Athlon/MSI mobo combo broken?
Hi!
I have now had two MSI mainboards and two Athlons with 1330 MHz, and none of them works as advertised. When I compile an Athlon kernel (I previously had an Athlon 900 with an Epox board, which was flaky and did not boot reliably and the USB subsystem was unreliable, too), the resulting kernel will boot only partially, get spurious errors like a divide error in the reiserfs code trying to mount a reiserfs volume, and finally panic on me because it tried to kill the swapper.
When I compile the same kernel for Pentium Pro, it works. How can this be?
This is not my only problem with this combination, though. My network card (tulip) only works in two of the six PCI slots, my 1995 NCR SCSI controller only works two slots, and there is only one slot combination with my sound card that actually works. When I start Windoze on the hardware, it hangs trying to load drivers for the NCR controller.
Since this is now my second MSI board, my second power supply, my second Athlon and there appear to be no thermal problems (judging from the BIOS health display), I am out of guesses here. I basically replaces everything in my PC and it still won't work.
Felix
Quoted 2 lines · Felix von Leitner
I have now had two MSI mainboards and two Athlons with 1330 MHz, and none of them works as advertised. When I compile an Athlon kernel (I
VIA chipset..
Quoted 2 lines
When I compile the same kernel for Pentium Pro, it works. How can this be?
Theory right now: Because the Athlon kernel does streaming memory copies at full bus performance. Not all VIA chipset boards seem to cope.
I'd be interested to know if 2.4.6-ac5 Athlon optimised works on your board. The reason for this is that it contains the official VIA fixes for their IDE corruption problem rather than our own.
Quoted 5 lines · Alan Cox
Quoted 2 lines · Felix von Leitner
When I compile the same kernel for Pentium Pro, it works. How can this be?
Theory right now: Because the Athlon kernel does streaming memory copies at full bus performance. Not all VIA chipset boards seem to cope.
I fixed my MSI/athlon stability problems by reducing PS load. My next purchase will be a 450W PS to replace my 300W.
Athlons are real power suckers, 300W is probably 'barely enough' for typical PC.
-Dan
Signature · 1 line
[-] Omae no subete no kichi wa ore no mono da. [-]
Quoted 5 lines · Dan Hollis
Quoted 3 lines · Alan Cox
Quoted 2 lines · Felix von Leitner
When I compile the same kernel for Pentium Pro, it works. How can this be?
Quoted 8 lines
Quoted 2 lines
Theory right now: Because the Athlon kernel does streaming memory copies at full bus performance. Not all VIA chipset boards seem to cope.
I fixed my MSI/athlon stability problems by reducing PS load. My next purchase will be a 450W PS to replace my 300W.
Athlons are real power suckers, 300W is probably 'barely enough' for typical PC.
I'd rather say it may be related to how your MSI board supplies power to the Athlon, rather than the Athlon itself. I've been using ABit KT7A + [email protected] for half a year now, without a _single_ stability problem - and this machine is my primary workstation, being switched off basically only when I sleep.
I have Matrox G400/YMF724/NIC/CDR/DVD/three HDs (two in RAID) and... 250W PS.
A friend has purchased basically the same config, he has one HD less but his CPU is overclocked to 1150MHz. He uses both Linux and Windows and also does not report any problems.
cheers,
Milek
Ps. One probably important thing is that I'm only using HPT370 IDE controller under Linux.
Signature · 3 lines
mailto:[email protected] | "Man in the Moon and other weird things" - http://wfmh.org.pl/~thorgal/ | see it at http://wfmh.org.pl/~thorgal/Moon/ Fight for the good cause: http://www.laubzega.com/dvd/
Hi,
Quoted 3 lines · Alan Cox
I'd be interested to know if 2.4.6-ac5 Athlon optimised works on your board. The reason for this is that it contains the official VIA fixes for their IDE corruption problem rather than our own.
