The Register reported this, you may boo and hiss, but there are PDFs on Intel's site backing them up so this time they're right. 
One involved AGP I/O voltage levels apparantly not being within specs, but was fixable through changes to the motherboard. Possibly this could be a stability issue, but as Intel has said they did not plan to fix it, chances are pretty much that this errata is a non-issue once the board change is in place.
Somewhat worse is that the chipset AGP prefetch buffer isn't working as it should, and has to be disabled via the system BIOS. Now, the article doesn't say what kind of prefetch buffer this is, if it's something other than the normal FIFO buffers present in chipsets, or what exactly it prefetches. It would be fun to speculate a bit on the possible repercussions this might have as far as performance is concerned...
Anyone have any kind of info on what this prefetch buffer actually does? Oh, and by the way... Intel doesn't plan to fix this one either. Maybe not because the performance hit is insignificant (if it was, why design a buffer in the first place?), but because it seems the affected chipsets won't have much of a shelf-life (nasty flashbacks of socket 423...), and perhaps the issue has been resolved in the successor.
*G*
One involved AGP I/O voltage levels apparantly not being within specs, but was fixable through changes to the motherboard. Possibly this could be a stability issue, but as Intel has said they did not plan to fix it, chances are pretty much that this errata is a non-issue once the board change is in place.
Somewhat worse is that the chipset AGP prefetch buffer isn't working as it should, and has to be disabled via the system BIOS. Now, the article doesn't say what kind of prefetch buffer this is, if it's something other than the normal FIFO buffers present in chipsets, or what exactly it prefetches. It would be fun to speculate a bit on the possible repercussions this might have as far as performance is concerned...
Anyone have any kind of info on what this prefetch buffer actually does? Oh, and by the way... Intel doesn't plan to fix this one either. Maybe not because the performance hit is insignificant (if it was, why design a buffer in the first place?), but because it seems the affected chipsets won't have much of a shelf-life (nasty flashbacks of socket 423...), and perhaps the issue has been resolved in the successor.
*G*
