Memory OCing... would like some opinions

Bluefinger

Distinguished
Mar 10, 2006
531
0
18,980
Right... this weekend i plan to spend on getting my pc's memory OC'ed a little so to bring up the performance up a little. At the moment, i have a paired set of Corsair XMS Dual Channel PC3200CL2PT 1GB modules running at the standard 3-4-4-8 2T timings set on the mobo, and i know it can run better, but before i do so, is there anything i need to know beforehand? I already have memtest, but is there any other software i require just to make sure i have stable settings? Its obvious i'm a little new to memory OCing, though i have OC'ed GPUs and CPUs before, but here's what i want to try and achieve:

1)Firstly get the memory to 1T and 2-3-3-6 mem timings.
2)Adjust the CL to allow the memory frequency to be increased.
3)Get it to run at a stable level (very important... I can't stress enough on this point )

So... any help? Links to appropriate guides and opinions appreciated.
 

ChipDeath

Splendid
May 16, 2002
4,307
0
22,790
Leave command rate on 1T. I recall reading an article somewhere that showed you needed a HUGE amount more speed to make up for 2T timings - something like with 2T you needed 230Mhz to perform the same as 200Mhz with 1T. Maybe not quite that much, but it was suprisingly large.

Aside from that.... There's probably not much point going to CL2 and immediately relaxing it so you can shoot for more mhz... why bother?
 

bront

Distinguished
Oct 16, 2001
2,122
0
19,780
I have the 512 modules of the same memory, you can check out my journal thread on overclocking. It has a few good articles in it linked, and I'm posting more as I find them.

Suposedly 1T vs 2T can result in as much as a 30% throughput increase in memory, but mostly for synthetic benchmarks. As far as how much it will effect games and such, that's more along the lines of 5-10%, So 216 should be about what you need to make up the difference in that timing, while the other timings can have 2-3% adjustments.

I'm testing my memory at 3-3-3-8 1T overvolted at 2.8 (Stock is 2.75, and I think 2.8 actualy does 2.85 with these chips, but I'm not sure) at 250 (DDR500). 2.5-3-3-8 1T only passed Prime95 for 5 hours, which is pretty good, but I want at least 10. However, there is some luck in this, so you may have more play, or less play.

Be sure to throttle down your CPU before you toy with the memory, and if your memory multiplier is on Auto, set it to 1:1 before you use the system clock to tweek.
 

Bluefinger

Distinguished
Mar 10, 2006
531
0
18,980
Thanks, i'll be sure to keep an eye on the journal. At the moment, my memory voltage seems to be at 2.6V, so one of the first things i'll be doing will be pushing it up to 2.7V at least and then trying the 1T setting at 3-4-4-8 timings at 200Mhz.

EDIT: Right... i tweaked the RAM to 2.7V and then got it to 1T with 3-3-3-8 timings... and it ran stable. Tested it on Prime95 for about an hour and a half, and no errors. Then i tweaked it more to get the memory timings to 2.5-3-3-7, on 1T, and it also was stable, and then made sure by repeating the one hour and a half session on Prime95. Tonight, i'm gonna leave my pc on for the night with the tweaked mem settings, but also with the CPU OCed, and do a 12-hour stability stress-test to make sure it will work.
 

denster

Distinguished
Aug 18, 2005
158
0
18,680
That is very interesting, because I have the same ram and everytime I try to tweak it to 1T, I end up resetting CMOS...

My best efforts have left me at AUTO RAM settings for a superb OC of 2.2Ghz to 2.7Ghz!!! :)
 

denster

Distinguished
Aug 18, 2005
158
0
18,680
By the way, now with WUSY's help, my RAM is at 3V and steady. You may be able to get a higher OC by boosting DRAMV :wink:
 

Bluefinger

Distinguished
Mar 10, 2006
531
0
18,980
I had the mem running at 2.6V auto... but at the mo, its running without errors at 2.7V 2.5-3-3-7 1T.... i may want to increase the voltage to 2.8V if i were to increase the frequency a little, but i need to benchmark the system now and then test the current settings for stability before i push for higher frequencies, if at all, because if the overall performance boost is high enough with these settings, i might just leave them at that.
 

Bluefinger

Distinguished
Mar 10, 2006
531
0
18,980
Thinking about that... yeah you're right... but then again, i've got a pretty good idea of what i want in my head at the moment, so i'm just trying to get my pc to inch its way there since its close enough already.
 

bront

Distinguished
Oct 16, 2001
2,122
0
19,780
I had the mem running at 2.6V auto... but at the mo, its running without errors at 2.7V 2.5-3-3-7 1T.... i may want to increase the voltage to 2.8V if i were to increase the frequency a little, but i need to benchmark the system now and then test the current settings for stability before i push for higher frequencies, if at all, because if the overall performance boost is high enough with these settings, i might just leave them at that.
The Newegg specs list XMS voltage at 2.75, so setting it at 2.7 or 2.8 is probably going to yield the best settings.

I got 2-3-3-6 T2 stable at 2.7 on stock speeds, T1 wasn't stable, but still booted, probably would have worked on 2.8.

I'm running 3-3-3-8 1T on 250 Mhz on 2.8V at the moment, and that's rock stable. (14 hours of Prime95 + a few hours of memtest).