*Official Retail G0 Q6600 Overclocking Thread*

[Pacha]

Humm.. so you removed the thermal pads
The mosfets heatsinks are covering 4 PWM each. This will cause in all cases a loose contact sometimes on some mosfets, despite the tightening of the push-pins. This is due to 4 things:

- The PWM surfaces are never even at a same level. Making that will need very expensive manufacturing process and is useless because of the bending of PCB. This first issue is the most important and impossible to correct, unless you lap evenly all mosfets with a glass
- The mb PCB is not hard, it bends, always... less than 1mm, but it bends
- Fixing of the heatsink is made by two push-pins on the extremities only, so impossible to solder to middle of the heatsink when the PCB bends
- Finally, ASUS heatsinks are really light aluminium without any surface polishing or special care.

It's not without a reason that all reliable third party manufactures of optional mosfets cooling kits recommand thermal pads for cooling mosfets.
This is also the case on VGA cards when cooling memory chips: you need either separate heatsinks per module, or the use of a thermal pad for many modules under the same heatsink because of the always uneven surface between two chips

Hope you understand better now what I meant

I made the same mod as he did with my first and second Commando and I must say the contact is quite good, putting enough as5 for the sink to cover them all, the results seem correct.

Anyway, there are mosfets just right near the mosfets' sink and others a little further that doesn't have any default asus cooling ont it, so this installation may be enough to not worry about it

Or you may want to put a fan blowing on the mosfets near you cpu to feal at ease.

[jeeka]

ah, yes, ok I get you now.Totally makes sense. I, too, realized that the mosfets were staggered height and was wondering how much that would effect cooling. It's worth a try to put the thermal padding back on...but obviously mine is rolled up in some trash heap somewhere.....

Where would you recommend, if you can actually buy them, thermal padding?
The only stuff I see are thermal compounds and not physical padding kits...

Thanks!

[jonny_ftm]

I made the same mod as he did with my first and second Commando and I must say the contact is quite good, putting enough as5 for the sink to cover them all, the results seem correct.

Anyway, there are mosfets just right near the mosfets' sink and others a little further that doesn't have any default asus cooling ont it, so this installation may be enough to not worry about it

Or you may want to put a fan blowing on the mosfets near you cpu to feal at ease.

It can be enough for a C2D, but clearly not optimal. And it won't be enough on a quadcore with vcores above 1.45-1.50v, overclock will be limited

ah, yes, ok I get you now.Totally makes sense. I, too, realized that the mosfets were staggered height and was wondering how much that would effect cooling. It's worth a try to put the thermal padding back on...but obviously mine is rolled up in some trash heap somewhere.....

Where would you recommend, if you can actually buy them, thermal padding?
The only stuff I see are thermal compounds and not physical padding kits...

Thanks!

In france, at watercooling.fr, they have adhesive pads. In NA, nextag searches gives some shops too, maybe google it but they do exist. By the way, even thermalright products (no one can criticize their heatsinks top quality products, right?) do ship all their mosfets heatsinks with pads

[jeeka]

Sitting at 3800Mhz on a q6600 has not been an issue with the Ceramique applied, but I will look into thermal padding to get some contact with the shorter mosfets, the ones that the Ceramique probably won't have great contact with.

[Unoid]

WIth a 1.325vid q6600 what would the ideal 24x7 vcore be safe under good water?
I'm guessing 1.55? maybe 1.6? Planning on the chip lasting a couple years.

I'll be needing 1.5V likely just to hit 3.6ghz on my L804

[Utnorris]

Since everyone is talking about their vid, mine is 1.2625v and I need 1.47vcore to be stable at 3.6Ghz, either 450x8 or 400x9, doesn't matter. I have tried 4Ghz, 450x9, but even with 1.65vcore it was not stable, although I was just doing some quick testing. I am going to try and get 3.8Ghz stable, but I think the consensus is that anything over 1.55vcore (actual) 24/7 is not good for the chip regardless of how good your cooling is. Speaking of which, under OCCT and P95 both running, I don't exceed 62c. That's with an ambient of 24c. My batch by the way is L726A820 with a pack date of 8/21/07. I haven't found anyone else with that batch, so I do not have a reference on what I should be getting, but base on some other vid's that are close, I think I am doing pretty good.
Utnorris

[jb-]

Hey Utnorris, I've got the same vid...these are the settings I'm running at (this pic was with P95 doing blend in the background):



Small FFTs temps are around 65C with the current voltage, and upper 20s idle. I ran P95 overnight at 1.47v (bios) and it failed, but my NB voltage was only at 1.4 so that could have been the problem. I'm trying again with 1.45 NB and the next step up on the vcore and I'll let you know how it goes.

