William Kenworthy wrote:
I'd recommend that anyone experiencing complete lock ups recompile their kernel using gcc 2.95.3 (or download an RPM of a 2.95.3 compiled kernel). Any peripheral modules such as alsa or nvidia-kernel will of course have to be recompiled as well.I rebuilt glibc with very basic (no extra!) flags/optimisations(and most of the system) via packages on an old cyrix 233 system (as I couldnt build them on the athlon) - no change :( Once underclocked, I rebuilt it all on the athlon with athon-tbird, -03, -pipe etc and its stable, and seemed faster than when clocked at 1.4 which is odd. If I wind the clock back to 1.4, the lockups come back: race condition dependent on speed?
I'm running on such a hybrid system, and my computer couldn't be more stable (it is a pain having to switch gcc's whenever I compile a kernel module, but that isn't too often).
GCC 3.2.[012] all occasionally die during emerges (not at consistent places or loads, may be a heat buildup problem, but I'd think that the 2 hour DIVX records I do of MST3k would illustrate problems as well). Probably just compiler bugs (filing a detailed bug report is kinda useless if it isn't reproducible).
I compile -march=i686 on my Athlon so it is interoperable with other CPUs in the house; the kernel is "Athlon" though.
I also don't use the nVidia drivers (I do have a GF2MX). Although I wasn't having problems as of 3123, I don't really use GL, and prefer to be able to use the nVidia fbdev.
It is also worth noting that AMD has had glitches that were dependant on speed (anyone who tried running Win95 on a k6/2 over 350Mhz knows what I mean). So while unlikely, a synchronization issue is worth looking at.
--
Public key available at pgp.mit.edu keyID:36F46FD0
--
[EMAIL PROTECTED] mailing list
