Hello
In my rig, I've recently discovered, that usual prime95/memtest/3dmark/etc. stress tests weren't enough in my case.
Purely by coincidence I disocvered problems with simple I/O operations - when I copied one larger (around 300mb) archive to the other disk, and 1 bit turned out to be different
.
So I made simple shell script that in loop copied some larger file, synced, and after that did binary comparison. Rinse, repeat. Turned out - that I had approx 2 1-bit errors per ~6 hours of testing. When I backed with fsb from 460ish to earlier 410, everything went back to normal.
The board in question is old rev. 1.0 ga-p965-dq6 (with E6400, TwinX2048-6400c4) - so pretty old for today, and without too many bios settings re overclocking. As it was a while ago, I can't recall precise settings - still they were carefully set to achieve stability under typical testers, as mentioned above (tens of hours spent in orthos and memtest).
I still wonder though, if it was cpu/memory culprit missed by orthos/etc., or southbridge (both disks were on ich8r) being unable to work properly under these conditions, that my small test caught somehow (also, my board doesn't allow sb overvolting).
Still - putting my setup aside - does anyone of you perform these type of tests. If so - any specific tools you use for that purpose ?
In my rig, I've recently discovered, that usual prime95/memtest/3dmark/etc. stress tests weren't enough in my case.
Purely by coincidence I disocvered problems with simple I/O operations - when I copied one larger (around 300mb) archive to the other disk, and 1 bit turned out to be different
. So I made simple shell script that in loop copied some larger file, synced, and after that did binary comparison. Rinse, repeat. Turned out - that I had approx 2 1-bit errors per ~6 hours of testing. When I backed with fsb from 460ish to earlier 410, everything went back to normal.
The board in question is old rev. 1.0 ga-p965-dq6 (with E6400, TwinX2048-6400c4) - so pretty old for today, and without too many bios settings re overclocking. As it was a while ago, I can't recall precise settings - still they were carefully set to achieve stability under typical testers, as mentioned above (tens of hours spent in orthos and memtest).
I still wonder though, if it was cpu/memory culprit missed by orthos/etc., or southbridge (both disks were on ich8r) being unable to work properly under these conditions, that my small test caught somehow (also, my board doesn't allow sb overvolting).
Still - putting my setup aside - does anyone of you perform these type of tests. If so - any specific tools you use for that purpose ?
Facebook
Twitter