I keep seeing techs in our shop skip the full memory test on a new build and it drives me nuts

They'll run a quick POST check and call it good. But last month, I had a custom rig for a video editor that kept crashing in Premiere. Everything else tested fine. Ran MemTest86 overnight and it found a single bad address in one of the four brand new sticks. Swapped it, problem gone. That was a $2500 machine and we almost sent it out the door. A ten minute boot test is not enough. How do you guys handle memory validation on high end systems?

2 comments

2 Comments

iris_dixon9129d ago

It's not just about finding bad sticks. A full test can show you borderline stability that a quick POST misses. I've seen systems pass POST but fail when the memory heats up under a real load. That video editor's machine was probably writing to that bad address during a render. The extra time is cheaper than a callback and a lost client who thinks you built them a lemon.

clairebaker29d ago

That "borderline stability" point is interesting... what kind of full test do you run? I've used memtest86 overnight, but I'm never sure if a few passes is really enough to catch those heat-up issues you mentioned.