Weird result? Please enlighten me...

[TMonte]

I am running 2x512 PDP 6400LL in a Bad Axe 2 with an E6600. I was just messing around with the RAM and doing some simple benches with SuperPI. I currently have it running 9x380 (3.42) with the RAM at 1:1 and wanted to see how tight I could time them. With 2.3V, I can run 3-3-3-8...however, using 1M, I'm finding 4-3-3-8 and 4-3-3-10 to be quicker - not by much, but it's repeatable.

3-3-3-8: 15.02s
4-3-3-10: 14.93s
4-3-3-8: 14.84s

Why is that? I realize there are other timings I can't manipulate directly from the BIOS, and I've yet to try memset, but this doesn't make much sense. I figured CL3 would make a positive difference, not go in the other direction. It's cranking through 32M now just for fun...

Also, I may take it back up to 9x400. What's the tightest you've seen the PDP LL sticks go at 400+?


Thanks!
Tom

[TMonte]

32M just finished. Another interesting result...

3.6 (9x400) 1:1, 4-4-4-12: 14m 59.562s
3.42 (9x380) 1:1, 4-3-3-8: 15m 14.375s

~180MHz = 15s difference? I expected it to be more significant than that. Does the latency make that much difference, or am I making more out of this than I should? I do admit to have not generated a complete data set...just wondering what everyone else thinks.

[TMonte]

Just ran 32M at 9x400, 1:1 4-3-3-8: 14m 31.469s

Nearly 30s difference, I wonder if it will run on a lower strap...