Importance of tRD

Yes, tRD affects performance of memory pretty much second only to tCL.

It's an important subtiming that those of who tweak our RAM for best performance :)


I run tRD 8 + all pull-ins enabled (equates to tRD 7) @ 8x442 [5:6] DDR2-1061.

I found that was actually a bit faster than tRD 6 + all pull-ins enabled @ 10x353 [2:3] DDR2-1059
 
sardonic.n7 said:
Yes, tRD affects performance of memory pretty much second only to tCL.

It's an important subtiming that those of who tweak our RAM for best performance :)


I run tRD 8 + all pull-ins enabled (equates to tRD 7) @ 8x442 [5:6] DDR2-1061.

I found that was actually a bit faster than tRD 6 + all pull-ins enabled @ 10x353 [2:3] DDR2-1059
Click to expand...
you have any links for tCL tweaking?
 
tCL = CAS.

The only tweaking you need to do with it is lower it if you can stably ;)

Of course, other than certain nForce chipset situations, you always want bandwidth over timings for max performance anyways.

IOW, you're better off running DDR2-1000 5-5-5 vs. DDR2-800 4-4-4, etc, as a general rule.

Basically, first determine max CPU speed.
Then determine max RAM speed.

Do this independently so as to avoid both being too high.

I usually test CPU for max with RAM underclocked, & then underclock CPU & test for max RAM.

Once you know those values, you can then determine the best possible combination.

Example in my case: CPU do ~ 3.5 GHz 1.4v stable, max i'm willing to use for voltage.
RAM can do ~ DDR2-1080 5-5-5 <2.1v stable

So i can do:
10x350 2:3 (DDR2-1050) tRD 6 + pull-ins enabled
9x389 3:4 (DDR2-1037) never bothered testing, but likely tRD 7 + pull-ins
8x438 5:6 (DDR2-1051) tRD 8 + pull-ins enabled

Overall, i already mentioned what i found to provide the best performance.

Prioritize CPU speed first, always.
Then RAM speed.
Then RAM timings.
FSB speed can affect things as well...you'll have to test for yourself to determine which one works the best.
 
You must log in or register to reply here.
Back
Top