I have a MSI K7T Turbo (MS-6330) mobo (VIA-KT133A chipset) and had a go with the 2.4.6-ac5 kernel with Athlon optimisations on. The overall impression is that the combination is slightly more stable than kernels without the Southbridge fix. I.e. I now manage to boot in single user mode, however running anything as advanced as gcc is out of the question: oopses all over the place and also some complaints from the VM-system.
I am in no way qualified enough to go hacking around in the kernel myself, but am quite willing to test any patches that might help towards solving the problem.
Cheers,
Gerbrand van der Zouw
If you have PC100 SDRAM try to replace it with PC133.
This worked for my motherboard. (Gigabyte GA-7ZXR, VIA-KT133A, Athlon 1.1MHz)
Cheers too
Gerold
Quoted 17 lines · Gerbrand van der Zouw
Hi,
Quoted 3 lines · Alan Cox
I'd be interested to know if 2.4.6-ac5 Athlon optimised works on your board. The reason for this is that it contains the official VIA fixes for their IDE corruption problem rather than our own.
I have a MSI K7T Turbo (MS-6330) mobo (VIA-KT133A chipset) and had a go with the 2.4.6-ac5 kernel with Athlon optimisations on. The overall impression is that the combination is slightly more stable than kernels without the Southbridge fix. I.e. I now manage to boot in single user mode, however running anything as advanced as gcc is out of the question: oopses all over the place and also some complaints from the VM-system.
I am in no way qualified enough to go hacking around in the kernel myself, but am quite willing to test any patches that might help towards solving the problem.
Cheers,
Gerbrand van der Zouw
Quoted 1 line
If you have PC100 SDRAM try to replace it with PC133.
ugh, this is the same (mistaken) approach as turning off CONFIG_MK7:
cripple performance enough that your system works. turning off CONFIG_MK7 disables Arjan's nice code in mmx.c which delivers 2-4x the copy/zero-page bandwidth...
there *are* stable via/athlon systems, and that indicates that the problem is not inherent to the chipset. I have a gigabyte ga-7zx, duron/600, pc133 system that has always been rock-solid. and I recently built an Asus a7v-133, tbird/1199, pc133 system that is also entirely stable. both run cas2 pc133, CONFIG_MK7 kernels, etc. both have fairly generous power supplies.
so far, the only plausible theory is that some individual factor(s) (MB, bios settings, power quality, dram quality, etc) causes the instability that some people report.
regards, mark hahn.
I did not say chipset at all.
You seem to miss the fact, that >= 40 % of the SDRAM modules are not even close to the spec. The german magazin ct can sing a song about it. http://www.heise.de/ct
Software has nothing to do with this.
The windows installation on my PC suffert from the same effects.
It was just harder to spot there below ... .
Every memory tester i tried was showing no problems at all.
It may be triggered more likely or not, but bad SDRAM cannot be fixed in software. Thats my own only plausible theory.
regards
gerold
Quoted 16 lines · Mark Hahn
Quoted 2 lines
If you have PC100 SDRAM try to replace it with PC133.
ugh, this is the same (mistaken) approach as turning off
CONFIG_MK7:
cripple performance enough that your system works. turning offCONFIG_MK7disables Arjan's nice code inmmx.cwhich delivers 2-4x the copy/zero-page bandwidth...there *are* stable via/athlon systems, and that indicates that the problem is not inherent to the chipset. I have a gigabyte ga-7zx, duron/600, pc133 system that has always been rock-solid. and I recently built an Asus a7v-133, tbird/1199, pc133 system that is also entirely stable. both run cas2 pc133,
CONFIG_MK7kernels, etc.
both have fairly generous power supplies.so far, the only plausible theory is that some individual factor(s)
(MB, bios settings, power quality, dram quality, etc) causes the instability that some people report.regards, mark hahn.
Quoted 2 lines · gjury
You seem to miss the fact, that >= 40 % of the SDRAM modules are not even close to the spec.
There's something to be said for sticking with brand name memory or at least memory from reputable vendors...