Intel’s Profusion: The Wait Continues
In November 1997 Intel Corp. purchased a little-known, Irvine, Calif.-based company named Corollary Corp. At the time, Intel indicated that it hoped to leverage Corollary’s eight-way SMP architecture -- dubbed "Profusion" -- as a means to manufacture standard, high-volume eight-way Intel PC servers.
More than a year has elapsed since the acquisition of Corollary, however, and there is no sign of the Profusion technology or standard, high-volume Intel PC servers based on the Profusion architecture.
At the time of the acquisition, Corollary president George White indicated that the Profusion architecture probably wouldn’t debut within the timeframe of the Pentium Pro, the microprocessor for which Profusion was originally built around. White suggested Profusion would likely appear in conjunction with Intel’s forthcoming Deschutes microprocessor.
Deschutes -- formally named Xeon -- shipped in June 1998, but officials at Intel and Corollary have remained mum on the subject of when Profusion will appear.
White, through a spokeswoman, did not respond to requests for an interview, and officials from Intel would not discuss a delivery time for the long-expected Profusion architecture.
One possible reason for Profusion’s absence is that with the impressive performance of Intel’s next-generation Xeon microprocessor, Intel isn’t feeling pressure to get Profusion out the door. Performance numbers released by the Transaction Processing Performance Council (TPC, www.tpc.org), for example, show four-way Xeon-based systems performing nearly 26 percent better than eight-way Pentium Pro-based systems in TPC’s standard TPC-C benchmark test. An HP NetServer LXr 8000 with four 400-MHz Xeon processors scored 20,433.93 tpmC, with a cost of $30 per tpmC, on the benchmark. That beats the 16,257.20 tpmC at $34 per tpmC score of the reigning Pentium-Pro-based performance leader, HP’s NetServer LXr Pro8.
James Gruener, a senior analyst with the IT consulting firm Aberdeen Group (www.aberdeen.com), says the issue is not merely a matter of pure processing power. The market for eight-way Pentium Pro systems evolved gradually as IT departments identified specific tasks that required more processing power than four-way Pentium Pro systems. Gruener predicts the market for eight-way Xeon processors will likely develop the same way. "If you look back at what the Pentium Pro eight-ways were like, that was a high profile but very small market. So I think it’s going to take some time for customers to find where they feel the four-way Xeons are not cutting it and determine that more processing power is needed," Gruener says.
Dan Kusnetzky, director of worldwide operating environments with International Data Corp. (IDC, www.idc.com) elaborates on Gruener’s assessment. He says because of the situations in which Windows NT is deployed, there really isn’t a requirement for high-end or exotic SMP configurations. "When we look at the missions people are giving to NT, scalability isn’t a problem because people are carefully selecting the missions they give NT to carefully match what they perceive the product can do," Kusnetzky says. "If we look at the missions that people are sending over to NT, it doesn’t look like there’s a requirement for either high-end SMP or clustering."
And then there are lingering questions over Windows NT’s ability to scale across more than four processors in an SMP configuration. According to Aberdeen’s Gruener, while Microsoft has made strides in augmenting the scalability of Windows NT, both Redmond and independent software vendors in general still have much work to do before Win32 applications and the Windows NT operating system can fully scale to take advantage of high-end SMP configurations.
"In the short time -- especially in the NT market -- there’s a limited number of apps that are able to effectively take advantage of the eight-way architecture," Gruener contends. "I think that we’ve had time over the last several years for a lot of ISVs to amend and improve their apps to take advantage of these architectures. And when Windows 2000 comes out, systems will likely perform better because they’ll be able to take advantage of Microsoft’s work to improve the scalability of the operating system."