Paul, Re: "The Xeon Bug could have been a showstopper. However, I have heard from
reliable sources that Intel was able to identify the cause of the "bug" (a L2 cache
conflict when 4 CPUs were in use with the 450 NX) and have been able to correct
(or fix) it.
One explanation that was offered was that Intel has made appropriate changes to
the PROGRAMMABLE MICROCODE features in their Pentium II device - to
correct the fault."
As I recall, when Intel introduced the Pentium II, they said that there would be the feature of writable, programmable, or flashable microcode, whose purpose was to fix bugs quickly. At the time, Intel said that there would be 2K (bytes) available for this. At the time, I had a couple of questions about it:
1. Is 2K enough? I guess you have to use it judiciously.
2. How do you get this microcode "close enough" to the logic it needs to fix? Generally, microcode, or control store is positioned all over the chip, to be as close to the logic it needs to control as it can. (Maybe it's static?).
In any event, if Intel did use this writable control store to fix the Xeon bug, they are to be commended, as you say, for putting a feature into the Pentium II core that paid off (as they continue to use more and more mainframe tricks all the time. ;-))
Tony |
