Q6600 heat as a function of clockrate (9x266, 9x333, 9x400)

[graysky]

EDIT - this experiment was flawed. Leaving the core voltage set to auto allows it to vary and thus tainted the results. I didn't realize this until after the fact. I'll need to repeat these runs @ 1.35 that's fixed and report back.

Tests are in progress now and finished tonight or tomorrow afternoon so stay tuned; I'll just edit this first post of the thread with the constant corev data.

Thanks to IntelUser2000 and GuitarDaddy for pointing this out!

 

[IntelUser2000]

I suspect I can get it to finish if I bump up the core voltage, but I've seen enough

You didn't change core voltages to achieve higher clocks right?? If you did, then the results would be less interesting than it sounds like.

 

[GuitarDaddy]

Originally posted by: IntelUser2000

I suspect I can get it to finish if I bump up the core voltage, but I've seen enough

You didn't change core voltages to achieve higher clocks right?? If you did, then the results would be less interesting than it sounds like.

He didn't manually change the volts, but he let the board do it by leaving it on Auto.
My P5B-E raises the voltage quickly as you increase the FSB if left on Auto. I get a little chuckle everytime someone posts "I overclocked to XXXXmhz on stock volts, I left it on Auto" Stock volts are 1.325 set manually, not Auto.

 

[graysky]

Interesting, CPU-Z reported 1.304 for what that's worth. If I'm not mistaken, it doesn't go past the stock zone, only down when speedstep kicks in. I'll have to look at that. Too bad I didn't log the damn voltage.

 

[graysky]

I just dropped down to 2.4 GHz and the Vcore is set to 1.12 so you guys are right. I'll repeat these experiment tomorrow.

 

[graysky]

Sorry guys, even w/ the vcore set to 1.350 in full manual mode and with all the powersaving options disabled, the voltage will not hold constant @ 2.4 GHz (drops). Same is true for 3.0 GHz. Guess I'll give up on this idea

 

[lopri]

What was exactly the purpose of this post? Sorry it's hard to tell it by the title and edited first post. If it's about heat limiting quad-core overclocking, I definitely agree. I got a Q6600 a few days ago and although I'm happy with the result I got, there certainly is a room for improvement. The most limiting factors in quad-core overclocking is:

1. Heat
2. Motherboard
3. Uneven potential among cores

I settled at 3.30GHz @1.37~1.38V), which takes the load temp to ~65C in open-air test bed. I didn't put the system in a case yet, so I might need to lower it to 3.20GHz which is doable with 1.30~1.32V. While the CPU has no problem booting to 3.60~3.70GHz and I can even run benches but it's only because there aren't many benches that pushes all 4 cores to max. But if I run quad-Priem95, one of the cores will fail in less than 5 secs.

Motherboards and/or BIOS are also hugely responsible for quad-core overclocking. My impression overall is that it feels as if overclocking E6600 with this board when it first came out. FSB holes and mysterious behavior except a few magic spots. I guess it'll take some time the boards will mature over time. Vdroop is also huge on most boards with quad-core CPUs. On my board it's almost 0.07~0.08V between idle/load.

And my Q6600 has 3 good cores and 1 bad core. Unfortunately, the CPU in whole is only as good as the worst core it has. Individual (by disabling cores via BIOS) testing shows 3 cores are able to reach 3.60GHz with relatively low vCore/temps, but one core is severely lacking compared to the other three. I'm glad that at least it can do the clocks I originally aimed with this CPU.

All in all, I think many boards' are not ready for quad-core. My board can stabilize the quad up to 425FSB, but that's about it. (And it's known to be one of the better boards for quad-core overclocking) Various performance tweaks are either not possible, or even if it is, it doesn't seem to affect the performance at all. With dual-core CPUs, one could squeeze more performance out via BIOS tweaking under a certain CPU frequency. Interestingly though, I noticed that stressing memory and/or NB will greatly (compared to dual-core) increase the vCore requirement and temperatures - without much performance benefit. I think it has a long way to go until we have a truly 'performance-oriented' BIOS for quad-core CPUs.

 

[AkumaX]

my goal is to eventually own a q6600 @ 3.0ghz, so a thread like this would be interesting indeed

 

[graysky]

9x333 is very easy... 9x366 is as well.

 

[aigomorla]

i want to say anything up to 375 will be fairly easy. But as lori said, your temps climb up very very quickly.

This is the best my high class waterkit can do this sucker: 3375 @ 1.425 bios 1.37 actual
http://i125.photobucket.com/albums/p73/aigomorla/X3220.jpg


At 400fsb i need about 1.45V actual in bios. That means i need about 1.57V in bios.

My load temps climbed to high 60s. Thats something im not happy with 24/7


@lori my board has a stupid wall from 375-399fsb.

I dont know WHY. I cant boot at ANY voltage from 1.4-1.65 on those FSB's. But i can boot up at 400fsb @ 1.57V

BAH stupid 680i i might go pick me up a P35 Bearlake when they come out.

 

[lopri]

Hey, aigomorla - what is that app that you're using to stress all 4 cores? I'd appreciate if you can give me a link or something. It's pain in the a** to run 4 instances of Prime95.

 

[lopri]

Originally posted by: aigomorla
@lopri my board has a stupid wall from 375-399fsb.

I dont know WHY. I cant boot at ANY voltage from 1.4-1.65 on those FSB's. But i can boot up at 400fsb @ 1.57V

BAH stupid 680i i might go pick me up a P35 Bearlake when they come out.

My board has lots of FSB holes, too. But it's not as bad as what you're experiencing. If you give me your e-mail address, I can send you my settings just in case.

Oh and you might as well wait till X38 comes out and see how they do. (Granted I don't know whether X38 will support DDR2 or only DDR3)

Edit: Have you tried ClockGen to up the FSB in Windows?

 

[aigomorla]

Originally posted by: lopri

Originally posted by: aigomorla
@lopri my board has a stupid wall from 375-399fsb.

I dont know WHY. I cant boot at ANY voltage from 1.4-1.65 on those FSB's. But i can boot up at 400fsb @ 1.57V

BAH stupid 680i i might go pick me up a P35 Bearlake when they come out.

My board has lots of FSB holes, too. But it's not as bad as what you're experiencing. If you give me your e-mail address, I can send you my settings just in case.

Oh and you might as well wait till X38 comes out and see how they do. (Granted I don't know whether X38 will support DDR2 or only DDR3)

Edit: Have you tried ClockGen to up the FSB in Windows?

AHAHAHA yeah i know about the loading its a PITA.

Orthos can do it if you split up priority. I however use Bionic. Its a World Community Grid project like F@H. If your system is terribly wrong, it usually reboots your comp when you run it.

And if your overclock is windows only stable, you'll spit out invalid results. Currently she's a monster doing tons of work and im very impressed. At my current speeds im almost keeping up with dual clovers.


cant use clockgen also, cant risk a 2 hour crunch job going bad because i wanted to see if i can squeeze 150mhz extra out of her. Not worth it.

She's getting a major redue on her internals. RD-30 + PA120.3 should keep her nice and cool @ 3.6ghz