA is a post-process filtering technique, accelerated with DirectCompute and compatible with any application from DirectX 9 to 11 (in theory). After a frame is rendered, it is passed through the morphological AA shader that looks for high-contrast edges and patterns consistent with aliasing. It then blends the colors of adjacent pixels to approximate a smooth transition along a line instead of aliased steps. This means that the smoothing effect isn’t limited to the edges of geometry or alpha textures like CFAA; it applies to all of the pixels in the scene.
Conceptually, this method promises results similar to super-sampling, but with performance comparable to edge-detect AA. AMD suggests that some applications will look better than others, and that the technique is not ideal for all scenes and games. So, we tried it ourselves to see what the actual result looks like.
Morphological anti-aliasing works independent of regular anti-aliasing settings, so controlling it might seem confusing at first. Even with the “Use application settings” checkbox enabled, the feature should work, so long as it’s enabled. It can even be used in conjunction with regular AA modes, if you want.
With morphological AA enabled, we fired up Battlefield Bad Company 2. The post-processing filter didn’t seem to make any difference at all. Careful checking verified that, indeed, it does not seem to work with this title. So much for automatic compatibility with all DirectX 9 to 11 games.
This brings us to StarCraft 2, a title in which we're quite interested due to the fact that Radeons are currently known for slow AA performance in this game. With morphological AA enabled, the results are visually obvious, and yet performance remains almost as fast as it is with the feature turned off (roughly twice as fast as it is with 4x multi-sampling, and about four times as fast as 4x super-sampling). This is a much more impressive result.
Performance can only be as impressive as the image quality however. And while morphological AA does smooth out the edges, it can have some less-than-ideal effects on the appearance of the scene. I noticed a little crawling on the edges of moving objects--a very slight shimmering. This is not surprising, as the post-process effect does not have access to geometry details. It’s doing its best on a frame-by-frame basis, but there is likely no temporal information stored to smooth edges between frames. The effect becomes less noticeable as the resolution increases, as with most aliasing artifacts.
The worst-case scenario for morphological AA is low resolutions combined with text. The effect appears to work on a per-pixel basis, which makes sense. But the unfortunate side effect is that, since there are fewer pixels in a low-res scene, there can be undesirable smoothing on things that you’d prefer would stay sharp. Text is a prime example of this. The smoothing isn't as apparent at higher resolutions. Here are some examples of what I mean:
It’s important to note that because morphological AA is a post-process effect, the resulting output can’t be captured with a regular screen capture utility like FRAPS. The comparison images above were created with an application that AMD provided. It uses the same code in the driver to modify the output, just as morphological AA does when running a game.
In any case, morphological AA is an interesting addition to the existing AA tool set. Folks running high resolutions may find themselves enabling the feature on almost everything because of its low performance impact. Users with displays limited to lower resolutions might want to consider regular AA modes due to the visual artifacts on text. The good news here is that more conventional AA techniques tend to be more playable in low-resolution environment anyway.
For the most part, Eyefinity as a feature is largely unchanged in the shift to AMD's Radeon HD 6800-series. That makes sense. After all, it's still uncontested by Nvidia, which can only enable two simultaneous display outputs, even on its highest-end boards.
DisplayPort 1.2 is one capability that AMD did add, enhancing bandwidth over the interface, integrating support for high-bit rate audio alongside video, and defining compatibility with the next generation of 120 Hz 3D monitors. Moreover, DisplayPort 1.2 is able to drive multiple displays from each output. That's right. DisplayPort displays can now be daisy-chained. That is to say a single DisplayPort output can use a Multi-Stream Transport (MST) hub to split off into multiple DP/VGA/DVI/HDMI displays of your choosing. Special Eyefinity edition cards are no longer necessary for six-monitor use.
Note that the reference Radeon HD 6800-series outputs include two DVI outputs, an HDMI output, and two mini-DisplayPort outputs. While the number of supported displays has increased, there has been a trade-off in that only one DVI output is dual-link. The DVI output below the exhaust port is a single-link limited to 1920x1080 resolution. In return, you're able to directly attach four monitors today: two DisplayPort and two using a combination of DVI and HDMI.
There are some interesting Eyefinity features coming down the pipeline in the future, including a 5x1 portrait display group that would fill the player’s periphery. The possibility of combining Eyefinity display groups and HD3D is also something AMD staff hinted at.
