Overclocking Q6600 healthy temperatures?

[VirtualLarry]

Originally posted by: Idontcare
To run a G0 stable at 3.6GHz with small FFT in Prime95 you need something around 1.5V or higher which is why the temperature shoots so high.

Only if you have a 100% crappy chip. Mine does 3.6 stable at 1.325v.

 

[Schmide]

Originally posted by: VirtualLarry

Originally posted by: Idontcare
To run a G0 stable at 3.6GHz with small FFT in Prime95 you need something around 1.5V or higher which is why the temperature shoots so high.

Only if you have a 100% crappy chip. Mine does 3.6 stable at 1.325v.

3.6ghz at 1.325v is not the norm dude, in fact I would call it an exception.

 

[Idontcare]

Originally posted by: Schmide

Originally posted by: VirtualLarry

Originally posted by: Idontcare
To run a G0 stable at 3.6GHz with small FFT in Prime95 you need something around 1.5V or higher which is why the temperature shoots so high.

Only if you have a 100% crappy chip. Mine does 3.6 stable at 1.325v.

3.6ghz at 1.325v is not the norm dude, in fact I would call it an exception.

I concur. I've got five G0's here, all lapped and with lapped tuniqs on them. To hit 3.3GHz all of them need right around 1.30-1.33Volts to be small FFT stable.

If you are getting 3.6GHz out of the same volts mine need to hit 3.3GHz then you have a very nice chip.

To call my five chips crappy based on your experience with one chip is not exactly a value-add use of the adjective "crappy".

 

[twjr]

I have a lapped G0 and lapped TRUE and to get 3.33 I need 1.45v. Mine is a little crappy but saying that any that don't reach 3.6 as crappy is probably a little wrong.

 

[kazuyakun]

I can't tell if people that are getting 3.6 ghz on less than 1.4 or even at 1.4V are trolls or got golden chips. Mine does it at 1.43V and the temps idles at 57C and loads at 85C. That normal? Sadly even with an aftermarket CPS9700LED the temps are considerably higher than most folks.

 

[Reven]

Sorry for my absence, here is an update on how things are going:

I did what you guys said and turned back the clocks and started again. I'm trying a 3.5ghz OC at the moment. Voltage is at 1.39 and I've ran Prime95 for 4 hours with no errors so far. Average 100% load temps appear to be 73-71C ( Cores 0 and 1 are on average 2-3 C hotter). The highest recorded temps are 79 for Cores 0 and 1 and 77 for Cores 2 and 3. It should be noted that seems to be the exception as I havent seen those during the many times I checked up on the PC over the last 4 hours.

All case fans are at max. Room temperature is 23C.

 

[Idontcare]

Originally posted by: Reven
I'm trying a 3.5ghz OC at the moment. Voltage is at 1.39 and I've ran Prime95 for 4 hours with no errors so far.

Is that small FFT, large FFT, or Blend?

For your purposes here (verifying stability of CPU at current clockspeed/temp/Vcore) you should be running small FFT.

 

[Schmide]

I give the chip 4 months...

 

[njdevilsfan87]

Mine idles at 45C and loads at 73C. This is just a general average. They can be +/- 5C depending on room temperature. This is at 3.6GHz, 1.35V small FTT stable.

 

[Reven]

I was using Large FTT as it says it produces the most heat. When I come home I'll switch it to small FTT if you wish.

It has been stable for 13 hours now. Highest temps were 81C and 78C. Again though, it seems to flucate a bit as when I left for work this morning it was at 71 and 69C.

My assumption is that if I can just survive 24 hours I can work on reducing voltage, which should in turn reduce the heat a bit more. If I'm not stable I can go down another 100mhz.

I've actually had the chip since 2007 and its served me well. I dont see why you'd feel it will die in 4 months, after all its not as if the PC is at 100% load the entire time and most of the time its under 75C. Even then, in games I dont think all 4 cores are used...

If it is your recomendation however, I could lower the FSB a bit to hit a 3.4 speed.

