Intel Says Some 'McKinley' Itanium 2 Chips Have Bugs

by Timothy Prickett Morgan

If you have a server or workstation using the 900 MHz or 1 GHz "McKinley" Itanium 2 processor from Intel, pay attention. According to an errata report that Intel is just about to post to its support site, there is an electrical issue with the McKinley processors that, under very specific circumstances, can cause machines to crash. The exact nature of the problem and the conditions under which it manifests are not clear.

According to Barbara Grimes, a spokesperson for Intel, the problem has to do with the actual circuitry inside the McKinley processors, not with operating system or application software that runs on them. While Grimes would not say what percentage of Itanium 2 processors might be affected by the glitch, what she did say is that only a percentage of the Itanium 2 processors made to date have a glitch, and a slightly smaller percentage of them would experience instability or a crash because of the electrical circuit problem because specific instructions in the Itanium core and specific kinds of data have to be manipulated by a faulty chip for problems to become apparent.

Intel has developed a test to check for the problem, which OEMs making McKinley-based machines and customers already using them in the field can use. This test will be available through Intel's support organization and through OEMs. Having identified a potentially bad chip, customers have two options. First, they can clock down the Itanium 2 processor to 800 MHz. The problem does not manifest itself on chips that are slowed down from 900 MHz or 1 GHz to this speed. Or, they can contact their OEM or Intel to get a replacement Itanium 2 processor, free of charge, that has been screened and identified as glitch-free.

Intel does not, says Grimes, plan to suspend shipments of McKinley processors. She also says the company is using the same test to ensure that the chips it ships now don't have the problem. Jim Dunlap, a spokesperson for Itanium partner Hewlett-Packard, which helped create the Itanium family of chips, says HP has not stopped shipping its McKinley-based products and has worked with Intel to implement a methodology to identify potentially troublesome chips and prevent them from crashing customer machines. IBM, which just two weeks ago announced the xSeries 450, its first Itanium 2-based server, has not, according to company sources, in any way decommitted to Itanium, but the company is working with Intel to absolutely ensure that the test for the bug works perfectly before it resumes shipments of the xSeries 450, which it halted last Friday.

According to Grimes, earlier this year, an OEM partner (which she would not name, and which neither IBM nor HP would admit to being) reported some issues with the McKinley processor as part of its own stress testing on the processors. After months of trying to figure out if it was a hardware or software problem, Intel concluded it is a problem in the chips themselves. The 900 MHz /1.5 MB L3 cache, 1 GHz/1.5 MB L3 cache, and 1 GHz/3 MB L3 cache versions of the McKinley are all affected by the glitch. Grimes says the problem is not found in the forthcoming 1.5 GHz/6 MB L3 cache "Madison" Itanium 2 processor that Intel is set to debut in mid-2003. This latter processor has been giving RISC/Unix servers a run for the money is recent weeks on popular benchmark tests. That the McKinley chips have a glitch is a problem, but if the Madison chips did, it would be a real headache for Intel and its OEM partners.

This is not the first problem Intel has had with its Itanium processors since they launched, by the way. In the summer of 2001, Compaq, now part of HP, identified a bug in the 733 MHz and 800 MHz "Merced" Itanium processors, the original Itanium chips, during some intense stress testing. HP and Intel said at the time that they had not found any customers who had experienced any problems like those it saw in the labs, but just to be on the safe side, the two vendors provided free Itaniums to customers who identified theirs as being susceptible to the bug.

Sponsored By HEWLETT-PACKARD

DEMAND MORE... Demand more from IT than its ever delivered before. Make it prove its value, make it pay. Demand a new IT architecture: one that is open, modular and flexible; one that adapts, and adapts quickly, to every IT event triggered by every business decision. Demand that technology yield to the disciplines of business and be subject to the same practices and return analysis as any other business decision. Demand an alternative to the way IT and IT services have been purchased, implemented and operated for the last two decades. Demand accountability, rather than account control, from your IT partner. Demand on-demand computing, the real thing, right now. On-demand computing really does exist, right now. You can see it. Demand the ultimate state of IT fitness: Insist that business and IT be perfectly synchronized, and speed the evolution toward an adaptive enterprise. Demand more from IT. And find out who, really, can deliver. Click the links below for more information on: adaptive enterprise IT consolidation business continuity management utility data center