Ummm... why?

This is all because Bobby told me perhaps it was my RAM limiting my o/c. If it wasn't for you man, I would never of thought of doing the following things:

Well, this is real weird.

First off, I'd like to update everyone with my o/cing adventures.

I have the following config:

2.4C 3318A Costa
P4P800 Deluxe
2x256MB OCZ PC3500 Basic

Now, when I first started o/cing, I went to 3GHz at stock vcore (250MHz FSB, 5:4 RAM ratio) and everything went perfectly. I then went on to 260MHz FSB, but I had to up the voltages slighty.

Then the fun all of a sudden stopped. I could get into Windows at 265MHz FSB, but there was no stability whatsoever. Upping the voltages to even 1.65V didn't provide perfect stability. Running at 3:2 RAM ratio with any sort of timings didn't help. 270MHz FSB could not even POST even at 1.7V.

So, I became sad () and stuck with the 3GHz at default vcore.

Until now. I realized that perhaps HT was crapping out my o/c. I decided to disable that... During my time of thinking, I also thought that perhaps the RAM was limiting my o/c as vcore didn't seem to help at all. So I decided to disable HT and also take out one of the 256MB sticks. Booted up at 270MHz FSB and 3:2 ratio, RAM at SPD... and 2.75V vdimm, and what do you know? Got into Windows fine and everything. So I therefore checked to see if it was disabling HT or taking out the RAM that helped. Seems like it was taking out that RAM. So I said, maybe I can now run the RAM at the 5:4 divider since I'm now running single channel, so I try. No amount of voltage lets me. So I say, well, then, 3:2 it is, but can I run at a lower vdimm? Lowering it to 2.65V wouldn't even allow me to POST. So seemingly, dual channel for me does not work at these high FSB even using a divider and neither does the rated speed at default vdimm. So I said, ok, let's try the next best thing, stick the RAM into a slot so that I can at least get 512MB single channel. Well, seems like that's a no go, not at any voltage. So, looks like I got one pretty bad stick, that will at least run, but at a higher voltage, and one plain crappy stick that won't even run, no matter what the vdimm.

Now for the weird part. I can run 3.12GHz, 260MHz FSB, dual channel DDR416 at 2.75V full stablilty, 24/7, yet I cannot run 270MHz FSB at even 3:2, dual channel and 2.85V vdimm.

Anyone have any thoughts? I guess I could run the comp like this, 256MB RAM at slow slow DDR360 single channel and 3.24GHz, but that doesn't seem right for what I paid for this RAM...

BTW, I'm typing this at 3.24GHz, HT, 256MB DDR360 SC. I have not tried a higher FSB yet since my RAM seems to be crapping out.

Thanks to anyone who can help!

Thanks for pointing me to this thread.

Running dual channel will always lower your o/c'ing. The problem is that when you run a single stick on the 875/865 boards, you lose dual channel which I think is more important than higher fsb. Almost all your programs and games will benefit with the higher memory bandwidth.

Running at 3.24ghz is cool and all but your overall system is probably slower than if you ran at 3.1ghz with dual channel. If you really want to get higher fsb's, then you need to buy better memory modules. Usually if you buy 2 sticks at the same time, more likely than not they'll be from the same batch. Dual channel kits are, for the most part, marketing bs that allows the companies to charge more.

Most people can get to 3.4ghz at stock vcore. Pretty much any memory will require 2.75v or more to be completely stable.

If I were you, I'd throw both sticks in and run at 260 fsb or 265 fsb. If you have the money, go with better memory but make sure you look at the IC's used in the modules.

I'm waiting on my Twinmos PC3700 w/4.3ns IC's.

-Bobby

Thanks for replying.

I agree that running dual channel is a lot faster than single channel. Sometimes it's as though I can feel the sluggishness.

I did the buy the 2 sticks at the same time from the same place so I had hoped they'd o/c close.

But still one thing I don't understand, is this problem sort of like the CH-5 not working well when the divider is set to 5:4 on these 865/875 boards? Because I can run DDR400 fine dual channel 5:4, yet I can do DDR360 single channel at 3:2 stably with both sticks in.

Hmm okay thanks... this is pretty close to the same problems. At least one of them. 18488 3dmarks is still as much as me... and i have 9800pro. :/ and u on 9700np...

Ok, how are your RAM timings? Faster RAM timings really get you faster performance. I think I got ~300 3DMarks from just lowering the timings from 2.5-4-4-8 to 2-3-3-6.

Also, I have my computer pretty tweaked out. My video card wasn't running as fast as yours, but I was running 369/660. I disabled all apps running in the background and also disabled a lot of unnecessary processes like sound and my meyboard/mouse software.

Also, I tweaked WindowsXP for better performance. I ran no theme, just Windows Classic and no background on the desktop, just a plain blue screen.

Also, I set the image quality in the drivers all down to fastest performance.

Also, I seem to be getting very good scores for my card, a lot higher than is average for my card, so I wouldn't hang on that for too long. But you can try the things above too.

Ok, seems like I've nailed down that the RAM was the problem. Ran at DDR434 and 2.5-3-3-7 which is the rated latency and default voltage and memtest86 just stops on test #5... Doesn't want to do any further, but I dont get errors, just... dies...

Hmmm... come more updating:

Ram 1 stick single channel 256MB at DDR433 2.5-3-3-7 at default voltage and it gets by test 5 fine, but stops at test 7... I'm now testing at DDR347 CAS2.5-3-3-7.

I get about 15000 in 3dMark01 with a Ti4200

What version of Memtest86 are you using? There's some problems with the earlier versions on our boards. I'm kinda skeptical about it though. There's been a lot of problems with test 5. I have pairs of memory that error out or freeze on test 5 but can run prime for over 18 hrs.

I can't run my memory at 3:2 at all. What bios are you using? I'm still at 1008 but 1010 betas are out now. I don't know if they fixed the mem ratio problems.

-Bobby

I'm running memtest 3.0. Thought that perhaps this was a bug, so I tried 2.9 with the exact same results.

I'm using BIOS v1010 beta at the moment. I just did a fresh install so I'm gonna go test memtest again.