In what is being seen as a second wave for the unit, Intel Corp is preparing to take some of its Supercomputing Systems Division’s technologies to market (CI No 2,439), but there remains a question mark over the future of its first-wave Paragon supercomputer and quite how the company will be able to bundle up separate Supercomputing Systems technologies that were designed to work together. While Supercomputing Systems, in Beaverton, Oregon, has always had a high level of US government Defence Advanced Research Projects Agency funding, which it has spent on some long-term, deep science development, and which has been most often sold back to defence research labs and establishments, the market for high-end scientific supercomputers has topped out, according to scalable system watcher Smaby Group Inc. Insiders believe it was the appointment of former Cray Research Inc executive Ed Masi as an Intel first officer – president of Supercomputing Systems – a couple of years ago that saved the unit’s life. Intel argues its been quietly perfecting the technologies that competitors such as Cray Research, Convex Computer and Kendall Square Research Inc have had to press, sometimes hastily, into commercial service as quickly as possible. However, whether there is or will be a market for scalable commercial systems large enough to support the existing crowd, let alone the likes of IBM Corp, Intel or its partners too, remains in question. Certainly the vendors are doing what they can to talk just such a market into existence, but their dogmatic pursuit of triumph at the benchmark table may be misplaced, as recent studies suggest that users interested in those kinds of technologies care less about raw performance than of the price-performance benefits – they won’t buy unless the price is right. Between two and four times better price performance than existing systems is said to be a median requirement. The scalable crowd is also still very much hung up on the componentry of their architectures, of one interconnect system versus another, and according to all the indicators, that is likely to become much less important.
