While testing my new Xeon build for Socket 775, a bug in Memtest86 5.01 sent me on a wild goose chase.
I had just installed 4x 2GB DDR2 1066MHz memory modules in my Asus P5Q motherboard, a modified Intel Xeon X5460 processor from China, and was on my way to enjoying the ultimate LGA775 goodness when Memtest86 kept freezing at about 21% completion.
I used the F2 SMT option to speed up the test by running on all four cores, and repeatedly it would crash during Block Move Test #7, right after one core had finished and was waiting on the other cores. Every frigging time…
Naturally, I went on a BIOS tweaking spree that involved lowering a whole bunch of frequency settings while simultaneously raising a whole bunch of voltages, until the settings looked completely ridiculous. And yet the cursed test continued to freeze at the same spot.
It wasn’t until after much searching and useless tweaking that I decided to run the test at its default settings, without SMT. And much to my relief, the system passed without a single problem.
Relieved that I don’t have faulty components or risk killing my system with needlessly high voltage, I went back to the BIOS undoing my tweaks, and a few tests later, I was pleasantly surprised that I could undervolt my CPU and memory, and still get a stable system. That’s one happy ending I’ll raise a beer to.