Clock Twister Question............
I noticed a strange phenomenon when attempting to overclock my new ram today. My current settings are as follows.
FSB: 450
STRAP: 333
Clock Twister: STRONGER
PL 7
VNB 1.57v
VDimm 1.90v
The ram in question is OCZ 2x2gb Reaper 1800mhz. I was attempting to set my FSB to 460 to gain another 100mhz out of the CPU and another 40mhz out of the ram. My ram timings are set to stock 8-8-8-27-2N with the Trfc set to 72 My system has been totally stable. I have let prime95 run for as many as 5 hours amd also for at least 3 hours after tightening the Trfc from 88 to 72.
The rather high NB voltage is keeping me from wanting to push things much further, btu today, I set the FSB to 460, bumped the CPU Vcore a few clicks, and raised the VDimm to 1.94 to see if it would run 460. I also loosened the Clock twister form STRONGER to AUTO. I made it through 2 memtest passes without error, but I would get a rounding error in Prime95 after only 5 minutes. I am worried about adding any more voltage to the NB, so I realize that may be holding me back. Here is where I get to my point. \
When I run an Everest bandwidth and latency benchmark, no 2 runs are exactly the same, there is a small variance, which I assume is normal, but when I tested the bandwidth with the clock twister set to AUTO, the scores, on average seem better than the scores on STRONGER. The Read and Latency scores are the only ones to improve for sure, but the Write and Copy scores stayed the same. For the heck of it, I also tried STRONG, which I thought would perform about the same as AUTO, and it did.
I guess, I will leave the Clock Twister set to auto if it is still stable. I cant figure out why the scores would improve with the Clock Twister set to Auto or Strong as compared to Stronger. I am running Bios ver 0503. If the performance is better, I probably wont get to drop that NB voltage, which was one of the reasons I tried the Auto setting in the first place.
For the record, I also tried MODERATE, and that dropped the Read score 300-350mb/sec and killed the latency score.
Anyone seen this before?