Eyefinity remains one of AMD's most interesting value-adds for gamers and productivity-oriented enthusiasts who use more than two displays. Nvidia requires an SLI configuration to enable its Surround mode. The fact that we can use triple-display configurations (now quad-display) from an AMD card is outright compelling.
The Radeon HD 6000-series cards include AMD’s third-generation Unified Video Decoder (UVD) hardware. Building on the UVD1 and UVD2 feature set, the new version adds acceleration for MPEG-2 bitstream, MPEG-4 Part 2 Advanced Simple Profile , and the Multi-View Codec (MVC) decode to the mix. Remember that the MVC codec is used for Blu-ray 3D playback, an important feature for the Radeon HD 6000s. Simply, UVD3 accelerates, in hardware, additional steps in the decode pipeline that weren't being sped-up before.
How will AMD's UVD3 measure up to the fixed-function decode hardware built into Intel's upcoming Sandy Bridge-based CPUs? It's still early to tell, but we have it on good authority that Intel put a lot of effort into that particular block of logic, and that both AMD and Nvidia are going to be way behind in this regard. This would coincide with what we experienced first-hand from Intel at IDF 2010. Naturally, that'll be a shock to many people--ATI, back when it was ATI, was routinely known for its uncontested video playback quality and performance.
AMD APP Technology
ZoomATI Stream is dead. We're guessing that name got dragged around for too long while general purpose GPU-oriented development languished. It was time for something new. Meet AMD Accelerated Parralel Processing (APP).
Yes, the brand was renamed to a more descriptive term. But aside from that, there isn’t much to address in today's story. There are certainly some interesting applications in the pipeline, and we’re hoping to get our hands on some nifty OpenCL- and DirectCompute-based applications to test in the future. Perhaps the most relevant demonstration of APP we saw at the launch was use of the Bullet physics library to fracture and destroy geometry in real-time. Then again, we were hearing about Bullet a year ago. Before that, Stream was the topic of discussion at the Radeon HD 4870 launch.
A muscle car on slicks in the snow could have found traction by now, and yet the tenets of Stream continue to prove amorphous at best. This is something that AMD is going to need/want to figure out soon. We're staring down the barrel of its Fusion initiative, so the timing couldn't be any better for apps to start materializing with more force. We're waiting.
To compare raw compute ability with existing products we do have the results of a Collatz@Home distributed computing application test to examine:
With the Radeon HD 6800-series cards sharing essentially the same architecture as the 5000s, the results aren’t exactly surprising. As expected, the combination of fewer stream cores combined with a higher clock rate brings the Radeon HD 6870 performance very close to the Radeon HD 5850.
AMD is proud of its angle-independent anisotropic filtering method, and the Radeon HD 6000-series takes it a step closer to crystal-clear by smoothing the transition between filter levels. Eric Demers clarified that, despite some concerns about the Radeon HD 5000-series’ filtering, the new 6800s use the exact same angle-independent linear LOD as its predecessor, and the problem was limited to uneven shifting of kernels within a MIP level. This has been improved with better transitions in the same MIP layer.
The size of the reference Radeon HD 6870 is very similar to the reference Radeon HD 5850 (the card in the middle). Both models are about 10 ¼ inches long. The Radeon HD 6870 is slightly longer, but it appears even more so because the Radeon HD 5850’s cooler is tapered off at the end. The Radeon HD 6870 weighs in at 1 lb., 16 oz.--two ounces more than the reference 5850.
The Radeon HD 6870 requires the same two 6-pin auxiliary power cables, but the placement is much more appropriate (on the top of the card, instead of the rear). Note that the Radeon HD 6870 only has a single CrossFire bridge, limiting the card to two-way configurations. The Radeon HD 5850 reference card had two of these bridges, enabling triple- and quadruple-card setups.
AMD arms its Radeon HD 6870 reference card with two DVI, one HDMI, and two mini-DisplayPort outputs. As we’ve mentioned, the DVI output beside the exhaust port is a single-link output, while the other is dual-link-capable.
