It has generally been assumed that the potential impact of the Pentium flaw was significantly greater than Intel Corp suggests, but not as serious as IBM Corp contends, and now the PC Week and PC Magazine laboratories have run their own tests, which lend weight to the assumptions: they say errors using floating point divide are 200,000 times more likely to occur than Intel claims and 200 times less likely than IBM claims; engineers and scientists will probably find the level of possible errors unacceptable, as will most people using a spreadsheet for complex financial calculations said PC Week advanced technologies analyst Peter Coffee.