[Hans Gruber]

Clubit.com just went out of business. I bought by "GO" stepping from Clubit.com Is there any other online etailer that guarantees "GO" stepping processors or are the noobs just rolling the dice?

[Gen]

you can try TankGuys @277.99 they are guaranteed G0 steppings I believe, if not then contact them directly.

I would be suprised if there are any B3's still rolling around.

[ZenEffect]

Clubit.com just went out of business. I bought by "GO" stepping from Clubit.com Is there any other online etailer that guarantees "GO" stepping processors or are the noobs just rolling the dice?

pretty much every e-tailer is g0, listed or not. it would be extremely rare to get a new b3.

[Utnorris]

Hey Utnorris, I've got the same vid...these are the settings I'm running at (this pic was with P95 doing blend in the background):



Small FFTs temps are around 65C with the current voltage, and upper 20s idle. I ran P95 overnight at 1.47v (bios) and it failed, but my NB voltage was only at 1.4 so that could have been the problem. I'm trying again with 1.45 NB and the next step up on the vcore and I'll let you know how it goes.

Hey thanks for the info. I did get 3825 OCCT stable @ 425x9 which is pretty good. I did have to put my vcore to 1.55v (1.6v in bios), but my temps are just barely over 70c under full load (ambient is 28c). When I was at 3.6Ghz I had to be at 1.47v to be OCCT stable. Also, what is your cooling solution? 20's seem really good as I sit at 40c under idle @ 3.8Ghz. At 3.6Ghz I was idling at 35c. Oh, my NB was at 1.58v.

The only issue I am finding is that my ram will only run stable at the 333 strap on my Blitz board. I have some G-Skill coming in rated for 1000Mhz and I will see how they do. What are you running your ram at?
Thanks,
Utnorris

[Utnorris]

By the way here is a screen shot of OCCT with the 3.8Ghz stable.

Utnorris

By the way, how did you get real temp to work? I keep getting "driver not installed" error when trying to run it.

[jb-]

Instead of explaining my cooling setup; here's an awesome picture (joking, lol):

Lapped w/ 21-22C ambients (explaining our difference in temps since yours is 6C higher)

I have the G.Skill F2-8000CL5D, but I'm only running it at 800mhz and 4-4-4-12 timings right now. I haven't tested the bandwidth on that compared to 1000mhz yet. Also, it seems like the jump in voltage/temps on our VID from 3.6-3.7ghz to 3.8+ is more than it's worth...but then again I plan to run my setup all day for gaming so to each his own I guess.

I didn't do anything special to get real temp working.


Oh before I forget, that was one nice lapping Pacha!! I think that beats the lap...dance..I got at club la vela, lmao! Care to explain the method you used with the glass/paper?

[Pacha]

ok here are a few results of my L804A515, pcb starting with 2L7, very disappointing at the moment I must say :

Idle 3.3G 1.4V vcore :



so it is 1.325V as others L80X


a few results :

3.2G 1.40V stable
3.3G 1.40V not stable
3.3G 1.425V stable (with 1.3V vFSB)
3.4G 1.425V not stable (5 seconds OCCT)
3.4G 1.450V not stable (BSOD after 5 seconds)

I'm running with 1.35V on NB for all of these, raising it up doesn't help. RAM is OK. Temps on full and idle are all right, and load temps don't increase much with vcore as some people reported on high VID CPUs.
It's like I'm reaching the max at 3.4, I'm going to try with more vcore or other settings but it's definitely not a good chip.


Do you guys have any recommandations?


EDIT : 3.4G 1.4875V not stable, and 3,456G 1.5V BSOD during boot
seems like it cannot even reach 3.4G... such a crap

[vws]

Pacha, thanks for posting your results - they look pretty much in line with the L804A I tested earlier. I, too, could hardly stabilize it at 3.4GHz with 1.425v in BIOS (+ load line calibration enabled).

So when all 4 cores are fully loaded do you notice any major difference among their top temperatures? Mine showed about 2 - 3 C at most (@3.4GHz load), even though at idle the T difference was greater than that (don't know exactly why). If you see big gap in core temps you probably want to re-seat the heatsink. Otherwise it means the seating is already optimized.

