Well, after digging into several searches and guides, I still don't seem to have an answer, just more questions. I've been through the Xbit article, and a few here too. This is my first OC system and I mainly game on it. Heres my setup:
E5200 3.7 ghz (296x12.5) 1.4v bios, 1.376v CPU-Z
Asus P5Q-Pro mobo
OCZ DDR2-800 Ram Its rated at 4-4-4-12 T1 @ 2.1v, but I have it on auto until I know what I'm trying to do. The MB picked it up at 5-5-5-13 T2 @ 2.1v.
Correct me if I'm wrong, but ram performance is a balance between speed (frequency) and timing (lantency.) On Core2 intel systems, bandwith is limited by the NB and FSB and can only be increased by high FSB speeds. The E5200 isn't know for big FSB speeds but uses the high multiplier for outstanding clock speeds. I think I have two options for improving my ram performance.
I can keep the same 3.7 ghz by lowing the multi and increasing the FSB. This will let me use more bandwith but still doesn't come close to stressing the DDR2-800. But I would be running asynchronous with okay timings.
Keep the CPU overclock as it is and run the ram synchronously at ddr2-592. I'm guessing underclocking like that will let me tighten the crap out of the timings.
Will having the ram correctly set allow me to push the CPU further without any more voltage, or have I already found the capability of the chip?
You timings are already pretty tight and to really improve the performance of RAM the timings are a big part. Changing them to 4-4-4-12 wont do that much. If you increased your RAM speed then you would probably have to loosen the timings so yea in a real world situation you probably wont get any noticeable improvement.
And I agree with what both guys said. I "tested" my e5200 on p5qpro with 2gb ram few months back, have a look. Maybe those arent perfect apps for such purpose but in a way they confirm the "little or no gains" theory(at least on this setup):
Crysis map loading times were also within 0.5 second from each other
Ram timings at low speed is 4-4-4-12, high speed is 5-5-5-15
Also, crysis map loading times were within 0.5 second from each other
No problem. Damn, I wish my 4850 could go that high
Well, I ran the benchmark for World in Conflict artifact free at that speed, but I've found a few maps in counter strike source that display some. Mainly when there's a lot of dynamic lighting on the map. So I'm backing it off bit by bit till that goes away. It was a Asus TOP and came at 680/1050, so there was a great start already. What really supprises me after reading how hot these got was I've been below 52 degrees on apps. When trying the catylast auto tune, it ran it up to 63, but the core speed was at 790 then. Of course that wasn't stable even though it passed catylast test.
After looking closer at your chart, I see that Crysis CPU has a full extra FPS with syncronous, slower, tight timings. With that in mind, I am trying to find 3.75 ghz stable with a lower multi and then get as tight as I can on timings. I tried 3-3-3-8 at 592 and it was windows stable but failed prime 95.
Since this is my first OC and I'm using it as an extreme learning platform, I'm going to push forward to see what else I can get out of this realizing there won't be any noticable performance gains. My biggest disappointment is the P5Q-PRO won't allow me to set the command rate to T1. Live and Learn.