New PC, random BSOD. Multiple Errors.

[osiato]

Worker 1 failed in 13 minutes, the other 3 are still running. "FATAL ERROR: Rounding was 0.498046875, expected less than 0.4"

CPU Temp is 90*C in HWMonitor, I don't think I can see it in speedfan (no CPU line), 90*C in Speccy. Will update when/if the other workers fail.

I have been informed that 90* is high enough to melt circuitry and have stopped the test. Should I start again?

 

[cluberti]

Failure is not good, although overclocking can cause failures due to voltage over or under issues, and increased temperatures (heat and voltage are what cause CPU issues). Consistent temperatures above 80C or so with an AMD A8 are probably a bit too high as well, although the newer versions of Prime95 are a lot more stressful on the CPU than older versions (prior to v27):
32bit: ftp://mersenne.org/gimps/p95v266.zip
64bit: ftp://mersenne.org/gimps/p64v266.zip

If you're overclocking - stop, and revert to stock voltage and multipliers before doing anything else. If you're at stock, and if you run ~2 hours with even 1 worker failure, try again and see if things continue to fail. You want to see the CPU staying under 80C or so (I prefer to see consistent temps under 80C for AMD, and 70C for Intel, before saying cooling is adequate for current clock/voltage settings), and if you can't, you need to think about more adequate cooling for the CPU. If you see consistent failures at stock, however, you should probably talk to the vendor about replacement, unless you're willing/able to underclock and/or undervolt until the machine passes regularly without errors.

 

[osiato]

Failure is not good, although overclocking can cause failures due to voltage over or under issues, and increased temperatures (heat and voltage are what cause CPU issues). Consistent temperatures above 80C or so with an AMD A8 are probably a bit too high as well, although the newer versions of Prime95 are a lot more stressful on the CPU than older versions (prior to v27):
32bit: ftp://mersenne.org/gimps/p95v266.zip
64bit: ftp://mersenne.org/gimps/p64v266.zip

If you're overclocking - stop, and revert to stock voltage and multipliers before doing anything else. If you're at stock, and if you run ~2 hours with even 1 worker failure, try again and see if things continue to fail. You want to see the CPU staying under 80C or so (I prefer to see consistent temps under 80C for AMD, and 70C for Intel, before saying cooling is adequate for current clock/voltage settings), and if you can't, you need to think about more adequate cooling for the CPU. If you see consistent failures at stock, however, you should probably talk to the vendor about replacement, unless you're willing/able to underclock and/or undervolt until the machine passes regularly without errors.

I need to go for a few hours, but will look into underclocking in a few hours until I can sort out cooling. I don't think I'm overclocking.