|
||
|
|
|
              |
|
 |
 |
||||||||||||
|
 |
|
||||||||||
By William Van Winkle |
||||||||||||
PCI Express 2.0 Earlier, we covered the move to PCIe 2.0 in the 7 Series chipsets. Know that the HD 3800 series takes advantage of this by moving to the newer bus signaling. You might think that most apps don’t really need or would use that extra bandwidth, but the fact is that many do. AMD set up a test system with a 2.1 GHz Phenom on a 790FX platform, using a Radeon HD 3850 configured for PCIe 1.1 versus 2.0. Across a range of 3D games running at 1600x1280 or higher, the mid-level HD 3850 with 256MB showed roughly 5% to 25% performance gain just by utilizing the wider bus. This is because in situations when a game produces more data than the display adapter can handle, the overflow spills off into system RAM. In a sense, the situation is analogous to using flash memory to supplement system memory. If you used a larger amount of memory on the video card (or a lower resolution), there would be less need for overflow resources. But not everybody wants to pay the extra for more memory on their cards. Meanwhile, there’s no sign of application reducing anytime soon. So for those who want to keep costs down while having a card better able to meet the demands of today’s and tomorrow’s software, going with PCIe 2.0 accomplishes this essentially for free.
AVIVO HD AND UVD Avivo was ATI’s umbrella name for several image optimization technologies that spanned stream capture as well as post-processing and playback. In terms of today’s HD 3800 cards, we can set aside discussions about video capture. Likewise, very little has been added to image post-processing from the Avivo of the Radeon X1000 series to today’s HD 3800. (That said, keep in mind that the subtle improvements behind Avivo HD are what enable the HD 3800 line to achieve a perfect score in the HD HQV benchmark, which tests the quality of digital video playback.) Instead, the current focus is on high-def decoding and how the GPU can shoulder processing loads from the CPU in order to boost overall system performance. Specifically, the issue revolves around decoding of VC-1 and H.264/AVC content. These are the two main technologies used on HD DVD and Blu-ray discs, just as MPEG-2 was the standard codec for DVDs. VC-1 is a format backed by Microsoft, and when you keep in mind that Microsoft has been one of the key backers of HD DVD, it’s no surprise that, according to High-Def Digest, 87% of HD DVD titles use VC-1 and only 10% use H.264 (which is based on the MPEG-4 standard). In contrast, Blu-ray Disc comes in at 26% VC-1 and 30% H.264. And while we at RAM are a long way from picking a winner in the blue laser format war, the evidence is mounting that the PC industry—-and Apple not least of all—-is backing Blu-ray/H.264 despite Microsoft’s efforts to the contrary. But again, the issue today is about not picking a winner; it’s about providing equal support to both sides. “You hear us talk a lot about HD because it’s really what sets us apart from our competitors, both the blue and green guys,” says AMD’s McNaughton. “From a video perspective, we’re delivering HD video through both VC-1 and H.264 decoding, whereas our competitor doesn’t. So we support Blu-ray and HD DVD. Our competitor, even with their newest part, the 8800 GT, doesn’t decode VC-1. Their feet are all in one camp whereas we’re willing to spend a little more money to be more customer-centric.”
Heads would roll at NVIDIA if this were entirely true. What McNaughton means is that NVIDIA’s PureVideo 2 engine, found in the latest NVIDIA GPUs, doesn’t offload the entire VC-1 decode pipeline from the CPU. PureVideo 2 offloads the motion compensation and inverse discrete cosine transform (iDCT) stages, but the CPU still has to shoulder the bitstream processing/entropy stage of VC-1 decoding. Avido HD’s UVD component handles all of these stages, which is why you often see about a 10% delta in AMD’s favor on lower CPU utilization during VC-1 playback. Also note that Avivo HD includes having integrated HDCP decryption circuitry for each DVI or HDMI port. Given that both the HD 2000 and HD 3800 series support this, it’s safe to say that the era of protected high-def content not being supported on PCs is nearly over. Tied to this is AMD’s nifty innovation of planting an audio controller in the GPU. When an HDMI cable gets connected to the graphics card, AMD’s integrated audio controller takes over for the Windows system’s default controller, routing audio through the graphics card and out to the external decoder/amplifier over the same cable carrying the video stream. In media center settings, this is a brilliant way to reduce cable clutter. POWER EFFICIENCY Customers who care about high-def playback are also likely to care about power efficiency. After all, those who want to enjoy video typically want to hear the movie, not fan noise. One of the biggest criticisms of the HD 2900 XT was that, like several high-end ATI cards of times past, it was a noisy brute. The HD 3800 units remedy that. The HD 3870 specs out at less than half of the peak power and less than 40% of the leakage of its HD 2900 XT predecessor. That’s a lot less heat generated and thus substantially fewer fan RPMs needed to remove that waste.