And after you've done pursuing the top end, try to find the minimal Vcore that stabilizes the highest clock speed before the "Vcore big jump" kicks in. That would be the point where the CPU runs in the most efficient way (in terms of performance/watt). In my case, I found ~2.8GHz is that speed for the L804A. With a mere 1.175v the hottest core barely hits 47C while other cores < 45C, air-cooled. My guess is that better CPUs will have superior clock/voltage ratios for that point, which could be used as an index to illustrate a CPU's potential. Honestly, I'd use that optimal point as my 24/7 settings.

[Hans Gruber]

You need to crank up your northbridge to at least 1.5v and your vtt to 1.4v I think your northbridge voltage is what is really holding you back. Mine is set at 1.55v in my bios.

ok here are a few results of my L804A515, pcb starting with 2L7, very disappointing at the moment I must say :

Idle 3.3G 1.4V vcore :



so it is 1.325V as others L80X


a few results :

3.2G 1.40V stable
3.3G 1.40V not stable
3.3G 1.425V stable (with 1.3V vFSB)
3.4G 1.425V not stable (5 seconds OCCT)
3.4G 1.450V not stable (BSOD after 5 seconds)

I'm running with 1.35V on NB for all of these, raising it up doesn't help. RAM is OK. Temps on full and idle are all right, and load temps don't increase much with vcore as some people reported on high VID CPUs.
It's like I'm reaching the max at 3.4, I'm going to try with more vcore or other settings but it's definitely not a good chip.


Do you guys have any recommandations?


EDIT : 3.4G 1.4875V not stable, and 3,456G 1.5V BSOD during boot
seems like it cannot even reach 3.4G... such a crap

[Pacha]

Pacha, thanks for posting your results - they look pretty much in line with the L804A I tested earlier. I, too, could hardly stabilize it at 3.4GHz with 1.425v in BIOS (+ load line calibration enabled).

So when all 4 cores are fully loaded do you notice any major difference among their top temperatures? Mine showed about 2 - 3 C at most (@3.4GHz load), even though at idle the T difference was greater than that (don't know exactly why). If you see big gap in core temps you probably want to re-seat the heatsink. Otherwise it means the seating is already optimized.

And after you've done pursuing the top end, try to find the minimal Vcore that stabilizes the highest clock speed before the "Vcore big jump" kicks in. That would be the point where the CPU runs in the most efficient way (in terms of performance/watt). In my case, I found ~2.8GHz is that speed for the L804A. With a mere 1.175v the hottest core barely hits 47C while other cores < 45C, air-cooled. My guess is that better CPUs will have superior clock/voltage ratios for that point, which could be used as an index to illustrate a CPU's potential. Honestly, I'd use that optimal point as my 24/7 settings.

The load temps follow the same repartition as in idle, but they are rather closer one to each other (1-3&#176;C)
the temps are 43-45-42-42 at 3.3G 1.425V BIOS and 1.460V real (I have to tune my vcore VR)

So in fact it's even worse than I tought, because it's only stable at 3.3G with 1.46V!! what a

As temps are OK i'm going to let this one to 3.3 but with regrets, I'd have liked to pull 3.6G of it but it's a long way to go...


Why don't you try to pick another one instead of running at 2.8? That, is pretty low for a Q6600.
The point now is that apparently all the new Q6600 have 1.325V VID and are bad overclockers, so it'll be hard to find a good one to replace it.

[Pacha]

You need to crank up your northbridge to at least 1.5v and your vtt to 1.4v I think your northbridge voltage is what is really holding you back. Mine is set at 1.55v in my bios.

I'm going to try 1.4V vFSB and NB but without any good expectation.
And I'm not going beyond 1.4V vNB as I burnt my previous Commando's NB with 1.45V after 1 week. (real voltage is the same as BIOS but Commando is well known for doing BBQ with NB).

[Pacha]

I've just made some new tests :

3393 (377*9) 1.425V 1.4V vFSB -> OCCT stops after 5 seconds
3393 (377*9) 1.425V 1.4V vFSB 1.4V vNB -> same result
3400 (425*8) 1.425V etc -> same result

I guess it's the upper limit...

