Q6600 4th core keeps failing in Prime95: need help

ApocNarok

Distinguished
Aug 16, 2008
15
0
18,510
My rig:

Asus Maximus Formula X38 Mobo
Q6600 currently trying to OC at 3.4Ghz
TRUE with Scythe Fan SFF21F

OCList.jpg


I set everything at it's lowest possible setting, except the NB voltage (1.41V)

I'm adjusting the cpu vcore as you can see, and a funny pattern is coming out of it. The fourth core (Core #3) keeps failing on me. Any clues as to why this is happening and what I can do to make it stable at 3.4ghz? BTW, the VID is 1.3250 (sad I know) so is it just possible that I'll just need a really high vcore (>1.5?) I can get 3.3ghz (9*366) stable with about 1.416V vcore in windows. Suggestions would be appreciated. Thanks.
 

prow

Distinguished
Jun 17, 2008
81
0
18,630
what is the rest of your bios set at? multiplier, FSB stuff? did disable all the extra stuff ei...spread spectrum ect.....
 

BustedSony

Distinguished
Apr 24, 2006
576
0
18,980
First, try an 8X multiplier, the higher fsb and lower core frequently adds a bit more clocking headroom to the CPU.

Secondly the CPU has probably reached its limit. The slew rate of the weakest core can't be decreased by upping the voltage, since that just increases the saturation region.
http://powerelectronics.com/power_systems/test_measurement/test_saturation_voltage_efficiency/

My own G0 Q6600 maxes out at 3.29 Ghz with two cores failing Prime95 above that. The voltage before vdroop is 1.4125. after vdroop on the Maximus Formula that's 1.36-something. Increasing beyond that gets a lot more heat but almost no increase in speed.

Assuming all the CPU "features" have been disabled, including any memory speed boosters, try 3.33 Ghz. If that's stable, be happy.
 

ApocNarok

Distinguished
Aug 16, 2008
15
0
18,510
Yep, spectrums disabled, c1e and all that is disabled. loadline calibration enabled to reduce vdroop effect.

@prow, the graph shows 9*378 at the top (or 3.4Ghz).

I will try lowering the multiplier. I read in another article that bad ram could also be another factor? at 425FSB, you think the value I'm at right now (1.41 bios, a little higher in windows) would be a good starting value? Oops, I forgot to include the FSB termination voltage. That's also at the lowest setting possible. Should I up that too for stability?
 

BustedSony

Distinguished
Apr 24, 2006
576
0
18,980
We don't know what your ram is, but certainly the memory can limit the OC. Run it at 1:1 at rated voltage. Up the fsb one notch as I did, but I don't think it made a difference.
 

ApocNarok

Distinguished
Aug 16, 2008
15
0
18,510
Ram is ocz HPC Reaper 2X2GB. Running Vista 64 ultimate. Specs are 4-4-4-15 at 2.1 Volts. It is currently at 1:1. I manually set these timings myself. An auto, the timings are 5-5-5-15 I believe. I tried lowering the multiplier down to 8. It still doesn't seem to be running stable.

I upped the NB for the higher FSB, and also upped the VTT to about 1.45V. CPU vcore went up to about the same level. I get the strange message of hardware failure and that I need to reinstall windows. I forget what low voltage is making that happen. I think have a low NB voltage will give that message? I upped that until I could get log into windows. About 1.45 NB voltage in bios.

It keeps crashing within the first ten minutes of prime95. fourth core seems to fail sometimes too. I thought the 8x multiplier allowed more headroom for cpu? I guess it didn't work.

Right now I'm trying to get 9X370 (3.33 GHz) and it seems to be doing better. I'm still crashing, but I probably just need to add some more volts. 4th core alone doesn't seem to be affected with this small boost.
 

prow

Distinguished
Jun 17, 2008
81
0
18,630
i have p5q pro board and a Q6600, i got it stable at 3.6GHz, but did not like the vcore which was 1.55. at 3.4GHz same as yours with a 9*378. i had to set my NB 1.40 to get it stable, vcore 1.44. all other voltages set at auto. try to change your ram settings to 5-5-5-18, that ram setting was the key for mine. i also had issue with the ram set 5-5-5-15. give the 5-5-5-18 a try and see what happens.
 

prow

Distinguished
Jun 17, 2008
81
0
18,630
O hey do you have the 45nm or 65nm Q6600? i have the 65nm. are you running core temp or anything else? what are your temps hitting?
 

ApocNarok

Distinguished
Aug 16, 2008
15
0
18,510
65nm G0. Let's see, at max load at the highest voltage I was testing, they were 66,66,62,62. I was able to find a pretty stable vcore of 1.424 in windows for a 3.33GHz OC. I'm still testing for stability, but so far I haven't had to up any other voltages (VCC, NB, PLL, etc) which is okay for me. I made the mistake of leaving those settings on auto. luckily I didn't keep it that way too long. Lesson learned for my first overclocking attempt, but I'll probably still keep this rig even when Nehalem arrives.
 

iluvgillgill

Splendid
Jan 1, 2007
3,732
0
22,790
the Q6600 is ONLY produce in 65nm.prow

@OP those temp seems fine. but i would recommend you to do 8x multiplier so you can get a higher system bandwidth like i did when i realise i got a bum core and cant get a high enough OC at the FSB.
 

ApocNarok

Distinguished
Aug 16, 2008
15
0
18,510
@iluvgillgill:

Okay, I'll try that. Turns out I need 1.432V CPU Vcore in windows to get it pretty much stable. I've been running prime for about 30 minutes and it looks good so far... I'll plan to run OCCT after prime95.

I'm currently at 9*370. If I switch to 8*416, what other settings will I need to adjust? Is it just the NB Voltage? Or the VTT? Or both? VTT is currently at it's lowest manual setting of 1.2 and NB is currently at 1.41V bios.
 

prow

Distinguished
Jun 17, 2008
81
0
18,630
keep your ram at 1:1 until you get everything stable. so with that in mind, if you go with 8* you will be stuck with 3.2GHz. which is a good OC. i would go with NB@1.40v and vcore at 1.40-1.42~ish. all other voltages keep on auto unless you have the specs in front of you. you dont want to fry your board. also, it should have changed on its own but if not, set your PCIE to 100v.
 

ApocNarok

Distinguished
Aug 16, 2008
15
0
18,510
Hey, so it's been about prime-stable for about 10.5 hours now. When I went to check on it, I noticed that the first core was threading a lot slower than the other 3. It was still at 100% load, but it took about 2-4 hours to finish its last test. It didn't fail prime95, but I stopped it, and ran wprime the small one. It usually finishes at ~12.3 seconds. but it went up to ~16sec. I was confused, but when I turned the comp off and turned it back on, it was normal again. I checked the temperatures, and the highest it ever got throughout the 10 hour test was 63,63,60,60. Hmmmm I'm confused, but I'm running it again to see what happens. Any ideas? I tried googling this up, and no one seemed to have encountered the same problem as me.
 

iluvgillgill

Splendid
Jan 1, 2007
3,732
0
22,790
to check CPU stability - SMALL FFTS
to check mobo+ram stability - Blend test

those difference between core is normal. they are never in sync because somne works fast and slowing. but very tiny difference, but over a long time(10hrs like your) you can see the difference yourself. its nothing to worry about.
 

ApocNarok

Distinguished
Aug 16, 2008
15
0
18,510
yep, I'm running the small FTTs. But I still don't know why my wprime test went up 4 seconds from 12.3 to 16. I ran it multiple times and it was still around 16. Huh, well I guess if it's stable, then it's stable.