Intel Corp. on Friday confirmed its Pentium Pro and just-released Pentium II chips have a minor math bug but said it hasn't immediately devised a way to fix the problem.
Intel said it expects the bug will affect very few applications and that it was working with software companies to devise ways the software can work around the defect.Unlike its controversial handling of a similar flaw in an earlier chip in 1994, Intel this past week publicly acknowledged the bug soon after learning of it. In addition, Intel on Friday released statements of support from major softwaremakers, including the three biggest - Microsoft, IBM and Computer Associates - saying that so far they haven't seen any problems with the bug.
Many of the software companies said that if they find any problems, they would create ways around them in their programs.
Intel, confirming reports of the flaw, said the problem related to operations that convert floating point numbers - which express a number in two components, the significant digits and an exponent, without using a fixed decimal point - into integers, or whole numbers. The chip is supposed to issue an "overflow" warning when large floating point numbers will not fit into the integer format, but in certain circumstances it apparently failed to do so.
Reports of the flaw were first disclosed by Robert Collins, a Silicon Valley engineer who maintains a Web site called Intel Secrets.
Intel spokesman Tom Walden said the company would start shipping revised chips without the flaw in about six months.
The Pentium II, introduced by Intel this week, is based on its Pentium Pro, a powerful chip used in personal computers for businesses. But it also incorporates the multimedia-enhancing MMX technology Intel recently added to its fifth-generation Pentium chips, used primarily in consumer PCs.