About two months ago,I put together a new computer,and everything was fine for about a month and week.
After that, I started getting crashes (random reboots) and errors in orthos and memtest (about 80 errors or so, and always in test 4. Memory addresses were on both sticks), whereas previously I had been able to run both with no problems for 10+ hours each.
I noticed that orthos errored much faster when running large fft/blend rather than small, so I figured it was a ram issue. I rotated the ram, which didn't do anything, and then switched them to the other two ram slots, which seemed fine. I ran memtest and orthos for 12 hours again each, and everything was fine...for about a week.
This morning I woke up and noticed that windows had rebooted and was at the login screen. Upon logging in, I got an error about a corrupted file and that I'd have to run chkdsk. I ran orthos on large fft, and it immediately rebooted. I went into memtest, and it started spitting out a lot of errors. Not just in test 4, but also in the other 3 tests before it. About 1500 errors before even finishing test 4.
Is it possible that the problem is with the cpu or motherboard, and not the ram? I find it very odd that it runs for weeks at a time, and then suddenly I have issues. Later this week I'm going to try testing out the ram on a friend's machine,and his ram on mine, so hopefully, that will confirm a ram error.
Intel Core 2 Duo e7200 (overclocked, or stock, I still get ram errors)
2GB (2X1GB) DDR2 800 Crucial Ballistix, being run at Crucial's specified settings.
If it is a ram issue, will I be able to RMA it? I no longer have the packaging or receipts, since I sent them in for a rebate, which I'm currently waiting on.
I didn't have anything installed. Hm. Autoupdate though.
I also realized that it was on bios defaults, and incorrect settings, so I went in and manually changed the ram voltage and timings to the rated ones.
My pc is plugged into a surge protector.
Currently running small fft test.
Here's the error:
1:Test 1, 1000 Lucas-Lehmer iterations of M62390271 using 3584K FFT length.
1:FATAL ERROR: Rounding was 0.4951171875, expected less than 0.5
1:Hardware failure detected, consult stress.txt file.
1:Torture Test ran 19 hours, 19 minutes 10 seconds -1 errors, 0 warnings.
2:Torture Test ran 19 hours, 19 minutes 10 seconds - 0 errors, 0 warnings.
Autoupdate can interference with stress testing cuz it will download updates and install them.
You need to not touch anything in bios or else it can change the outcome of a stress test. Auto/default doesn't mean manual. Rated specs don't always work. i.e. specs says 4-4-4-12. 5-4-4-12 may be more stable. Go back to bios & load default.
Then d/l this, burn as image to a cd/dvd, boot it up, go to CPU Tests, Mersenne Prime Test, let it run for 24 hours. Make sure there's no sudden change to room temp. Gradual change is ok. Leave the door open.
Everything was until monday night. It hard booted, and now I periodically get errors in firefox, corrupted disk messages, and occasionally it doesn't boot the first time (gets to windows xp screen with the bar and reboots). Memtest shows errors within seconds.
This is such a mess.
I'm just going to buy some new ram, and if that doesn't work, new mobo, and if that still doesn't work...a new cpu.
Hopefully I can stop at the ram, but I strongly suspect it's something more serious than that.