MSI’s take on the Radeon HD 6870 is closely based on AMD's reference design. It comes bundled with two Molex-to-PCIe power adapters, a mini-DisplayPort-to-DisplayPort adapter, a DVI-to-VGA adapter, a manual, the user guide, a driver/utility CD, and a CrossFire bridge. The card is clocked at the reference rates and is equipped with a gigabyte of GDDR5 memory.
The HIS Radeon HD 6870 comes equipped with a CrossFire bridge, dual Molex-to-PCIe power adapters, a DVI-to-VGA adapter, and a CD folder that includes a manual, driver/utility CD, and a HIS sticker. Like all of the vanguard Radeon HD 6870 cards, it is set to reference specs with a 1 GB of memory.
Despite the RV870 GPU at the heart of both cards, the Radeon HD 6850 is much smaller than AMD's Radeon HD 6870. Its reference card is about 9 ½ inches long and the card weighs much less--six ounces more than a pound.
The Radeon HD 6850 needs only a single auxiliary PCIe power cable as a result of its relatively low clock rates, and this probably makes it the most powerful reference card available with one 6-pin connector. Unfortunately the connector is on the back of the card and not the top. Like the Radeon HD 6870, the 6850 has a single CrossFire bridge and is only capable of dual-card configurations.
The HIS Radeon HD 6850 features an output configuration identical to the reference card: two DVI (one single-link and one dual-link), one HDMI, and two mini-DisplayPort outputs.
HIS' bundle includes a DVI-to-VGA adapter, a CrossFire bridge, a Molex-to-PCIe power adapter, and the CD folder including driver/utility CD, manual, and a HIS sticker. This model is set to reference clocks on both the core and GDDR5 memory.
Surprisingly, the HIS Radeon HD 6850 appears to be significantly different from the reference model. It's a little smaller and lighter at just over a pound, with a custom cooler underneath the hood. Moreover, the air intake port is closer to the center of the card rather than offset to the rear.
DiRT 2 appears to put the new Radeon HD 6800 series' tessellation enhancements to good use, allowing them to pass their predecessors and GeForce competition alike.
Just Cause 2 gives an advantage to the GeForce cards, but the new Radeons continue to outperform the 5800-series cards.
Built on the same 40 nm process as their predecessors, the new Radeon HD 6800-series cards should have similar overclocking headroom as their Radeon HD 5800-series predecessors, and that means 1 GHz is an achievement, even with a voltage increase.
The Radeon HD 6850 turns out to be quite the overclocker, which is not much of a surprise when you consider its stock 775 MHz core clock. This card manages to go all the way to 900 MHz (core), which requires the use of MSI's Afterburner utility. AMD's Overdrive utility caps the Radeon HD 6850 at 850 MHz. The card's memory isn't nearly as accommodating, allowing a mere 50 MHz increase.
On the other hand, the Radeon HD 6870 is a relative overclocking miser, restricting stable running to a mere 50 MHz boost (core). This results in a 950 MHz clock speed, quite close to the 1 GHz ceiling we mentioned earlier. It would be interesting to see what a voltage increase would allow. As for the memory, the card was completely unhappy with anything but the stock 1050 MHz speed.
Note that we've included the EVGA GeForce GTX 460 FTW Edition card in these benchmarks to show what a highly-clocked GeForce GTX 460 1 GB can do. Again, we've allowed the Radeon HD 5870 to strut it's stuff as well.
Aliens vs Predator is an ideal benchmark for an overclocking test because its performance is decidedly GPU-limited. In this test, we see a notable gain from the Radeon HD 6850 overclock, but not much from the Radeon HD 6870's best effort.
The EVGA GeForce GTX 460 FTW edition card gives a very strong performance compared to the regular GeForce GTX 460, suggesting that this is, indeed, an overclocked card to be reckoned with.
The $350+ Radeon HD 5870 walks away with an easy win, showing exactly what 1600 shader cores bring to the game.
While our review is pressed short for time, we did manage to run the HQV 2.0 benchmark on the Radeon HD 6870.
We left most of the driver settings at their default values but increased the edge enhancement setting to 35%, to our taste.
