Hi,
I have a small cluster of systems with:
dual AMD MP 2000+
ASUS A7M266-D motherboard
2GB RAM... DIMMs are Crucial 512MB, DDR, 266Mhz, CL2.5, ECC
I'm running 2.4.20 (Debian, or vanilla, didn't seem to matter)
The only stable configuration I can seem to get is 1 CPU and 1 DIMM.
Every other setup Cerberus just blows up in anywhere from 10 minutes to
a little over 2 hrs. Sometimes I see an oops, other times it just stops
dead and no interrupts are handled (caps lock doesn't turn on the little
light on the keyboard). I saw a similar conversations in the archive
about a setup similar to this and various solutions, one being to
underclock the CPUs, suggesting, I guess that CPU heat is a problem
(even though I never see an alarm out of lm_sensors). I did update the
board to the latest BIOS (1009), that doesn't seem to have any effect.
testing with Cerberus:
1 CPU/ 1 DIMM run all the time
2 CPU/ 1 DIMM might run for 3 hours
1 or 2 CPU with more than 1 DIMM, lucky to last 20 minutes, usually
less.
I have tried each of the four DIMMs individually, they all appear to be
fine.
So I'm wondering if anyone has any insight into what the problem might
be. Is underclocking the chips all I can do?
--
Dale Harris
[email protected]
/.-)
On Fri, 14 Mar 2003 15:12:27 -0800
Dale Harris <[email protected]> wrote:
> So I'm wondering if anyone has any insight into what the problem might
> be. Is underclocking the chips all I can do?
Have you checked your power supplies?
On Sat, Mar 15, 2003 at 11:34:29AM +0900, Bruce Harada elucidated:
>
> Have you checked your power supplies?
Well, it's supposed to be a 400W power supply. But no, I haven't done
much in that area yet. I suppose I better get a P3 Kill-A-Watt:
http://www.efi.org/products/power/p3watt.html
--
Dale Harris
[email protected]
/.-)
On Saturday 15 March 2003 03:59, Dale Harris wrote:
> On Sat, Mar 15, 2003 at 11:34:29AM +0900, Bruce Harada elucidated:
> > Have you checked your power supplies?
>
> Well, it's supposed to be a 400W power supply. But no, I haven't done
> much in that area yet. I suppose I better get a P3 Kill-A-Watt:
Did you enable "AMD 76x native power management"? Setting CONFIG_AMD_PM768=Y
leads to the results you described for me, among others. (Mainboard is Tyan
S2466, same chipset, AMD-760MPX). Disabling or loading as a module works fine.
Speaking of power: my rig, with 2x Athlon XP 2000+, 1GB of ram, a GF2 GTS and
two hds, draws about 190 watts under full load, somewhat more when the
optical drives are active. But the most I ever registered was 211W (at the
mains level, typical PSU efficiency is around 70%, iirc).
MfG, Ulrich
On Fri, 2003-03-14 at 23:12, Dale Harris wrote:
> 1 or 2 CPU with more than 1 DIMM, lucky to last 20 minutes, usually
> less.
>
> I have tried each of the four DIMMs individually, they all appear to be
> fine.
>
> So I'm wondering if anyone has any insight into what the problem might
> be. Is underclocking the chips all I can do?
The A7M266-D is infamously picky about the PSU, about the DIMM quality
and about the heat handling. Its one of the reasons I suspect that the
dual athlon didn't get AMD into the enterprise market - goes like a
rocket, requires space shuttle grade components and produces as much
heat.
I am using dual cpu and two dimms (ecc, registered). Contrary to the
docs I've never managed to get it stable with 3 dimms.
Also make sure you use a recommended PSU, many 400W PSU's don't
produce the right amount of power on the right voltage lines.
Alan