IBM and AMD are expected to announce an alliance that will lead to mainstream IBM servers based on the Opteron processor. A deeper partnership between IBM and AMD could take several forms, ranging from joint technology development to work on mainstream dual-processor servers. It might even cover the creation of more-powerful multiprocessor machines that could rival IBM’s current Intel-based 32-processor, the System x3950. Update: AMD also posted its quarterly results.
Well I, for one, am happy to see AMD get such a large ally. The stronger AMD gets the fiercer the compitition between them and Intel and that means better products for me. And THAT, makes me happy.
Seems like a lot of historically non-AMD system vendors are starting to find their close ties to Intel much less compelling than they used to be. At least in the mid-range to high-end server and workstation markets.
First Dell finally severed its exclusivity with Intel, for real this time. Then came the GAO rejection that caused Intel to lose a lot of US government contracts (due to the RAID 5 issues in Woodcrest… this is IMHO an under-publicized story). Now it appears that IBM is losing faith.
AMD seems to be doing fine without “Colonel” Sanders, but I think Intel is missing something now that Andy Grove (whom I really respect, btw) is out of the picture.
With AMD now licensing its coherent HyperTransport interface, and IBM pursuing its Cell architecture, this alliance could result in some truly interesting technical advances. At least on their surface, these two architectures seem to have a lot of synergistic potential.
Interesting… I’m not so sure I see the synergies as clearly as you do. Are you suggesting that the STI alliance could produce a PE-less version of the Cell and plug it in as a coprocessor for an Opteron? There would need to be some serious work to get the DMA to the SPE local store working over HT, and they’d need to completely overhaul the GCC backend, but I guess it’s possible. I don’t think it’s worth it from a business perspective.
Actually, the direction IBM’s going with eClipz (POWER6) seems to be a better fit for a next-gen Cell stepping (they’d probably remove most of the FPU). Here’s the relatively meager public info as reported by wikipedia:
http://en.wikipedia.org/wiki/POWER6