Per Jessen wrote: > I've been using mprime as a stresstest on a new AMD 64bit quad-core > system I've put together. It's failing miserably, causing the machine > to automatically reboot within 15-20min. > It doesn't seem to be a thermal problem, as running a parallel compile > of the Linux kernel actually drives a higher CPU temperature.
If you've successfully run through ALL of the Memtest86+ tests TWICE, then other possibilities (in no particular order) are: Are you OC-ing? If so, have you backed off TWO steps to give one full step of margin? Are ALL the motherboard PSU supply sockets used? Have you got a PSU overload on one or more supply rail? Is there a bad thermal contact between the CPU and heatsink? (Transient temps can be very much higher than what is reported/measured. MPrime will heat the CPU far more than a program compile ever can! Very strange if you have a /compile/ showing a greater temperature... So try reseating your heatsink cleaned and with a VERY THIN layer of thermal grease.) Have you applied the Linux kernel patch and/or the motherboard BIOS patch for the Barcelona TLB bug (assuming that is what you have)? Have you a bad connection or short somewhere that is the real problem? (Try running with a minimal hardware/software system to run only the test.) Let us know what you find. Good luck, Martin _______________________________________________ Prime mailing list [email protected] http://hogranch.com/mailman/listinfo/prime
