Quote Originally Posted by nick2006 View Post
Thanks to Soapman 72, I tried setting manually Vnb at 1.61, and, so far, OCCT gave me 1 hour stable results.

But I'm concerned if by increasing that much the NB voltage, I'll eventually result in degrading the mobo. Please note that I squeezed between the 8800GTS and the mugen cooler an 80mm fan looking halfway down the northbridge heatpipe... But I dunno if this can reduce nb temperature considerably...

The downside of this overall effort is that Prime95 keeps giving me an error after 3-4 minutes... The core voltage according to CPUZ is 1.312. Is there a possibility of an error-free Prime95 with more core voltage ?? Or do I have to manually give some settings in other tweak options too ???

Thank you all for this enormous help to all of us !!!!!


Hi Guys,


First off, I want to make it clear that I run a Dual core CPU, and I am in no way calling myself an expert on overclocking quad cores CPU's. That said, I would try more Vcore (just for testing) My E8400 requires 1.33v (load) to remain stable while running prime95 for almost 6 hours. I stop the test there with no errors, but don't feel the need to test for 12-24 hours like some guys do. It is VERY hard on equipment when we run the kind of elevated voltages we run.

I would recommend that you bump up the Vcore and NB and run Memtest86 for multiple passes. Whlile trying to overclock (tighten timings) on my 1800mhz reapers, I have seen errors after 4-5 passes, all while the system would run prime95 and OCCT for hours on end. Some would still call the system stable, since those programs run without error. I set the ram back to recommended settings, and got no memtest errors, so THAT is what I call stable.

Once you know that your memory does not give errors, run prime95 (seems tougher than occt to me, at least for testing a CPU overclock) on large FFT, this creates the toughest load on the CPU. If it runs fine for several hours, you can move to the BLEND test. THis loads the ram (and Northbridge) much more while also loading the CPU. If that will also run for several hours, you are nearly done. It is now time to start reducing the VCORE or NB voltage, but only test one Item at a time, or you wont know which reduction caused your instabilty. P5E3 boards have VERY fine adjustment when it comes to VCORE, so I tend to test in 2 click increments. I never want to end up right on the edge of stability when it comes to adequate voltages, but I also don't want to run too much......you get the point, I'm sure. Do the same for the NB, btu you may not want to move 2 clicks at a time for that one.

NOW, If you get errors soon while running Prime95 Large FFT chances are you don't have enough VCORE, but you need to pay attention to WHICH core gives the error. if it is ALWAYS the same, you may be able to fix it by playing with the GTL for that core, but I have not had that poarticular problem, so I have not had to adjust GTL's on my board at all. Pay attention to your temps, but once you get stable with the Large FFT test, move to the Blend test. this involves the Memory subsystem much more, so you may still get errors on the BLEND test that you didn't get when testing with Laarge FFT's. Bump the NB till those go away.

I posted a question in another section about the Max safe NB voltage for this chipset, not alot of people seem to know where the safe limit is. I am not sure either, but it seems that a fan over the NB heatsink is a very good idea when running these kind of memory speeds. I found some recommended settings for X38/X48 NB voltages on Corsair's website, and for their 4gb 1800mhz ram kit they list the NB voltage at 1.71v!!! I thought that might be a mis-print, so I called them. They said that ACTIVE cooling would be required, but some boards would need that kind of voltage to run those kind of memory speeds. They also did say that many boards could be stable at lower voltages, but they recommended 1.61v to start for 4gb at 1800mhz, moving up slowly from there if stability still was not achieved.

At teh end of the day, Quad cores can be tough to get stable at 450mhz FSB from what I am reading, and even dropping the multiplyer may not fix the problem, since it is the FSB that is not completely stable in alot of cases. Also, don't forget, If you are doing alot of testing at lower multiplyers, if and when you decide to move up to a higher multiplyer, you will almost always need a bump in NB voltage to keep things stable.

I hope this helps!