 

[elconejito]

I think he meant because you were pumping ~1.5v into it. I don't think it is too big of a deal, but if you see instability even after reducing voltage/heat then I'd worry. Until then, don't sweat it.

If you experiment around a bit with your speed and voltages, you may find that there is a sweet spot where you get lots of speed for very little voltage. On my G0 that spot was about 3.2Ghz. Anything above that, and it required lots of voltage increases. I don't remember exactly the voltages, but let's say it took 1.25v for 3.2Ghz and 1.3 for 3.3Ghz, 1.375 for 3.4ghz and 1.45v for 3.5ghz. So in my case, the extra 300mhz wasn't worth the extra voltage/heat. You speed and voltages required to get there vary from chip to chip. If you OC properly you can find out what they are for your CPU.

 

[Reven]

OK, but I wasnt pumping 1.5v into it. Max I ever even tried was 1.41. Right now I'm 1.39.

Hopefully I'll be able to reduce the voltage with 3.5, as I think I its stable now... I have decided that 3.5 should be my limit, I wont be trying anything higher, my main goal is to reduce temps now. I'm not that bone-headed

 

[Idontcare]

Originally posted by: Reven
I was using Large FTT as it says it produces the most heat. When I come home I'll switch it to small FTT if you wish.

Small FFT is for testing your CPU stability, large FFT is for testing your NB/IMC/FSB stability.

Unless they've radically changed Prime95, small FF will produce the highest temps (in particular test #3), not large FFT.

Blend is a nearly useless setting as the program just oscillates between crunching small FFT's and Large FFT's with the result being that neither your CPU nor your NB/FSB/IMC experience consistent solid loadings for the hours on end that are needed to tease out those "on the edge" instability issues.

Typically when using Prime95 as a tool for optimizing your OC you start with using Large FFT, drop your CPU multiplier, and start pushing your FSB up until you find the edge of stability per Large FFT (which at that point could mean you need more NB voltage or more Vdimm, etc).

Then you go back to stock FSB frequency, boost your CPU multi to its max/stock setting and start testing with small FFT while increasing the FSB. Unless you get to a FSB value that was in the region of your ram/mobo limits as uncovered with large FFT, when small FFT gives you errors you can rest assured the limiting component in your rig at that point is the CPU and it either needs more Vcore or better cooling, etc.

At this moment if all you've tested with is Large FFT then all you should be confident about is that your memory and northbridge/IMC are not unstable at the current FSB. Large FFT makes no guarantees or assurances of CPU stability. And avoid blend at all costs unless you really want to give yourself a false sense of OC stability in all departments.

 

[Reven]

OK, I'll do a test in small FTT tonight. Thanks for the heads up.

 

[renozi]

Originally posted by: kazuyakun
I can't tell if people that are getting 3.6 ghz on less than 1.4 or even at 1.4V are trolls or got golden chips. Mine does it at 1.43V and the temps idles at 57C and loads at 85C. That normal? Sadly even with an aftermarket CPS9700LED the temps are considerably higher than most folks.

a bit on the high side but I guess that's what you'd expect with 1.43V
I get 3.6GHz on 1.36v and I'm no troll. Just luck dude, just luck.


and @OP small FTT would tell you rather quickly if your set up is stable. I had to increase to 1.36v from 1.35v @ 3.6GHz to not crash Prime95. Temps were better at 1.35v though,

I get 35-37C idle and just about 60C load and it's summer here in southern California. Although I did just clean out my heatsink and fanS. and use about the size of a grain of rice for your thermal paste; less is more!

 

[Reven]

OK, I have been stable on Small FTT for 9:30 hours now. The test is continuing, if its still stable by the time I get back home tonight it will be near the 20 hour range...

 

[clubfoot]

Just out of curiosity turn off the side panel fan,..see if your cpu temps go down, that's been my experience with the Ninehundred.

 

[BonzaiDuck]

Originally posted by: Schmide
I give the chip 4 months...

Originally posted by: Reven
I've actually had the chip since 2007 and its served me well. I dont see why you'd feel it will die in 4 months, after all its not as if the PC is at 100% load the entire time and most of the time its under 75C. Even then, in games I dont think all 4 cores are used...

