Memtest: not too useful for testing anymore?

[magellan]

I've noticed over the course of testing the overclocks of both my LGA775 q9550 and now my SB-E 3820 that memtest doesn't seem to be a very good indicator of whether or not memory is stable at any given set of timings and clock speed. I used to run 8 hour memtest tests that were successful and yet would fail orthos testing within minutes and the failure would be down to memory timings and or V-DIMM.

 

[wagex]

but was the memory bad, or was just the settings? i believe it just tests for bad spots or what ever in the ram.

 

[petteyg359]

memtest86(+) is not a stress test. It does not run your CPU at full power, nor does it read and write to RAM as fast as possible. It only tells you whether your RAM is functioning correctly. Of course it's not going to determine if an overclock is stable.

 

[magellan]

memtest86(+) is not a stress test. It does not run your CPU at full power, nor does it read and write to RAM as fast as possible. It only tells you whether your RAM is functioning correctly. Of course it's not going to determine if an overclock is stable.

I see your point, but the failures I was experiencing were NOT due to the CPU overclock, they were down to the RAM timings and/or V-DIMM. When I test memory timings at a given bus speed I lower the CPU multiplier as much as possible. I'm just kinda surprised that 8 hours of successful memtest testing is mostly irrelevant when orthos fails in minutes at the same exact memory timings and V-DIMM.

 

[Automata]

Think of memtest as a preliminary test before you boot into an operating system. It isn't as stressful as other tests (LinPack/LinX, etc), but it allows you to quickly test settings before booting into an operating system and prevents corruption of said operating system if the system is unstable. The test isn't perfect, but that doesn't mean it isn't useful.

 

[wingman99]

I like memtest86+ just to detect bad memory.