Today, AMD/ATI has released their next-generation graphics platform, the 5800 series of Radeon graphics cards. First to market with DirectX 11 and featuring more than 2.7 billion transistors on a single chip, the new graphics chips are almost inconceivably advanced, not to mention ridiculously powerful.
First, let’s take a look at a couple of the new GPU’s physical stats:
|Manufacturing process||55 nm||55 nm||55 nm||40 nm||40 nm|
|GPU clock||625 MHz||750 MHz||850 MHz||725MHz||850MHz|
|Transistor count||965 million||965 million||965 million||2.15 billion||2.15 billion|
AMD is touting this as a major hardware revision, and they’re right, it is: new (smaller) manufacturing process, higher clock speeds, much higher transistor counts, a new version of DirectX and greater power efficiency than ever before. While ATI’s former high-end cards, the 4850s, 4870s and 4890s of the world were known for being hot, loud and hungry for power, that unending electrical thirst was one of the key things worked on for this revision.
Part of that is down to the manufacturing process – smaller transistors generally means cooler chips that use less power. In this case, while the maximum power usage remains hovering around the 200W mark, the idle power usage is pretty astonishing, at less than 30 watts. It looks like ATI finally caught up to NVIDIA in that arena, whose idle power numbers for the GTX200 chips have been pretty low in their own right.
The new chip behind it all is the RV870, the most complex chip ATI has ever created. At over 2 billion transistors on the core, it really leapfrogged the previous generation’s accomplishments. Raw performance isn’t the only thing ATI decided to add to the mix, though; we recently talked about the new Eyefinity technology they’ve developed. It’s also been added to the new graphics cards, giving users unheard-of amounts of resolution.
The new cards will ship with the ability to hook up three displays each of up to 2560×1600 resolution (that’s the standard 30-inch display resolution, for those curious). A second version will follow later this year that includes the ability to hook up to six DisplayPort-enabled displays onto one card. We saw this in action at a recent ATI press event onboard the USS Hornet, in California: twenty-four HD displays (1920×1080) were connected to a single desktop computer that ran four Radeon HD5870s in CrossFire.
Obviously, it works via a number of adapters that connect to mini-DisplayPort ports onboard the card itself. Fortunately, a single card isn’t quite as messy; while the converters are still there, it’s much easier to see what’s going on. Seeing all of those monitors put together is pretty impressive; while the bezels are still something of a problem, the flight simulator being used to demo the technology felt much more immersive than when using just one or two monitors.
ATI is also working to market the new cards alongside an upcoming monitor being produced by Samsung. The new monitors, which feature a ~7mm bezel to the left and right, do a lot to overcome the divisive feeling multi-display setups usually bring. The gaps are still very much noticeable, but between the thin bezels, and the high-quality cPVA panels, it’s definitely less of an issue.
In addition to the Eyefinity tech inside, ATI is using the new cards as an excuse to relaunch their ATI Stream initiative, their version of GPGPU application acceleration. They’re working to woo programmers currently developing products for NVIDIA’s CUDA by using a more open architecture like OpenCL or DirectCompute. That way, companies can develop one codebase and take advantage of the acceleration features found in cards by both Big Green and Big Red.
So they’re more efficient, more powerful, packed full of eye-catching new technology, but none of that means too much without the final part of the equation: how well do they game? By all initial accounts, very well indeed. ATI made a number of boastful claims over the past few weeks, suggesting how well the cards perform in a variety of recent games. Fortunately, that boasting has quite a bit of substance to back it up. The new 5870 is now the fastest single-GPU solution in the market, soundly beating every other option out there. It even manages to pull away from the dual-GPU GTX295 in a number of benchmarks. Sometime before Christmas, we can likely expect to see the HD5870X2, ATI’s own dual-GPU solution that will almost assuredly become the gold standard for consumer graphics performance. Later in the spring we can expect to see more mainstream variants of the new cards, the 5000-series equivalents of cards like the 4650.
Chances are, if you’ve got a pretty high-end card from the last generation, like the 4890 or higher, it’s probably not worth an upgrade at this point in time. Certainly not until prices die down, with the 5870 coming in at $379.99 and it’s little brother the 5850 not too far behind at somewhere around $319.99. Not that you can buy one yet, anyway; launch-day availability is looking pretty grim, with users struggling to find a single store that has the cards in stock.
If nothing else, though, it looks as if AMD is really coming back to a lean, mean fighting stance; while many have been predicting their demise for some time, they’re really hitting a few out of the park. Their new Athlon II X4 quad-core chips are utterly phenomenal for the price, and combined with affordable 785G motherboards and Windows 7, provide a great opportunity for both system builders and OEMs. Add in a Phenom II x4 and one of these new Radeon 5800 cards, and you’ll have one of the fastest gaming systems on the planet. Hold your heads up high, guys. You’ve done good.