Overclocking reliability tester for Phenom-II x6 4GHz/1666MHz for CFD?

Hi Everyone: This may not be the right forum for this!

I am doing some numerical weather prediction on my Phenom II x6 overclocked to 4GHz under OpenSuSE 12.1 with 16 GB 1666MHz memory. I had tested the system with “burn” (Intel?) under WinXP-64 and memtest86, and all seems stable. Is there a recommended multi-CPU system tester for Linux? I need memory, memory bandwidth reliability, FPU calculations, and the interprocessor-communications tested. Games testers don’t test memory enough. I can get memory from memtest86, but it doesn’t really check the system as a whole, including the OS.

I’ve gotten a few segfaults recently, and need to recheck the system, to make sure it’s the CFD software, not my hardware. Once before I received a Phenom-II that I was able to isolate to a bad onchip memory controller, as well as bad multichannel memoy.

Thank You Very Much,

PS: I have a previous set of 4 ram sticks (16GB-1666MHz dual channel) in which one had a memory flaw (failed memtest86 always at the same address); however, it ran OK at 1333MHz, which is what the system clocked it down to during BIOS! Is this normal? What is the recommendation: should I try to isolate the bad stick and remove it (leaving 3) and clock the remaining at native (1666MHz), or whether I should use all 4 clocked at 1333MHz? My software is memory speed intensive. (computational fluid dynamics). Or maybe I should just try to replace the one bad stick? I’m not sure how that would affect dual channel memory.

Run prime95 on your system.

Got an nvidia gpu, use cuda instead?

Cheers Malcolm °¿° (Linux Counter #276890)
openSUSE 12.2 (x86_64) Kernel 3.4.6-2.10-desktop
up 1 day 20:10, 6 users, load average: 0.23, 0.19, 0.15
CPU Intel i5 CPU M520@2.40GHz | Intel Arrandale GPU