Is the VCore setting you quoted at BIOS or in Windows? Mine is stable at 3.6 but with VCore in Windows (cpuz) showing 1.408v. Also, what are your ram timings? AFAIK Tracers need 2.2v for absolute stability. Moreover, you are clocking your DDR2 800 very high, so at 1.9v they might be crapping out at full load.

Try Prime95 or Orthos with all the individual torture tests so you can isolate the problem. By using blend only, you won't be able to tell if the prob is with the rams or cpu. Once you've isolated and fixed the prob then only run blend for absolute stability checking.

How much did you raise the FSB? And how much extra vcore? Or are you saying that you upped the proc speed with no increase in vcore and vdimm from your current settings for 3.2GHz?

Another way to do this is to up the vcore to the max at current FSB setting, i.e., at full load see how much vcore the heat sink is able to handle without touching 70c. Once you've determined that, slowly raise the FSB until it crashes, then dial it down a bit. Best that you loosen the ram timings first, or at least raise the vdimm so you'll be assured the rams won't hinder your efforts.

That's not true, cos I was on air previously running 3.6 with a TRUE for 24/7, prime and OCCT stable.

IMO the P35 chipset is a good companion to the Q6600. And with DFI boards being legendary overclockers I don't see why not.

You have to test it for yourself. Set stock vcore, then pump up the fsb to see how far it'll go. I'm guessing 3GHz should be quite do-able. Anything more than that, it's your luck bro.

CM Hyper 212 = RM130.00
Arctic Cooling Freezer 7 Pro = RM140.00
CM GeminII = RM110.00

But if you want super quiet and want to OC, get CoolIT Freezone

---

Added on May 16, 2008, 3:57 amAccording to official specs http://www.coolermaster.com/products/produ...=detail&id=3096 the cooler itself only weighs 710gm. The rest of the weight is the box, screws, brackets for both LGA775 and AM2, manual etc. A 12cm case fan weighs about 120gm thereabouts. So at most, the cooler and attached fan will weigh around 830gm oni

This post has been edited by gregy: May 16 2008, 03:57 AM

Dunno about lapping, but for me when I was running a TRUE, the temp performance dropped slightly after lapping. That's why I tu-lan and switched back to Swiftech Apogee GT with a bowed base.

The way I see it, the best performance for lapping must be done on both surfaces, otherwise one flat and one convex surface won't achieve proper contact.

Wow, after hearing all the sad stories I'm happy with mine, VID 1.25v and can do 3.25GHz with 1.26v in BIOS

Maybe it's my board, a "GayGaybyte" P35-DQ6 LOL!

IMO it is easier to use 9x and a lower multi for everyday usage, easier to boot etc. A slight loss in performance won't be as bad as struggling with stability and heat issues.

Oops, my bad. I meant highest multiplier and lower fsb. LOL!

Not entirely true, one can pass OCCT but still fail Prime. The same can also be said, someone passing Prime may fail OCCT. It will all depend on where the problem lies. Is it the fsb, vcore, rams, psu etc... Best to do all the tests.

Open Prime95, look at the individual tests and you'll see that different tests will test different areas. For example, small FFTs don't test much ram, only concentrates on FPU stressing. If you fail this test, it usually means the problem is at the CPU side, either not enough vcore or fsb too high or NB not enough voltage etc. Only when you successfully identify and overcome this stage do you then move on to the next step.

IINM Large FFTs will test your CPU's overall stability with regards to vcore, voltage rail stability and heat dissipation. You'll find that with large FFTs the load temps are slightly higher here than in small FFTs. So sometimes when ppl say their load temp is such and such, you can't really say for sure that that is the highest temp, unless you run large FFTs for at least an hour or one round of Orthos (half an hour run, then analyse the chart).

Blend tests everything, so once you've successfully completed small and large FFTs then only try out Blend. Cos if you go straight to Blend and your system crashes, it would be harder to pinpoint the problem.

After completing a number of hrs of Blend, then just to be very sure go try out OCCT. If you can get through OCCT as well, then you can safely say you have a stable OC.

I dunno la, but that's how I do it all this while and this method has never failed me. Oh btw, before running any of the tests above I also run Memtest for at least an hour, just to be sure my ram settings are stable.

Some possible "diagnostics" for the following scenarios under load (my unscientific findings LOL!) :

1) BSOD - Rams not stable, either timings too tight or vdimm not enough or temps too high
2) Shutdown - Vcore or voltage rail unstable, could also be excessive vdroop
3) Hang without shutdown - Could either be rams or vcore, possibly borderline case. It might mean your voltage settings are close to optimum, just need a bit more juice

You are right, perhaps you should follow the advice of some OCers who leave Prime running for at least 24hrs so as to cycle through the day's ambient temps. This is good advice especially if you plan on running your rig 24/7 for folding etc.

Mine, running Prime at 1.39 vcore in Windows for 3.6GHz, typically gets around 60c for hot days and around 52c at night with the aircond running at 27c. But when the water gets hot over an extended period, I typically see an additional 4-5c on a hot day, and about 2c extra at night. My WC setup is not ideal, and the water does tend to get warm if I fully stress the PC for more than 5 hrs at a stretch, which I normally don't so no big deal. I'm a light gamer and usually don't game for more than 3 hrs at a go

IMO 3.6GHz on my 65nm Quad is about the most stable for 24/7 based on my own experience. Last time I even used to run it on a TRUE, temps were higher but still within 70c (but then again, I almost never run my PC with the aircond turned off). Perhaps it was stable also because I had 4 Vizo 120mm fans blowing from the side and a 100CFM 120mm Gigabyte fan (which I salvaged from my 3D Galaxy II WC system) running almost at full power, sucking out the heat from within. My TRUE also had 2 Silverstones in a push-pull config. Imagine the noise pollution!

Wow, 4GHz is good stuff. I could only push my ex Q6600 to 3.87 max for benching *sweat*.