One of the questions raised over the lawsuits is just what is Intel after? Intel can't be after money as a reason for the spin-off is that AMD doesn't have that sort of money to spend. While Intel could be maneuvering to kill AMD, the fact is that current cross-licensing agreement is set to expire in 2011. If Intel was looking to kill AMD by withholding their x86 license and refusing to negotiate a new license, their's little doubt Intel could afford to wait another year and really put the screws to AMD's shareholders and investors. There's also the huge political mess that would result as anti-monopoly lawsuits would be flying left and right, and Intel could be in the possible legal position of being forced to open the x86 hardware architecture up with no license agreement. So I find it unlikely that Intel is actively trying to outright kill AMD with the lawsuits.
The answer as to what Intel might be after came from of all places, Schlock Mercenary. The story in question has General Xinchub posing and posturing as he tries to get the mercenaries to do some work he himself cannot accomplish. So, thinking outside of the box, what does AMD have that Intel doesn't have, or can't get?
Well, there is HyperTransport, but that's backed by a consortium that Intel could join at any time without having to resort to legal threats. Not to mention that Intel is also working on their own Peer-to-peer topographical connection, and while Hypertransport is years ahead of Intel's efforts, the technical superiority on paper hasn't manifested performance superiority.
There is Crossfire, but Intel already had Crossfire capable motherboards in the past, and while the Crossfire license has been renewed, Intel surely could have gotten that updated without legal wrangling. In any case, since Intel isn't in the graphics card business, it made sense for them to support both multi-GPU solutions.
Only, Intel IS getting into the graphics card market with their upcoming Project Larrabee. Epic, which has pummeled Intel in the past for their graphics, had good things to say about Larrabee. Larrabee has been revealed to be little more than multiple P5 cores with an extreme die-shrink, and is thus little more than an x86 software rendering solution. Granted, as TechArp's Intel page indicates, Intel has been offloading critical graphical instructions to the central processor for years. Larrabee is thus nothing more than an extension of the software rendering Intel has been using, only with dedicated hardware.
However, ATi's extensive patent and product portfolio is deep enough to be of extreme Interest to Intel's graphical ambitions. It is possible that AMD could extract a high-dollar figure from Intel in order to have access to these patents, and a cross-licensing agreement could be Intel's ticket. That's getting close enough to a reason, but there still needs to be something to clench the deal.
Only, Intel IS getting into the graphics card market with their upcoming Project Larrabee. Epic, which has pummeled Intel in the past for their graphics, had good things to say about Larrabee. Larrabee has been revealed to be little more than multiple P5 cores with an extreme die-shrink, and is thus little more than an x86 software rendering solution. Granted, as TechArp's Intel page indicates, Intel has been offloading critical graphical instructions to the central processor for years. Larrabee is thus nothing more than an extension of the software rendering Intel has been using, only with dedicated hardware.
However, ATi's extensive patent and product portfolio is deep enough to be of extreme Interest to Intel's graphical ambitions. It is possible that AMD could extract a high-dollar figure from Intel in order to have access to these patents, and a cross-licensing agreement could be Intel's ticket. That's getting close enough to a reason, but there still needs to be something to clench the deal.
Intel's lawsuit also involves Globalfoundaries. Keeping the idea that Intel wants something AMD and Globalfoundries have, and that has to be licensed from both, or from just Globalfoundries, there is one important item in the Globalfoundries stack. I think Intel's after AMD's APM, or Automated Precision Manufacturing. AMD's fab technology has been both a blessing and a hinderance over the past years.
APM has allowed AMD to achieve astoundingly high yields on their processors. Intel, while having more fab capacity, and their own fab optimization techiques, is widely reguarded to have less than stellar yield ratios. This is actually a blessing to Intel as Intel can rebin failed products of one production range as lower-end products. I'm thinking stuff like Pentium to Celeron. AMD, on the other hand, has had a consectutive problem of having lots of mid-range parts in the market because their yields are so good, but nothing really on the top-end because they can't get everything else together. While AMD is open to licensing out APM, there's little doubt that APM for Intel would be set to a cost high enough it would be cheaper for Intel to just develop fab techniques internally.
With Intel extending their x86 designs into low-power architectures like Atom, and the upcoming project Larrabee's, APM could save Intel time and money on their external processor designs. Considering that Intel is open to licensing some of their hardware designs to other fab companies, the possibility is raised that Intel could be looking at a similar option for Globalfoundries.
Together, Intel's aspirations to get into the graphics market coupled with their desire to improve their own foundry technology might explain why Intel has gone after AMD so aggressively over the spin-off of Globalfoundries.
APM has allowed AMD to achieve astoundingly high yields on their processors. Intel, while having more fab capacity, and their own fab optimization techiques, is widely reguarded to have less than stellar yield ratios. This is actually a blessing to Intel as Intel can rebin failed products of one production range as lower-end products. I'm thinking stuff like Pentium to Celeron. AMD, on the other hand, has had a consectutive problem of having lots of mid-range parts in the market because their yields are so good, but nothing really on the top-end because they can't get everything else together. While AMD is open to licensing out APM, there's little doubt that APM for Intel would be set to a cost high enough it would be cheaper for Intel to just develop fab techniques internally.
With Intel extending their x86 designs into low-power architectures like Atom, and the upcoming project Larrabee's, APM could save Intel time and money on their external processor designs. Considering that Intel is open to licensing some of their hardware designs to other fab companies, the possibility is raised that Intel could be looking at a similar option for Globalfoundries.
Together, Intel's aspirations to get into the graphics market coupled with their desire to improve their own foundry technology might explain why Intel has gone after AMD so aggressively over the spin-off of Globalfoundries.
No comments:
Post a Comment