Thuban/Leo Final Thoughts

Members of the computer hardware industry have been forced to watch a real-life episode of Keeping up with the Joneses repeat itself over and over for the past several decades. More recently this behavior has become central to the standard operation procedure, and now companies use it to plan their future product strategies. For example, when Intel releases a new CPU, in no time at all AMD produces a similar product at a lower price point. In response, Intel then launches a bigger, better, faster product - to which AMD answers by repeating the process. The same scenario is true for ATI, AMD's graphics unit. If NVIDIA produces a new GPU, AMD comes back with their own at a better price point. It's become such a predictable show, that by this point every manufacturer has shared in the act of conspicuous production.

AMD has become the master of 'keeping up', and Thuban is a perfect example of their desire to deliver Intel-level products with AMD-level pricing. When Intel designed Gulftown they didn't just add two cores and call it done. The 32nm Westmere processor fabrication allowed two additional cores to be added onto a shrunken Nehalem design to form the foundation for Gulftown, which then added additional shared L3 cache to the unit for proper scaling. Thuban simply adds two more cores, and leaves everything else virtually identical to the way it was on their Phenom-II X4 series. This might sound like I'm talking down AMD's six-core Thuban series, but in reality they've just managed to create an Intel Core i7-980X Extreme Edition counterpart for only 30% of the price to consumers.

Of course, AMD didn't end Thuban with the inclusion of two new CPU cores. New to the Phenom-II series is Turbo CORE, AMD's answer to Intel Turbo Boost. AMD Turbo CORE is a hardware-based C-state performance boost that raises the clock speed of up to three processor cores when they are in a resting idle state. In theory this might sound exactly like Intel's Turbo Boost technology, but in reality they're each very different in execution. I'm summarizing here, but where Intel's TB tech can scale core clock speeds and shut down cores as needed, AMD raises the speed of three cores in unison. So what's next, HyperThreading? Probably not, since AMD has long contended that HT is counter-productive processor technology.

On an similar note, AMD has mastered is the art of consumer-friendly platform branding. The AMD Dragon platform really wasn't anything more than a press-friendly way of saying that their 790-series motherboards worked well with ATI Radeon HD 4000-series video cards and Phenom-II processors... all of which were released at approximately the same time. Combined into one synergistic 'Dragon' platform they supported the capabilities of each other, which is to say they were designed to work with each other. That sounds silly, but when you consider that Intel's been busy building incompatible platforms (P55 and X58) for the same consumer group for almost three years now, it begins to sound like a novel approach. The same concept is true for the AMD Leo platform, which combines six-core AMD Phenom-II Thuban processors with their 890-series motherboard and ATI Radeon HD 5000-series video card. The concept is so simple, it just might work.

Forward Thinking: AMD Processors

If you're sold on Intel processor's and would never stray to another brand, it still serves your best interests to cheer on AMD to produce better products. Imagine a world where only one company makes the product you depend on. Think deeply enough on the topic, and you'll realize that a strong market dominance, especially a monopoly, never serves to benefit the consumer. Value-pricing AMD Athlon/Phenom desktop processors is a winning strategy, and as consumers realize there's very little real-world difference the gap narrows, but at the end of the day PC hardware enthusiast are shrinking in number. If AMD is going to continue competing within the same market space as Intel, they'll need to expand beyond consumer-level desktop processors. The Magny-Cours AMD Opteron is an excellent example of renewed focus on the professional workstation and Enterprise server environment, but whenever Opteron products cannot compete with Intel Xeon CPUs on sheer performance, AMD should immediately fall back to the value-priced strategy.

Despite an absurd asking price, Intel has already swayed consumer opinion by producing the best-performing desktop processor available. The only way for AMD to salvage reputation and compete with this mentality is by offering a similarly powerful product to the consumer market that sells for a fraction of the comparative cost. This is likely the legacy AMD's Phenom-II X6 will inherit. Going forward, value-priced products are really AMD's only dominant strength over Intel. AMD processors have yet to match Intel's 32nm fabrication process, which makes product development a real sore spot for AMD when it comes to selling Dell and HP on the Opteron product line. As head to head performance becomes more comparable, AMD can relax discount pricing; but not any time before then.

One other area AMD could concentrate their efforts is advancing a specific technology beyond the competition. Intel's Advanced Encryption Standard (AES) Instructions Set R3, introduced with Gulftown, posts a 2000% performance gain over the previous Nehalem microarchitecture. While I'm not suggesting the AMD take on Intel's AES capability, which is pretty well untouchable judging from the Everest AES encryption results, there are other areas that serve to gain ground. One suggestion: command queuing and pre-order fetching. Since SSDs have significantly improved storage technology response time, processors should begin to find ways of sequencing data ahead of the user's request. Computer hardware needs more than speed and power; it needs purpose.

Random After-Thoughts

New and upcoming DirectX-11 software notwithstanding, video games have generally required the same graphical power as they needed over the past few years. Newer server and virtualization technology continues to refine efficiency and uses fewer CPU cycles. So essentially software is barely moving forward while hardware is making leaps and bounds. Which raises the question: to what end?

Software just hasn't been keeping up its end of the deal, and most people still use 32-bit technology (introduced back in Windows 95). it's sad but true. Games like Crysis helped give reason for advancing graphics technology, just like virtualization technology and Terminal Services helped push processor power. But the apex of software demands hasn't really changed in several years, and having massive amounts of system memory are only helpful if there's an application that requires it.

I used to overclock my CPU to get a few extra frames out of video games, or earn more work-units per day with Folding@Home. But for the past few years, just about any graphics card can fold proteins better than the CPU ever has. We now have desktop processors that can encode multimedia in a mere fraction of the time it used to take. While hardware progression hasn't hit a wall, the light at the end of the tunnel is getting very close.

I believe that If software doesn't come around soon, perhaps coming in the shape of widespread adoption towards 64-bit computing using Microsoft Windows 7, we'll soon share the same dilemma facing the automobile industry: high-horsepower engines with break-neck torque driving down roads with a 65-MPH speed limit. This has already been the case for far too long, which is why we're already sharing the other auto-industry dilemma: power efficiency.

<< Previous Page - Next Page >>