The G0 stepping is still a 65nm quad-core. As I recall, the Intel "safe-range" spec has an upper-bound of 1.50V.

I've had my B3-stepping twisted up to around 1.43+, and others -- still here and among us -- told me back in late '07 that you COULD risk frying the chip for going about 1.47V. [But the very honorable sage who told me this seemed to have more fried motherboards and memory modules, and still swears he never killed a CPU with his extreme overclocking.]

Now as to Reven's temperatures. You're peaking out at near 80C from the core sensors, which would make me a little uncomfortable. However, the thermal spec is based on the TCASE sensor, and the TCASE sensor should be reporting around 10C lower than the core-temperature average.

I'd have to go back and check to see how much the thermal spec of the G0 differs from the B3 -- the latter being around 63C. But I THOUGHT that the G0's thermal spec is closer to 72C. I KNOW that 72C is the thermal spec for Wolfies and Yorkies. You should check it out. The G0 has a TDP thermal wattage of about 95C -- lower than the B3.

I'd say that staying within the voltage spec and staying within the thermal spec -- you should get a lot more than four months -- likely more than four years.

But that's no certainty, and that's the risk we take, dude!

EDIT: UPDATE: Here you have it, Dude. Q6600 G0 spec summary

71C is the TCASE thermal spec, and the voltage range tops out at 1.5V.

 

[Reven]

OK...

Well I've been stable for over a week now, and my games have been working fine. Thanks for all the help, guys.

 

[Idontcare]

Originally posted by: BonzaiDuck
EDIT: UPDATE: Here you have it, Dude. Q6600 G0 spec summary

71C is the TCASE thermal spec, and the voltage range tops out at 1.5V.

Hey Bonzai, FWIW the Vcc max for Q6600 is 1.55V accounting for the allowed Vcc overshoot. The 1.5V you are looking at is the max VID that Intel would sell a Q6600 at.

Also I thought this Intel document on embedded apps was priceless in terms of defining, in Intel terms, exactly what they mean by TJmax and TCase as well as how to thoroughly characterize your system's thermal profile if you were a systems integrator.

I really like the tutorial on heat conduction, kinda nice for people to be exposed to that level of the details IMO. I should probably figure out a thread in the Cooling forum to toss this link into, I'm sure folks there would appreciate the info too.

 

[Shmee]

in my second computer, I have a Q6600 G0, VID 1.25. Got it stable at 3.4 GHz with around 1.425 or about vcore, and with Linx reached up to about 74 C. Using a Xigmatek dark knight on that computer. Thing is, I think the memory is unstable, cuz when running large FFT's it gives a failure, and windows memory testing also reports problems. Probably either the crucial going bad, or not playing well with the Gskill.

Anyways, with a good G0, I would expect about 1.4 V for around 3.4 GHz, mb higher depending on the board and the chip. Make sure you dont leave it on auto. Actually set a voltage. For 3.6, start with MB 1.45V, though may need more. Make sure the temps arent too high though, must have adequate cooling.

 

[Idontcare]

Originally posted by: Shmee
Thing is, I think the memory is unstable, cuz when running large FFT's it gives a failure, and windows memory testing also reports problems. Probably either the crucial going bad, or not playing well with the Gskill.

Shmee that could just as easily be your FSB/NB not handling the 378MHz FSB all that well, either to little voltage or too little cooling on the NB.

Easy way to rule out the ram is to leave your rig at 3.4GHz but in the BIOS force the ram multiplier to be "synch" or "DDR2-756" however it shows you the math. Basically clock it as low as you can while leaving the FSB as is and then test again for large FFT. If you still get errors then I'd investigate better cooling on the NB or bumping that MCH voltage a shmidge.

 

[Shmee]

Yup, that's what I did, I had it at sync, I even upped NB voltage, and left the timings very loose. I will try taking out the crucial though, and see if that helps.

Edit: Took out the Crucial, and it works fine now, even with the Gskill at 1008 Mhz.