1st February 2005, 15:48
| || |
| Originally posted by The Stilt |
When I started to test these modules, I did few preparations. I set multiplier to 14x instead of default 19X, raised VDDR voltage to 3.2V for starters, set latencies to 2-2-2-5-70µs-64µs (which were fastest with BH-5) and set FSB to 200MHz. Then I started Memtest and chose test #5 and saw it.... To fail over thousand times in the FIRST pass. I thought, ok one of those sticks are defective, lets try another pair. And guess what, same errors with new sticks. Then I tested the sticks one by one, but still got same errors. It´s unneccessary to tell how dissapointed I was, because I thought that these sticks contained the same stuff that the ValueRam that I had tested in late of 2004.
Actually I lied about changing multiplier to 14x, since 19x multi never worked of this [censored]ing piece of **** DFI
When I cleared CMOS, set same settings in bios, and tried Memtest again, guess what? It worked like charm up to 230MHz 2-2-2-5 without raising voltage from 3.2V. When I went back to bios, I noticed that I had forgot to set refresh mode to 64µs and left it to Auto instead. After testing this setting again, it was certain that setting refresh mode to 64µs totally killed overclockability of these sticks. On Athlon 64 platform overclokability was similary killed if tRFC was set to anything else than 15.