I had a RAM stick go bad. Debugged with user-mode memtester on Linux, as I couldn't get memtest86 or memtest86+ to work at all (maybe grub splash screen confused them? Maybe they're still x86 and not 64bit capable?).
Was causing spurious data corruption issues and maybe the hard lockups I was getting occasionally (will see about that).
I'm waiting on a reply about warranty from the retailer. Maybe I'll have to send both sticks of the pair back even though only one is faulty, which would mean desktop being out of action for a while.
@mathr you could check of it alone. Than you know better if both are gone, or maybe ram socket.
Replacing both isn't bad idea.
If you would receive new one with different timing it may slow down system
@hamjb yes I checked each stick alone in different sockets to narrow it down:
stick A in socket 2 and stick B in socket 4 -> error
stick A in socket 2 -> ok
stick B in socket 2 -> error
conclusion: stick B is faulty
I've been reading up on RAM timings, seems complicated. the particular timings of these sticks seems no longer available from the retailer, the most similar alternatives are slower (higher latency numbers same clock) or about the same (higher latency numbers higher clock -> same or a very tiny bit faster latency in nanoseconds). need to check if my motherboard+cpu can handle the higher clock...
Welcome to post.lurk.org, an instance for discussions around cultural freedom, experimental, new media art, net and computational culture, and things like that.