It sounds like you've tried pretty much all of the normal tricks that usually work. I recently helped a friend with his new Asus P5Q Pro. It was inexpensive but overclocked well and had a decent cooler on the northbridge similar to the P5B. I also like Gigabyte boards but haven't used one with the newer Intel chipsets so I can't help you too much.
My P5B was rock solid today with some new settings. Here's how it did running Prime small FFTs for the day:
This was with a bios setting of 1.55 volts for the northbridge. Here are the voltages and memory timings I used:
My E8400 (C0) runs OK but has always needed more voltage than most. The trips to 100C and beyond might have something to do with that! I think a fresh E0 stepping should run 4 GHz, Prime stable, with 1.30 volts or less.
Edit: I originally had the Static Read Control set to Faster but later learned that this can cause "unable to boot" issues when the FSB is at 333 MHz which is the default for an E8400. This changes the Performance Level as reported by MemSet. Not by much but enough to cause boot troubles with some combinations. At 533x7.5 on the Faster setting everything was fine but at lower FSB MHz it was problematic.
Nothing too complicated but check out the memory write bandwidth.
Newer DDR3 chipsets with aggressively overclocked CPUs would be envious of a number like that:
I was using my mismatched DDR2-800 ram but both sticks have D9GMH chips on them. I designed a new SPD table that I used so they would match up better and hopefully cause fewer problems at boot up. My Reapers have been pretty decent so far at booting up so I borrowed some of the settings from them and from the net and did lots of testing yesterday. The 965 chipset doesn't use the EPP settings so I got rid of all of them and I turned off the CL3 and CL4 SPD settings since I want this board always using CL5. Unfortunately, after a bunch of hard work, I learned that the P5B ignores most of the information in the SPD tables within the memory modules.
I thought boosting tRFC up to 56 and tRC up to 32 might help at higher frequencies. I created a SPD table with those settings as shown in my earlier post above but when I booted up, the P5B sets that to 42 so there is no advantage. You can adjust that with MemSet once you're in Windows but that isn't going to help any if you can't even boot up. My run above was with tRFC = 42.
I tried SPD timings of CL5-5-5-15 or -18. The P5B only sets CL5 and ignores the rest. Those two settings both end up as CL5-6-6-15 when running MemTest86+ or booting up into Windows. This might help explain why this board sometimes has problems booting up. It ignores most of the information in the memory SPD tables.
The modified SPD table I created is working OK for me today but with the P5B, tomorrow is another day. No lock ups recently other than ones I've deliberately tried to create. I might play some more tomorrow to see what else I can learn. I'll also post the SPD table I created if anyone wants to have a look at that. I'm not yet sure if it is helping me in any way since I had to use manually set timings when testing and most of the info is ignored by the P5B.
Edit: Resume from stand by is working 100% with these settings.
Reply With Quote
Bookmarks