AMD cools on Slot-B, but Alpha alliance still strong
Slot-B or no, there's a lot of API technology in AMD's roadmap...
Rumours of the demise of the deal between AMD and Alpha Processor Inc. (API) over Slot-B may have been somewhat exaggerated, according to sources close to the companies. AMD director of product marketing Steve Lapinski seems to have started this particular hare when unveiling the company's first 64-bit chip, "SledgeHammer," but he might just have said a little more than he intended to. "We're not going to pursue the direction of Slot-B," said Lapinski, while adding that API would be one of the main chipset suppliers for SledgeHammer. If this constitutes complete abandonment of Slot-B then it's bad news for API, but equally it's bad news for the AMD-API relationship. Slot-B is a part of API's effort to make commodity components available for system builders, and so AMD support is important to the company. But in terms of co-development the two are already in deep - too deep, probably, for AMD to risk trashing the relationship. For example, although AMD played down the significance of API's involvement in the Lightning Data Transport (LDT) I/O architecture it announced last week, API is actually co-developer of LDT, which obviously means that API and AMD will both be using it. LDT is intended for both I/O and co-processors, so might to some extent be viewed as an alternative to Slot-B. But although there's some apparent overlap (API's UP 2000 board uses Slot-B to house twin Alphas), AMD is still going to need packaging, and Slot-B or a derivative might be a good place to start. API European Business Director Miles Chesney suggests that given the context (Lapinski was essentially talking about the X86-64), he "may not have intended to be quite so definitive about Slot-B." Effectively, says Chesney, Slot-B is a technology that AMD doesn't need yet, but that doesn't mean it won't need it in the future. Equally Chesney can afford to be phlegmatic. API would find it very helpful to have AMD shipping Slot-B, but hasn't been planning on AMD doing so in the immediate future. "At this stage they are not yet ready to address Xeon, or its successors, so they do not need to take a position on packaging large L2 caches. They will need to in the future. When they are ready to compete in the Xeon marketplace they too will need to offer a packaging technology such as Slot-B." AMD did of course also announce a Xeon competitor (AMD unveils Athlon 'Xeon') and positioned LDT as the architecture for connecting multiple paired CPUs in servers. If Chesney's right, then we could probably envisage AMD going for LDT for its high end server architectures, but still finding a role for Slot-B in intermediate and workstation implementations. Slot-B has the advantage that it's already here, whereas LDT and AMD's 'Xeon killer' are scheduled simply for "some time in 2000." So Slot-B could find itself a role in forthcoming Athlon Ultra workstations and servers. Part of the problem, of course, is that the partners are currently heading into crossover territory, with API attempting to broaden its product line at the base, while AMD tries to move upwards. API already has a reasonable high end (which it positions around the base of Compaq's current Alpha offerings), so needs to commoditise. AMD has volume product and - courtesy of Athlon - a fighting chance of grabbing high-spec PC and workstation sales. But it needs to make a dramatic leap to establish credibility in the server arena, which is where LDT comes in. ® See also: API-AMD confirm collaboration
Sponsored: Benefits from the lessons learned in HPC