Overclocking Results/where do i go from here?

[dabiggoober]

I finally put together my new system. Components in my sig file. I overclocked the pc3200 memory last night to 219 FSB or 438mhz with a multiplier of 11 for a total of 2409 mhz. Its still running at 2-2-2-5.

Is this normal? I'm getting no errors even after playing MOH for an hour. My temps at full load are 47 degrees (C).

What should I do next? My understanding is if I increase the FSB much more I will need to increase the latencies of the memory? Like to 3-3-3-8???? In real performance will this be worth it? I mean more FSB but slower cycles to get things done.....


Also, I noticed last night after installing lots of updates, programs etc... after my clean OS install, I am unable to create restore points in Windows XP PRO anymore. Any way to fix that without repairing windows installation?

Thoughts and advice appreciated.

 

[rogue_jedi]

is that the pc3200 XL? if so you can probably get 225 or so fsb without relaxing timings. if you have to, i think relaxing them to 2-3-3-11 will grant the best performance apart from having low latencies.

i don't really use winXP restore points, so i don't know about them.

 

[dabiggoober]

is that the pc3200 XL? if so you can probably get 225 or so fsb without relaxing timings. if you have to, i think relaxing them to 2-3-3-11 will grant the best performance apart from having low latencies.

i don't really use winXP restore points, so i don't know about them.

It is the XL, thanks!

 

[IM [H]ard]

Heh, you made the front page man!...later

 

[Hurin]

According to a review (who's link I lost, but I jotted these down for my own use), here is what one particular pair of XL could do:

2-2-2-5 Timings: 200-215

2.5-3-3-6 Timings: 216-253.0

I am currently running my pair of XL at 2.5-3-3-6 @ 240. No problems at all. Default voltage and "turbo" selected under Abit's "GAT" memory optimization settings.

Keep in mind that the literature at corsair for these modules states that they were running at 2.75v. I haven't needed to increase my voltage. But I found that interesting.

Yes, I had to up the timings a bit. But that 25MHz in crease in FSB speed allowed me to get my CPU to 3.6GHz. Between the faster FSB and faster CPU, I think I more than made up for the slightly increased timings.

H

 

[uclajd]

What should I do next? My understanding is if I increase the FSB much more I will need to increase the latencies of the memory? Like to 3-3-3-8???? In real performance will this be worth it? I mean more FSB but slower cycles to get things done.....

I don't know Athlon boards, but can you switch the CPU-RAM ratio to 5:4 or 3:2? At least that way you can see how much headroom the CPU has. I understand that Athlon's tend not to OC as well as Pentiums. On Pentiums, sometimes it's worth running 5:4 just to get the higher FSB. But if your OC potential isn't high enough, it might be worth it only to test the CPU's ceiling.

 

[Parja]

I don't know Athlon boards, but can you switch the CPU-RAM ratio to 5:4 or 3:2? At least that way you can see how much headroom the CPU has. I understand that Athlon's tend not to OC as well as Pentiums. On Pentiums, sometimes it's worth running 5:4 just to get the higher FSB. But if your OC potential isn't high enough, it might be worth it only to test the CPU's ceiling.

Running memory asynchronously on an nForce2 board is a big no no...huge performance hit. Best performance on nForce2 is typically as high as you can get the memory without significantly sacrificing memory timings.

One piece of advice, though. At a 219 FSB, try running your memory at 2-2-2-11 or 2-2-2-12 timings. You'll find your memory bandwidth is actually higher due to the way memory timings are handled by the nForce2 chipset. Testing can be verified with a Sandra memory test, but at 219, one of the two timings I mentioned above should be fastest for you.

 

[Hurin]

Running memory asynchronously on an nForce2 board is a big no no...huge performance hit.

I have seen this repeatedly, and I believe it. However, can someone confirm that there is no performance hit for non-nforce2 boards when running asynchronously?

Thanks!

 

[uclajd]

Running memory asynchronously on an nForce2 board is a big no no...huge performance hit.

OK, but it is still a way to test a processor's clock ceiling without the RAM being a bottleneck. If going 5:4 gets him a higher clock, then he can get a RAM upgrade.

 

[dabiggoober]

Thanks for the help guys. I found out that although I have no problems running anything in windows or gaming, PRIME95 gave me an error in 3 minutes. I'll have to investigate further. I will also try out these suggested timings on the memory. Any other thoughts, please post.

 

[Hurin]

Well, when you get Prime95 errors, it's time to do one of four things:

1. up the voltage a tiny increment on either the CPU, RAM, or both. Try to determine which is crapping out by messing with your divider and taxing only one of them at a time.

2. Increase timings on the RAM (if RAM is the one crapping out).

3. Increase cooling. (not always very possible and not always helpful).

4. Dial back the overclock a small amount and retest.

Personally, I just had my OC listed below start crapping out on a hot day after 22 hours of Prime95. Yes, 22 hours should be fine. And some people poo-poo Prime95 as being any sort of test. But I just know that it will run forever with no overclock. So the OC must be doing something to affect stability. Anyways, I'm cranking up my fans to confirm it's heat related. After that, if it still fails, I'll turn off all Abit GAT settings. Then I'll dial back the OC if needed.

H