Q6600 Overclock

WingZero30 · 3 Dec 2010 at 23:28

bainbridge said:
Yeah thanks for your help mate, purring along at 3.6Ghz as I type

Had to go right up to 1.5v (1.5075 in BIOS) to run stable at 400x9 under load, anything below this would result in a crash, though I was able to boot to windows @ 3.6Ghz from 1.45v, anything below this would BSOD after posting.

Regarding the Northbridge all that was needed was a 1 increment nudge up in voltage from the default 1.25v to 1.33v. Before I did this I was unable to get stable under load with the clockspeed above 370mhz on a x9 multiplier.

Looking at your '£100 overclock thread', your VID is 1.2875v which is not very far from my 1.2625v. As you are on Abit and I am on Gigabyte, I think my results are likely to be different for [email protected].
Is the 1.5v the load voltage in cpu-z?

C64 · 3 Dec 2010 at 23:48

I'd run intel burn test before running prime but keep an eye on temps intel burn test will more than likely be +10c over prime temps.

Saves time really as if it gets through intel burn test passes it's more likely to pass prime.

I think on many of these chips there's a huge bump between the voltage required for 3.5ghz and 3.6ghz.

Have you tried 1.41 volts and 3.5ghz ?

setter · 3 Dec 2010 at 23:51

C64 said:
I'd run intel burn test before running prime.

+1, then run prime blend for a longer test, small fft could run for hours, but blend can fail in seconds if things arent set right.

MeatLoaf · 4 Dec 2010 at 00:16

Sean_UK said:
I'd be suprised if you need anywhere near 1.500v BIOS for 3.6ghz stable. To give you an idea my former Q6600 managed 3.7ghz with around 1.4v under load, and brothers Q6600 is 3.6ghz stable with only 1.360v.

Mine needs 1.35v in BIOS to get stable at 3.6Ghz

setter · 4 Dec 2010 at 00:19

I needed 1.4v for 3.6ghz, (1.2750 vid, middle of the road cpu tbh). 1.512 at 3.8ghz 8 hours prime stable, albeit the cpu and cooler were lapped, 2000 rpm fans on the cooler and a scythe kama bay with a 2000 rpm fan for extra intake in my old eclipse 62 case.

WingZero30 · 4 Dec 2010 at 01:42

C64 said:
I'd run intel burn test before running prime but keep an eye on temps intel burn test will more than likely be +10c over prime temps.

Saves time really as if it gets through intel burn test passes it's more likely to pass prime.

I think on many of these chips there's a huge bump between the voltage required for 3.5ghz and 3.6ghz.

Have you tried 1.41 volts and 3.5ghz ?

setter said:
+1, then run prime blend for a longer test, small fft could run for hours, but blend can fail in seconds if things arent set right.

-EIST and C1E disabled as usual
-Anti-virus disabled
-Chose Custom Stress test and inputted free physical ram.
-These GFlops values are the best I could get. They are with in 70% of the theoretical maximum Gflop value of 16x3.4 = 54.4Ghz
-Ran 20 passes of IBT and temps do indeed increase by 8C more especially on core 0 and core 1 as they run hotter anyway.
-Just upped the vcore by 2 notches more before running the test as I know it places way more stress on the cpu compared to Prime95.
-Surprisingly the load voltage still stays the same at 1.312V and even fluctuating to 1.296v lol.

So it can still be seen that vdroop is especially high. However the way I see it Intel Burn Test is more of a cyclic stress test program where it places the cpu under a lot more stress during calculations and when an iteration or a pass is finished, the load is momentarily reduced and then it goes back up again and in the process more heat is also produced.
However Prime95 although doesn't produce as much stress or heat is more consistent and constant at 100% loading.

Iirc I remember reading one of the don's post where he mentioned that it is better to do stress test with prime95 first. So if you receive an error then you know you just need to up the vcore more especially in the case of small ffts. However with IBT because it increases both the stress and heat, you don't know really if your testing failed due to your cooler not able to cope with the heat or if the vcore needed to increase more. So in this regard I completely agree with him

.

WingZero30 · 4 Dec 2010 at 01:47

setter said:
I needed 1.4v for 3.6ghz, (1.2750 vid, middle of the road cpu tbh). 1.512 at 3.8ghz 8 hours prime stable, albeit the cpu and cooler were lapped, 2000 rpm fans on the cooler and a scythe kama bay with a 2000 rpm fan for extra intake in my old eclipse 62 case.

Ae these the load voltages?

setter · 4 Dec 2010 at 01:53

With p95, small fft is ok for cpu only, blend can be a hard one to get stable, due to it testing ram as well as cpu, small fft will generate higher cpu temps, (ibt or LinX are a bit higher again), blend often requires tweaking of north bridge and ram voltage, timings etc, nice clock mate btw, temps are a little higher than what i was hitting, but your cpu has a lower vid than my q6600 was, lower vid chips, for some reason tended to run a bit warmer.

EDIT:Link to a thread were i posted my voltages.

http://forums.overclockers.co.uk/showthread.php?t=17779323&page=28

WingZero30 · 4 Dec 2010 at 02:15

setter said:
With p95, small fft is ok for cpu only, blend can be a hard one to get stable, due to it testing ram as well as cpu, small fft will generate higher cpu temps, (ibt or LinX are a bit higher again), blend often requires tweaking of north bridge and ram voltage, timings etc, nice clock mate btw, temps are a little higher than what i was hitting, but your cpu has a lower vid than my q6600 was, lower vid chips, for some reason tended to run a bit warmer.

EDIT:Link to a thread were i posted my voltages.

http://forums.overclockers.co.uk/showthread.php?t=17779323&page=28

Nice results. Judging by the thread even the earliest B3 seemed to perform very well. Does Asus P5Q Deluxe have 16 phase voltage regulator. I was reading that a larger phase regulator mean that your overclock wil be more stable, less heat will be generated, voltage will be distributed more efficiently to cpu and less vdroop will occur.
All the people who possess the P5Q Deluxe don't seem to report large vdroop unlike in my case. My gigabyte mobo has only 4 phase voltage regulator hence may be this is the reason why I am experiencing large vdrop/vdroop even though it is also supposed to be one of the best P45 mobos for overclocking the Quads.

I think I may be able run [email protected] prime stable though with higher coretemps. However with IBT I am sure the core temps will easily reach 75+C and thats where watercooling arena comes in.

Is there anyway of reducing Vdroop other than enabling load-line calibration or using pencil mod about which I have no clue?