Q6600 + Asus P5N32-E Plus Woes.

[TheVrolok]

Just can't get it stable. Prime95 hangs mostly on during the first test (running Small FFTs), but I've gotten it to hang during the second, and then during the third when I gave up.

I've got all the standard settings enabled/disabled as per the guide thread.

I've been trying my FSB at 333, with a 1:1.2 ratio, then my last ditch effort was a 1:1 with 1333:667 (333/333) when I got Prime95 to get to the third test before the system hung. The RAM is A-Data 2x2GB DDR2 800. 5-5-5-18. Kinda cheap RAM, perhaps this is the problem? Originally I tried a 333 FSB with DRAM at 400 (so 1:1.2), and memory timings all at auto. I altered my vcore from auto, to 1.2, 1.225, 1.265, 1.3, 1.35 and then called it quits. My last effort (the 1:1 was with a vcore set to auto). Oh, and the hottest I ever saw a core before the hang was 54 degrees (using CoreTemp). Sorry if this is incoherent, it's a bit late and I'm sleepy. Will probably give it another go tomorrow, just looking for some advice/suggestions.

If it matters:
Vista x64
8800 GT
Arctic 7 Freezer Pro on the Q6600

Oh, and I'm a huge OC'ing noob. First time I've legitimately trying to OC. So be gentle.

 

[myocardia]

That RAM passing MemTest at stock speeds and timings?

 

[sutahz]

Small FFT's test your cpu more then ram, large test your RAM pretty good.
Run your ram at 1:1 (266FSB, DDR2 running at 533MHz).
In all my less-then-amazing overclocking attempts I don't recall prime95 freezing : ( Reported errors and rebooting my computer for me (got lots of that last time).
As myocardia said, run memtest86+ on your memory at 800MHz (1:1.5?).
I was going to say "then test your cpu by running prime95" but thats failing already (derr). So if your ram pasts memtest, that means your cpu is bad, you have a BIOS setting thats screwing up everything or your motherboard is bad.

 

[TheVrolok]

Just got back to school now where I've had time to run MemTest. Running it now, 14 passes of test 5 (block move - 64 moves) and still going with 0 errors so far. I haven't noticed any problems with system stability that might indicate a motherboard/cpu problem (aside from the possible inability to OC). I wonder if there's some sort of software issues with any combination of Vista x64/Prime95/Cpu-Z/CoreTemp that's causing a software lockup that's unrelated to the OC'ing. I might just OC it again and just run the system as normal - play Crysis - etc. and see if I notice any stability issues. It wouldn't be as intensive as Prime95, but it would help indicate that the system is still stable, no?

 

[Avalon]

Hmm...if your memtest results are good, I'd either say you have a weak OC'er, or your Mobo is having issues at that FSB speed for whatever reason. If you think your prime95 errors are out of place, though, I'd say go ahead and run some Crysis for an extended period of time. If Prime95 is crashing on the first test rather quickly, Crysis won't take much more time to do the same thing.

 

[TheVrolok]

I'm going to try using the Asus AI Overclocking for now just to test the waters while playing Crysis. I've heard that overall there is a disadvantage to using such an automated OC'ing sytem as opposed to the manual method, is that about right? The thing I don't like about the AI OC'ing is that it forces memory timing changes that I wouldn't otherwise make. My default RAM timings are 5-5-5-18, AI has set them to 5-6-6-20. I'm wondering what the consequences of these changes are - I don't know much of anything about RAM timings. Does increasing CPU speed while slowing down RAM really make sense - will I come out ahead?

 

[BlueAcolyte]

RAM speed has barely any difference... But those are weird timings.

 

[TheVrolok]

Originally posted by: BlueAcolyte
RAM speed has barely any difference... But those are weird timings.

That's what I thought about 'em too. If I use the 10% AI OC'ing everything seems fine. Played Crysis for like an hour without a hiccup. Tried turning AI to 20% and the machine won't complete POST. The POST screen will pop up, I'll get 1 line (listing the BIOS make/version) then before it would normally scan the IDE/SATA drives it just hangs. I clear the CMOS using the jumper and I'm back in business. Bad board, chip, possible? I didn't really plan on OC'ing too heavily either, but I was hoping I could at least get 3ghz outta this Q6600.

 

[Idontcare]

I had a DS3L motherboard recently that for the life of me I could not get to hold a 3Ghz (9x333) overclock on a G0 Q6600.

This was perplexing the crap out of me as I had setup 3 other identical systems by that point (within the same week) and they were all hitting 9x367 with ease.

Then I checked Vdroop during small FFT before the system bombed. The Vdroop was ridiculous, >0.2V. (that is not a type, 1.4Vcore was pulling down below 1.24V on load)

I swapped the PSU - no improvement. Swapped the CPU with another G0 - no effect. Same massive Vdroop.

Then I pulled the mobo. From the top down it looked fine, pristine, no obvious electrical issue. Flipped it over to see if I had any long solder pins that might have been shorting to my case and there was nothing.

But then I noticed a disclored region on the mobo backside near the CPU region. One of the VRM's had burned out. Turned the very dark board even darker for an area maybe the size of the dime.

I RMA'ed the board, dropped in a new one and OC'ed the same G0 chip to 9x367 on stock Vcore, Vdroop is maybe 0.06V (a typical value).

Cliffs: even slightly malfunctioning mobo hardware can make your overclock potential suck...check it over closely, not all bad electrical issues are visible from the front of the mobo.

 

[TheVrolok]

I just took my board out, looked at it, and remounted it. Tried to OC, same awful results. Couldn't even get into Windows @ 2.7. I heard rumors about downgrading my BIOS. So I backed down to the 0501 and bam. OC'ing is going well so far. Tried a few less aggressive OCs and everything seemed fine in my preliminary testing so I'm now spending more time with what I have set right now.

CPU FSB is at 1333, RAM is at 667. (so 1:1)
RAM timings are 5-5-5-15 (default was 5-5-5-18). Didn't really want to mess with the RAM much.
Running Small FFTs for 40 mins now with no errors.
Core temps have been sitting steady at 60/60/57/57 since the CPU went under load. Idles were all around 38-40.

These two things slightly confuse me, though. I set the vcore to 1.2625 in the BIOS to try to see how low I could get the vcore and still be stable (to save heat) but CPU-Z is reporting it at 1.328 both idle and under stress. It was also 1.328 with the same settings when I had my vcore set to auto. Is that odd? Secondly, with Prime95 running torture tests, shouldn't my computer be sluggish? I can still browse the web/check email/etc. and I don't know too big a slowdown.

 

[TheVrolok]

Just wanted to update:

Running my Q6600@3.0 GC on air (A7FPro), idling with the hottest core at 36 degrees and the hottest core under load is 57. I keep my ambient room temp rather chilly tho. For reference, my case ambient temps idle right now are 23 CPU (sensor taped to the CPU HS), 26 HDD (taped to the top of the HDD), 36 Sys (taped to the side of my NB HS - note: I remember touching this thing after I first set my system up before OC'ing/I had all my fans set up and man it runs hot). Obviously these aren't the temps of the chips, but it gives you an idea of the ambient case temps near the sources of heat.

Thanks ATOC for all the help. I'm very satisfied now knowing that I could probably get 3.2 easily on air without overheating the chip.