Running RAM 1:1 still the best?

[biff]

I remember back in the socket A/nForce2 days that running the ram in sync with the FSB was actually faster than running the RAM out of sync at a higher clock. Does that hold true with current C2D systems or no?

With most setups that are over clocked we're usually forced to use the 1:1 anyway so I suppose this is a moot point but I'm just curious.

[t0m]

i have a 680i mobo, e6850 and some pc2-8500. my cpu and ram are 1:1 with 1480/740. my cpu is 9x370=3330mhz, and i have tried 8x416=3330mhz to get the ram up a bit faster, and it didnt make a difference. i have unlinked the cpu and ram and cranked the ram as high as i can get it (~1100mhz with sloppy timings) and it was about 1-3% faster on some parts of the memory benchmarks. not a very good trade for the extra heat/voltage required imo. also, at those speeds some of the other memory and cpu tests were slower due to the sloppy timings.

my system could be an anomaly though, this mobo has been rapidly deteriorating since the day it got here...i wouldnt be surprised to find out that its not acting like everyone elses lol

[Mars]

I don't have much experience with C2D/Q systems other than building them, and running them at stock speeds, so I'm hoping someone else will give a good answer for Intel setups.

For AMD (since the introduction of the hypertransport bus), since there's no front side bus, there is no such thing as running the ram in a 1:1 ratio anymore. Internally, memory runs on a divider from the total CPU speed, so every extra MHz gives higher performance.

[Spblue]

Well, for me, running the RAM at 4:5 gives much better test results in Everest than running it at 1:1, even with the sloppier timings. At 1:1, my RAM runs at 958Mhz at 4-4-4-12. At 4:5, the RAM runs at 1200Mhz at 5-6-6-15. Everest reports a good 10% advantage with the 4:5 setting. Even the latency is lower this way. For me it's almost 250Mhz faster though, so depending on your FSB you might see different results.

[Synth]

When using only 2 gigs of Firestix, I run a 5:6 ratio (465 CPU - 558 ram) and it's better than 1:1. Mind you, your results may vary, and your ram might not scale as high, so maybe not worth it for you.

[t0m]

ok just for the sake of curiosity i grabbed everest home (which appears to be really old) and tried testing at my above settings, then switching to 3:2 and default timings and everest shows a good boost in all 3 categories. the other benchmark software i was using (passmark performance test for 64bit) still reports a difference of +/- ~2%. i wonder if its just a crappy program? one would think it would notice the difference everest reports in my memory speeds (which is pretty big)?

what is the industry standard mem/cpu benchmark that 'everyone' trusts?

[Eldonko]

Super PI :)

[MpG]

As I understand it, C2D systems generally show benefits from faster RAM speeds up to about 1.25:1. Beyond, it's usually just the synthetics that benefit. Not really sure, but I'd imagine that Quad processors might not mind even faster speeds, maybe up to 1:1.5?

[Eldonko]

Dont forget it depends on chipset..

[enaberif]

grab Everest as it does memory tests and you can see the difference that 1:1 and up will make.

Intel relies on speed or appears that way.