No. Not really. The (really small) increase in memory i/o detected by a really memory benchmark will not be noticeable in any real world benchmark. And it may contribute to system instability.
Thanks for the answer.
I realized that tiny increase after doing some benchmarks with Sandra and AIDA64. I guess I'll set back the default timings (6-45) because of what you said, it may cause instability!
I've been doing some tests and right now I'm running Orthos to see how low can I go with Vcore.
The current settings are:
CPU @ 3200MHz @ 1.475Vcore @ FSB400
Memory @ 800MHz @ 2.1Vdimm @ 4-4-4-10 (default for the rest of the timings)
I can run them @ 4-4-4-9 but I guess there's no such big increase in performance.
Btw, one question. What's best: lower timings or higher speed (FSB) for the memory ?
I've read: Intel = speed and AMD = timings »» is it true ?
For a Core2 system, lower timings will give you better performance. But again, it will take a really good memory benchmark. And again, the increase in memory i/o does not have much effect on real world performance.
Case in point - my first Core2 system: 1st gen E6600 (2.4 GHz), 2 GB DDR2-1000 Crucial RAM, eVGA 680i motherboard. At 3.3 GHz, motherboard memory settings were 5-5-5-15-2T. Worked them down to 3-3-3-7-1T. Gave a 7% improvement in memory i/o. Had no effect on system performance.
When I replaced the ACF7P cooler with a TRUE, I increased the memory timings to 4-4-4-12-1T and OC'd the system to 3.6 GHz.
For some reason (I suspect the relatively large amount of L2 cache), memory performance does not have a major effect of system performance. Oh, don't get me wrong. The difference between high speed "gaming" RAM and low spec "value RAM" is noticeable in a well tuned Core2 system.