A lot of this power savings stems from the FV670’s fab shrink. But AMD introduces some other elements to sweeten the pot. You might recall ATI’s PowerPlay power management technology from the Mobility Radeon product line. As with other similar schemes, PowerPlay is able to drop power consumption by adjusting the core and memory clock speeds as well as the chip voltages based on GPU activity and load. There are three states-—intensive gaming, light gaming, and general use—-that you can see in our chart. These numbers were taken on a machine running an Athlon 6400+, 790FX chipset, 2GB of RAM, and Windows Vista 64-bit. Being able to cut power draw by 75% in average gaming scenarios is just mind-blowing, and when discussing “platform power draw,” the benefits of the HD 3800 group should definitely be taken into account. “That platform you use to generate 200 fps in FEAR is just wasting power from the wall that you could be saving,” says AMD’s Ian McNaughton. “The metric for what is and isn’t a good PC has really changed. It’s more about energy and less about balls-to-the-wall performance. Having a Ferrari able to hit 300 MPH in Los Angeles is not relevant. We want now to deliver features to the buyer that they can use and that will give them a better experience. It’s no longer about running a synthetic benchmark faster than our competitor. Our ultimate goal is platform experience.” This comment stands true for the entire Spider platform, not just the GPU element. Apple may be the modern master of marketing the PC as an “experience,” but AMD is making a run at the same mindset. Hopefully, this messaging will resonate with buyers and help them to adopt these latest technologies and all the new possibilities they offer. No one is better positioned than a local reseller to make this happen. RACE TO WIN
Truth be told, Spider is probably the most significant launch AMD has had since the original Opteron. By themselves, the arrival of a new CPU, chipset, or GPU probably wouldn’t make much more than a ripple. But now we’re talking about AMD’s first desktop platform, and that’s a game changer. We’ve already seen Intel dabble in its Extreme CPUs and chipsets with the ability for a desktop platform to offer extra functionality when a single manufacturer’s components are used in tandem. No one should be surprised if and when AMD starts to do the same. Such platforms won’t be proprietary per se-—they’ll still mix and match with third-party products—-but the user will simply have more feature-level incentives to keep everything under one roof. The benefits of buying a platform rather than parts don’t need to be only about features. Resellers should also be aware of more intangible benefits from single-sourcing, such as support and marketing assistance. “If you’re selling Spider as Spider with all of its parts, you’re going to get a better level of support from AMD from a marketing standpoint,” says AMD’s Gary Bixler. “The tools you get from my team to go sell those products—-things like messaging and images, Web materials and banners for your advertising, collateral and things to print, training for your staff, all those things—-will be better at the Spider level. If you pick and choose and do Spider with a leg missing, we’re still going to support that. Absolutely. But you’re not going to have the best marketing tools behind you that you would have had with the full Spider.” Dig through AMD’s channel marketing material now and you’ll find a mountain of resources designed around the Spider platform, including sell sheets, reference guides, partner training, and e-communications, all tied in a weblike fashion around the Spider platform message. For both AMD and its resellers, it’s smarter and more lucrative to push the platform than a single point product. We’ve given a lot of detail in this article about the technological advantages within Spider, but many customers will care less about these and more about the idea of having a cohesive component source behind the machines on which they stake their entertainment and productivity. Buyers intuitively grasp the value of knowing that the core components in a system were all designed, built, tested, and supported by the same team.
Do all of the platform incentives have to be intangibles? No. Distributors are likely to swing in with their own creative sales ideas, and while AMD may not be throwing piles of rebate money at bundle sales, there are still the usual coop and MDF benefits tied to AMD’s partner program, plus you may see some other pushes made along the way. “We may run some value-add promotions,” adds Bixler, “where if you buy the whole platform together, you get something extra. Buyers might get a free game or an upgrade to this or that. You’re a lot more likely to see us do that than offer X dollars off the whole bundle. It’s just not necessary to drive the market for a product there’s been really strong demand for.” Spider poses an interesting predicament for NVIDIA. Nobody disputes that the company makes excellent graphics and core logic chips. Even now, NVIDIA essentially rules the workstation graphics and Opteron chipset spaces. But with the 7 Series as a precedent, how far behind can an AMD server/workstation chipset be? And despite NVIDIA’s slowly escalating murmurs about using GPUs for general processing tasks, it seems unlikely that widespread “GPGPU” adoption will arrive before the single-package integration of CPU and GPU, a la AMD’s “Fusion” processor and whatever Intel has planned for the next decade. Again, we go back to that concept of a two-horse race. Given the competitive advantages of owning a complete platform, each month makes it harder to imagine why anything besides a two-horse race would evolve. And given AMD’s current second-place position in both processors and graphics, it would seem that the company, with its strong new lineup, is poised for the greatest growth. That’s good news for AMD, but it’s potentially even better news for AMD resellers. “We think Spider gives resellers an opportunity to sell a richer mix with AMD,” says AMD’s Bixler. “It’s about higher ASPs. This opportunity with the Radeon HD 3800 line to sell two, three, four graphics cards in one system is great for them. It drives things like HDMI sales and multiple monitors. This whole thing is a great tool for the channel to be able to sell more richer-margin products.” Of course, the surest way to bet on a two-horse race is to play both ponies. Both contestants have their unique advantages, and it would be foolish not to present buyers with whichever proved to be the best fit for their needs. No matter what, 2008 looks to be a radically different year for AMD than 2007, filled with more competitive products and better value for resellers and end-users across the board. Make sure you’re plugged in and capitalizing on the Spider opportunity. |
||||||||||||
 Back to top
|
||||||||||||
Copyright © 2007 RAM Magazine. All rights reserved.
Do not duplicate or redistribute in any form. |
||||||||||||
