Quote Originally Posted by cantankerous View Post
Thanks for your attachment C-N. Before I write anything off I always try a second time to check for consistency. Not so sure I am glad I did this as I left it running while at work and just came home and 10h have passed and it is still running! Why would one day it fails at 8.5 and the next still be running after 10? To me, if it were going to fail it should have failed close to the same point it did before, if not sooner. I am helping my dad move today so am going to keep it running and see if it is still going when I get back in a few hours. If so I have no clue what to say. I am wondering if it was heat related as the ambient temps were much warmer during the day when the 8.5 h run test failed where as the 10h run was at night when ambients dropped at least a good 5c. I will see how it keeps up today when the ambient temps start to rise again.

BTW, your prime95 attachment was about dead on. At the 10h mark I am well into 1024k which is the restarting point. Good job!

BTW, if it fails again could it possibly be a GTL slightly off tune? I would hate to have to use more voltage as temps are already getting up there. I am thinking since 8.5 is fairly long I may try and adjust the GTL value for the 4th core up/down by one and see how that makes out. Either way, at 8.5h minimum before it fails I am sure my system is stable enough. I highly doubt it will crash during 24/7 use do you?
Stable is whatever you call stable, we are all different, personally one full 9 1/2 hour Prime pass + a night of gaming is stable enough for me.

I found the Prime time line useful when the rig reboots during Priming rather than failing. Prime writes a log you can use to see what fft it failed on. You can then run a custom test at that fft rather than waiting 8 hours for the fault to re appear.

Any number of things could have caused the failure but for my money it would most likely be a RAM issue. I often find that running the 4096k test for an hour or so is enough to find memory issues. I have had plenty of rigs that pass 8 hours + then fail 3DMARK or fail mid game due to running the memory on the edge. I find one hour + of the 4096k test is usually enough to find voltage & timing issues.

CN