Plus 3393 = 377*9, and 377 FSB is pretty low imo to require more than 1.35V, with my E6600 I had no trouble running 412 FSB with 1.35V, and I had the same result trying 425 coeff 8 instead of 377 coeff 9, so I don't think vNB is the problem here. Just that this chip is garbage unfortunately...

[ChaosAD]

Pacha, my previous Q6600 had a VID of 1.325. To run it stable at 3.4Ghz (8x425) i needed 1.41v real for cpu and 1.56v real for NB. For 3.5Ghz i needed 1.51v real for cpu and 1.6v real for NB. I also had to increase FSB VTT some but dont remember how much exactly. My new Q6600 has a vid of 1.30. I need 1.39v real for cpu and 1.37v real for NB to run at 3.6GHZ (9x400).

[Oskaliber]

I've got my Q6600, it's L745B108 batch, anyone heard something about it? Haven't got rest of my hardware yet so can't test it.

[Pacha]

Pacha, my previous Q6600 had a VID of 1.325. To run it stable at 3.4Ghz (8x425) i needed 1.41v real for cpu and 1.56v real for NB. For 3.5Ghz i needed 1.51v real for cpu and 1.6v real for NB. I also had to increase FSB VTT some but dont remember how much exactly. My new Q6600 has a vid of 1.30. I need 1.39v real for cpu and 1.37v real for NB to run at 3.6GHZ (9x400).

Mh interesting. Thank you for this remark, so maybe mine would need that too.
What are the batches you have for these two quads? What vFSB do you use with your new quad?

Anyway, if it requires 1.55V to be stable at 3.4G it's not worth the overheating and the risk of burning my NB again for +100MHz, as I am stable at 3.3G 1.35V NB.

I don't catch very well why the quads need so much secondary voltages increase over dual, on my E6600 the vFSB didn't changed anything about stability, and I've seen other people reporting the same.
Maybe much more power drained and stronger signals are required on the FSB and so does the nb?
Btw, my MB capacitors whistle quite a lot while they didn't with my dual so I guess it's because this chip is a gas plant compared to the dual.

[jonny_ftm]

Mh interesting. Thank you for this remark, so maybe mine would need that too.
What are the batches you have for these two quads? What vFSB do you use with your new quad?

Anyway, if it requires 1.55V to be stable at 3.4G it's not worth the overheating and the risk of burning my NB again for +100MHz, as I am stable at 3.3G 1.35V NB.

Unless you madely overclock your memory, you use +4Gb RAM or your FSB is above 450MHz, usually, you won't need to increase vNB

Those voltages on the NB are clearly not appropriate for only the CPU
Also, at those frequencies, VTT above 1.30v won't help you as you noted

Many people still bump crazy voltages every where when they don't know what to do or what's holding them.
In your case, it is clearly the vcore limiting you. These L8x steppings act like B3 and early G0 chips. For a Q6600, I won't hesitate bumping 1.53v + loadline calibration if your temperature permits it. Intel specs are 1.50v, so with 10&#37; margin on 65nm chips you can go to 1.65v. Usually, you'll be limited to 1.53-1.55 with heat

I never heard of any Kentsfield reported deaths at those voltages and I ran my B3 at 1.53v for few months too at 3.6GHz

edit: I noted you have a commando, it is a very bad overclocker of quadcores and stability near and above 400FSB is a random dream

[SySTaXer]

My G0 Q6600 loses two cores every time I drop the multi to 8 and raise the FSB to 450 and over. It's been like this on both my P35-DQ6 and Maximus Rampage. I get four cores right back up when I switch back to 9 multi again.

It has been working just fine 9x419 for many months but could never get it to work under the same speed with 8 multi.

I am thinking my chip has a low FSB wall but is this the way all quads behave when they cannot sustain high FSB overclocks? They lose two cores temporarily? Seems a little odd to me.

[wyemarn]

I have a L804A248 with 1.325V VID. I have to admit its a bad clocker. Here are my stable settings:

3.08G (8x385) BIOS 1.25V, load 1.24V
3.20G (8x400) BIOS 1.30V, load 1.29V
3.30G (8x413) BIOS 1.40V, load 1.38V

Currently I'm running at 3.08G at 1.25V with temps around 60C load. My 2nd core is the hottest and 3rd is the lowest. Difference around 3-4C both idle and load. I find that upping MCH or VTT voltage doesn't help much in stability. Instead of failing OCCT or Prime, most of the time I will get blue screen if the CPU is not stable.