Brian's Brain

Continued from 'AMD: Time Keeps On Slippin', Slippin', Slippin'...'...

I've long suggested that AMD seems to have endemic 65 nm yield problems (speed, power, and/or functional), and the tri-core Phenom is pretty damning evidence from my perspective. Speaking of lithographies, if AMD was stumbling in a vacuum, its problems might be manageable. But put competitor Intel in the picture, and things look even more grim for AMD. Freed of its underwhelming NetBurst microarchitecture, fully on the impressive Core follow-on, and with 'tock' new microarchitecture Nehalem coming in two quarters, Intel has fully executed on its 'tick' 45 nm process conversion plan and is, at this point, at least one full process generation ahead of AMD, with a far more extensive fab network to boot.

Look, for example, at my assembled sampling of independent benchmark results published last week, on AMD B3-stepping material versus a range of Intel's CPUs spanning various price-vs-performance points:

• AnandTech
• Ars Technica
• Engadget (summary of several reviews)
• Maximum PC (via Gizmodo)

I've got AMD and Intel hardware in hand, as well, and plan to publish my own benchmarking results coincident with my upcoming late-May cover story. But for now, what's the summary? In spite of the 200 MHz core clock boost that the TLB flaw fix enables (thereby allowing Phenom to finally hit the speeds which were promised last November), and in spite of Phenom's 'true quad core' architecture versus Intel's supposedly inferior dual-die shared frontside bus approach...AMD's top-end Phenom got roundly performance-spanked even by mid-range Intel CPU counterparts.

AMD remains fairly competitive in the volume sweet spot of the market, admittedly, but whatever parity exists primarily comes about due to substantial Phenom price slashing versus Intel alternatives...price slashing that's particularly troubling if my poor-yield prognosis is correct. As I wrote last September:

I think you'll agree with me that short-term slashing does not a long-term sustainable strategy make.

Maybe Intel will once again become organizationally distracted (reminiscent of its decade-ago acquisition flurry) as it pushes into new markets with Atom, and AMD will benefit from Intel 'taking its eye off the ball'. Or maybe AMD's 45 nm process will come up much cleaner and quicker than its 65 nm predecessor. Or maybe the in-progress worldwide investigations of Intel's business practices will unveil damning evidence of malfeasance, leading to regulatory restraint that empowers its primary competitor. Or heck, barring these or other resurrection opportunities, maybe NVIDIA will buy AMD.

Regardless, I admittedly craft this writeup with more than a bit of a heavy heart. AMD and Intel have both been at their best when competition between them has been its healthiest, specifically across all segments of the PC business, and the tech industry has greatly benefited as a result of the spirited rivalry. We all need a rejuvenated AMD, Intel included. But I'm not seeing even a faint path to AMD's high-end rebound, far from a solid one...particularly considering the substantial amounts of cash that AMD's currently hemorrhaging, and the finite amount of surplus both currently available in the company's coffers and potentially coming from future investors.

Are you more upbeat on AMD? If so, shoot me your thoughts in the comments.