Today is all about acronyms: AMD, ARM, ATI, GloFo, and OPT. Haven’t heard of that last one? I just coined it, as “other peoples’ technology.” As in, AMD’s upcoming server chips won’t have a lot of AMD’s technology in them.
AMD announced today that it was becoming a full-fledged ARM licensee, with plans to make 64-bit ARM-based server chips by 2014. AMD isn’t giving up on x86; the company will offer both ARM- and x86-based server chips and let the customer decide which is better. Both flavors will include the on-chip network fabric that AMD acquired from SeaMicro a year ago, and some of chips will include graphics engines based on ATI’s popular Radeon architecture.
That’s swell, and probably a good strategic move for AMD. But I can’t help noticing that all of these gee-whiz features are somebody else’s technology. What, if anything, is AMD itself actually contributing?
The processor is ARM’s upcoming 64-bit “Atlas” design, because AMD chose not to take out an architectural license and design its own implementation. The graphics come from ATI, which AMD acquired a few years ago. The on-chip switch fabric comes from SeaMicro. And the silicon fabrication will be outsourced to GlobalFoundries and/or TSMC, two foundries in which AMD has invested in the past. In short, there’s no AMD-branded IP here. It’s an amalgamation of OPT: other peoples’ technology.
AMD calls this its “ambidextrous” strategy: ARM processors on the one hand, x86 processors on the other hand. Clever. And it’s a path that Intel certainly will not follow. But there are a half-dozen other CPU companies making ARM-based server chips, so AMD won’t have the market to itself. in fact, it’s kinda late. AMD will find itself competing with Marvell, Calxeda, Applied Micro, and others for the hot ARM-based server business. And AMD’s contribution to all this? Other peoples’ technology.