On Sunday 03 February 2008 10:21, Per Jessen wrote:
Last week I put together a new system:
Gigabyte GA-MA790FX-DS5 (BIOS F3 which includes the AMD patch). AMD Phenom 2.2GHz (quad-core) 4 x 1Gb Kingston KVR800D2N5/1G (on Gigabytes approved list). ATI Radeon X1650 graphics. 2 x Hitachi SATA disks. The usual stuff - DVD, CDRW, floppy.
The powersupply is rated for 350W. The cabinet has 2 extra fans.
I installed 10.3 on a software RAID1 of the two SATA drives. The system isn't overclocked or anything like that, it's mostly running on Gigabytes "optimized defaults".
Gigabyte may have overly aggressive "optimized defaults." I'd back off to stock values and then slowly tweak the overclocking up (if you're into that sort of thing).
After putting a new system together, especially one with fairly new components like this one, I usually do some stresstesting/burn-in. Nothing particularly fancy - memtest86, mprime, maybe some home-grown memory and IO-tests.
Well, this system is highly unstable under high load. If I start 4 x mprime torture tests, it will automatically reboot in 30-40minutes.
This suggests CPU overheating. Is the cooler installed properly? Is it operating properly?
So which one is a the faulty component? I suspected memory at first, as that has most often been the case. I tried memtest86+, but it didn't work, presumably due to lack of support for this chipset/CPU.
I ran a number of mprime torture tests with different memory configurations - 4G, 2G, 1Gb - which all produced the same result = auto-reboot after 30-40 minutes. I also reduced memory frequency to 667MHz and tried that - no change.
This to me pretty much exonerates the memory.
Then I thought - might the powersupply be just a little too small?
That was my first thought. Are the SATA drives 7200 RPM or 10,000 RPM? (Or even 15,000 RPM?? Are the such SATA drives?) They have significantly higher power consumption.
The Radeon X1650 graphics card gets pretty hot, so might it just be drawing a little too much? I replaced the Radeon card with an ancient ATI Wincharger PCI-card from around 1996. And ran the same tests on the same memory configs. Which produced exactly the same results.
Maybe. The power supply could be defective. Can you measure the voltages while the test runs?
Which would seem to indicate that the powersupply is doing just fine.
...
/Per Jessen, Zürich
Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org