Monza, I'm very embarrassed to say it's an ECS G33T-M2 board that I used to bench all these. I got the board virtually free from a Q9550 bundle deal back in Nov. last year, when I luckily scored that L835B078 chip. According to my inquiry with the ECS engineers the board supports CPUs of designed power no greater than ~95W, which means it can not sustain highly OC'd quads through long stress tests. That's why I mostly resort to short benches such as OCCT standard (not Linpack) 1hr tests. Memtest86+ is fine, but with the generic BIOS (with no controls of vCore, vNB, vTT, GTL, clock skews, etc., i.e. mostly by raw settings) I am only able to test FSB up to 453 MHz and memory up to 545 MHz (in 5:6 ratio). But I do try my best to optimize whatever is available to me. The highest bench-able speed for OCCT is FSB 442 / memory 531. Benching at this speed the Q9550's droop ~0.07v off their VID values (the board automatically sets vCore by stock VID), but the three Q9550's that I decided to keep have all passed the OCCT 1hr benches.