EPP Profiles, Linked & Synced, etc.

[sabregen]

So, I've had my E4300 for just over 6 months now. I have gone through 3 boards, and 3 sets of RAM. I have just ordered a set of OCZ PC2-8500 SLi Certified w/EPP profiles, to replace my Buffalo Firestix PC2-8000 kit that I think my 680i A1 may have borked when it died, 3 weeks ago. I got the replacement 680i A1 back yesterday from RMA, and am awaiting delivery on the OCZ. I've not ownede SLi certified RAM before, nor have I used any with EPP profiles. As I understand it (and I have seen the BIOS option for it) I will just need to flip the option in BIOS to have the board run the RAM (fairly automatically) at 1066 5-5-5-15 @2T, correct? Now, since I've only have my E4300 for 6months, and before that, I had been on AMD platforms for over 9 years, I'm still gathering my information regarding being Linked & Synced. So far, I've been running my RAM on a mutiplier (effectively, since my CPU caps out at about 9x365) to run the RAM (the Buffalo Firestix, at this point) at somewhere around 560MHz 5-5-5-15@2T. What is the benefit of running Linked & Synced, lower timings? If I did that, my PC2-8500 kit (new OCZ, on its way) would run at or below (depending on how bad I push the CPU) 730MHz but at lower timings. What's the best Scenario? Linked and Syned at 730, or not and around 1100? EPP or no EPP?

 

[Betrayer00GT]

I have Micron D9 chip based PC2-8500 from OCZ (what you ordered, I think) as well as from Kingston (HyperX) that I am using in my machine. I was able to get Sandra Memory tests on:
1:1 settings, 733Mhz (w/ my 1466.7Mhz FSB) @ 4-4-4-12-1T & 16TRC (would boot at 3-3-3-9-1T, but not run Windows for >60sec w/o crashing until back at 4-4-4-12)
916.7Mhz @ 4-4-4-12-1T & 17TRC
1066.7Mhz @ 5-5-5-15-2T & 22TRC

The 916.7Mhz/1T and 1066.7/2T got me essentially identical SSE bandwidth for Int and Float, respectively. The 733Mhz runs were closer than I would have expected, but not good enough to convince me to use it. The latency tests were pretty close, as well. 733Mhz did not test well in Sandra, over all, but was really good if you just compare it in a performance/Mhz to the 916.7 result. Also, the voltage difference between stable (Prime95) at 916.7 vs 1066.7 (a couple tenths higher for 1T @ 916.7Mhz) led me to actually use the higher bus speed/higher latency combo. Here is the whole enchilada.