Here are the results we observed:
|TEST CLASS 1:|
|Chapter 1: Video Resolution||19/20|
|Chapter 2: Film Resolution||10/10|
|Chapter 3: Overlay on Film||8/10|
| Chapter 4: Response Time||7/10|
|Chapter 5: Multi-Cadence||30/30|
|Chapter 6: Color Upsampling Errors||10/10|
|TEST CLASS 2:|
NOISE AND ARTIFACT REDUCTION
|Chapter 1: Random Noise||20/20|
|Chapter 2: Compression Artifacts||20/20|
|Chapter 3: Upscaled Compression Artifacts||20/20|
|TEST CLASS 3:|
IMAGE SCALING AND ENHANCEMENTS
|Chapter 1: Scaling and Filtering||15/15|
|Chapter 2: Resolution Enhancement||15/15|
|TEST CLASS 4:|
|Chapter 1: Contrast Enhancement||20/20|
|Chapter 2: Skin Tone Correction:||10/10|
As you can see, the Radeon HD 6870 achieved an almost-perfect score. While there is definitely some subjective variables for this test, our results are very close to AMD's 198/210 reported score. No matter how you slice it, the Radeon HD 6800-series is capable of superlative video playback enhancements.
The power usage numbers don't present any particular surprises, although we would have expected lower idle power from the new Radeon HD 6800-series cards. The Fermi-based GeForces are known to be power-hungry, and the results are no surprise. Note that the Radoen HD 5830 reference card we have on hand uses a Radeon HD 5870 PCB; this may account for the relatively high power usage.
GPU temperatures are fairly close across these products. Keep in mind that the GeForce cards we test here do not use the reference coolers.
The Radeon HD 6870 generates more noise than we would have expected, but the non-reference GeForce GTX 470 manages to out-do it. The rest of these cards were a pleasure to live with under load.
What’s in a name? That which we call a rose by any other name would smell as sweet.” It’s obvious that when Shakespeare penned this famous quip, he was foretelling the arrival of AMD’s Radeon HD 6800-series cards. Wherefore art thou, Radeon 6700?
All kidding aside, once you accept the outright-confusing naming scheme, the Radeon HD 6870 and 6850 have a lot going for them. These are the cards that lead AMD’s charge into the realm of stereoscopic 3D, the cards that make six-monitor Eyefinity a much more accessible possibility, the cards that add the first post-processing based AA mode, the cards that bring AMD’s tessellation performance much closer to the GeForce-based competition.
More important than all of those points put together, though, the Radeon HD 6800-series lowers the buy-in for very powerful graphics cards. Last week, the cheapest GeForce GTX 460 you could find was $220—the Radeon HD 6850 offers performance in the same league for an MSRP of $179.99. Not very long ago, the Radeon HD 5850 was priced at $300—and enthusiasts were happy to pay that much! Now, the Radeon HD 6870 ups the ante a little for the low price of $240.
Almost as a validation of the notable price/performance potential of AMD’s new Radeon models, Nvidia very quickly slashed its own prices in a very aggressive fashion. At $200, the GeForce GTX 460 is more powerful than the Radeon HD 6850 and offers some stiff competition. We’re already seeing these prices on Newegg, sometimes even lower with rebates. Even more impressive is the price reduction of the GeForce GTX 470, with at least one model reduced to $260 on Newegg—once again, even lower with rebates. AMD contends that these prices are temporary and will go back up after three weeks. Fine. Buy now while the value is there.
For anyone interesting in upping their game (or buying a Christmas present for a PC gamer), all of these cards are fantastic buys at prices we couldn’t have imagined back when they launched, Radeon and GeForce alike. It remains to be seen whether AMD can deliver the Radeon HD 6800-series in force to supply the inevitable demand, but the company claims that ten thousand cards are already waiting to be sold.
It also remains to be seen if Nvidia can maintain the long-term price war it recently declared. Every single GeForce GTX 470 is equipped with a monolithic GF100 GPU in the 530 square millimeter range. That’s close to twice the size of the Radeon HD 6870’s 255 mm2 die. How long can Nvidia keep up such a numbers-based fight? Not long, we’d guess, if there’s nothing else waiting in the wings. But this sure would be a good time to introduce a card with a fully-equipped GF104 and 384 CUDA cores enabled (Ed.: I can’t comment, but I know something that you don’t, Don).
But enough speculation. The bottom line is that the new Radeon HD 6800-series refreshes excellent performance. It also introduces a handful of notable features at lower prices that what we were seeing previously. It’s hard to complain about that. To misquote the bard, “Now go we in content!”