On Wednesday, January 3, 2018, 2:31:44 AM PST, Georgi Guninski <guninski@guninski.com> wrote: On Wed, Jan 03, 2018 at 08:48:12AM +0000, jim bell wrote:
For some reason, I'm reminded of the 486 math processor screwup of 1992 (?). As I vaguely recall, the math coprocessor might have errors in the fourth digit of significance. Intel offered to replace the affected chips.
I think it is the Pentium FDIV bug. IIRC only the server CPU was replaced in the whole office, don't remember why.
While doing a Google search, I saw reference to the Pentium problems, but there definitely was a 486 problem too. In fact, it came with a joke: The drug "RU486" had been relatively recently been released, and the joke was that the 486 "prevented you from multiplying". The corrected, replacement CPU was given a double-sigma after the 486, although I just found that a similar double-sigma was used for the replacement for a 386 bug as well.
From TFA: microcode can't fix it, lol. AMD is not affected. Shouldn't Intel do recall again?
Presumably, yes, especially due to the performance hit. I'm satisfied with AMD, but the people who insist on Intel processors claim that it's worth the extra money. But that argument gets destroyed if "fixing" the problem produces more than a small hit. Jim Bell