overclocking Q6600 w/ GA-P35-DS3L, 3.4 ghz on air!

billiardicus

Distinguished
Mar 2, 2008
186
0
18,680
Specs:
COOLER MASTER Centurion 5 CAC
Antec earthwatts EA500
GIGABYTE GA-P35-DS3L
Intel Core 2 Quad Q6600 Kentsfield 2.4GHz (GO)
ARCTIC COOLING Freezer 7 Pro
Crucial Ballistix 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 800
BFG Tech GeForce 8800GTS (G92) 512MB
Western Digital Caviar SE WD2500AAJS 250GB
SAMSUNG Black 20X DVD+R
Windows Vista Home Basic 32 bit (OEM)

I've left the multiplier at 9x:

- I started with the mobo automatically chosing voltages. I jumped to 3.0 ghz fine.
- Then I set the vcore to 1.3V. Oced to 3.1, 3.2 and 3.3. Core temps were fine.
- Oced to 3.4ghz, and the system would crash on PRIME95
- Increased vcore to 1.4, and achieved 3.4 ghz. Core temps were ~60C stressed. Ran Prime95 for ~10 minutes without problems
- Set it to 3.5 ghz. Bios would check out, but Vista wouldn't load.
- Dropped back down to 3.4 ghz. Dropped vcore to 1.35. Ran Prime95 ~10 minutes and 3dmark06 (scored 14832).
Core temps were all below 60C stressed.

I think 3.4 ghz is about my limit. 3.5ghz is not stable at 1.4V and my core temps were 60C stressed at 3.4ghz@1.4v. I don't want to exceed 65 C....

Questions:
1. The vcore requiements seem to ramp up as you push the cpu harder. How do guys get to 4.0 ghz without exceeding the 1.5V spec limit? I know they're using watercooling, etc but right now it seems I'll reach the 1.5v limit before the temperature limit.

2. I've got my ram voltage set a 2.2v per the spec. And the ram multiplier set at 2.0 for the 1:1 synchronous mode. I haven't messed with any other ram setting (couldn't find any other setting). Anyway, my RAM is now underclocked at 755 mhz (9*377=3.4ghz, so 2*377=755) instead of the original 800 mrz.

Thoughts?

This is my first build, and first oc. I've tried to do my homework, but don't expect me to be too smart just yet.


 

Evilonigiri

Splendid
Jun 8, 2007
4,381
0
22,780
1.) The voltage requirement is exponential as the cpu speed goes up. For example, if you need 1.3v for 3GHz, you may need 1.42V for 3.2GHz. You'll need to stress test for at least 6hours before declaring you load temps. 10min is not enough. To hit 4GHz, you'll need more than 1.5V.

2.) That's fine, just lower the timings accordingly and you won't see a diff in performance.
 

billiardicus

Distinguished
Mar 2, 2008
186
0
18,680
I read to stress test the cpu for 6 hours to make sure its stable, but I figured it would reach steady state temperature after a few minutes. I'm still trying to reach the right vcore. I got one error on prime95 3.4ghz@1.3375V after 15 minutes....
 

billiardicus

Distinguished
Mar 2, 2008
186
0
18,680
Okay...I was running Prime95, but the wrong test!
Running small FFT's makes it run hotter.

3.4ghz@1.35v = reboot after several minutes
3.4ghz@1.375=CPU Thermal Monitor 2 kicked in and changed my multiplier to 6x. This happens when the CPU gets too hot. The highest core temp I saw was 67C, but I've got the mobo set to protect them at 70C...
3.3ghz@1.3v=error after 1 min
3.3ghz@1.325v= no errors 1hr 25 min. Max core temp was 63 C.

I'll run this overnight, but it looks like I'm pretty much there. Still 3.3ghz w/ a $25 Arctic Freezer Pro 7 aint too shabby.
 

lschmidt

Distinguished
Sep 21, 2006
97
0
18,630
"Yes, the highest you can OC the Q6600 on a AC7 is 3.4GHz, so it seems about right."

I would have to say that's not necessarily true. I'm running stable at 3.2 Ghz and 1.25v with the AC Freezer Pro 7, after 15 hours of Small FFTs in prime95 temps are between 48 and 50 for all 4 cores (reported by coretemp). I was also able to boot into windows with no problems at 3.8 Ghz (and idle temps were in the low 30s), but I haven't done any stress testing yet at that speed.
 

Evilonigiri

Splendid
Jun 8, 2007
4,381
0
22,780

It's a rough estimate according to many people's experience. Just like how most people say the highest Oc for the Q6600 on stock cooling is 3.0GHz, it doesn't have to be true. Of course when we're talking about max OC, we're considering load temps, after all, I'm sure I can do 3.6GHz on stock cooling and get 65C idle.
 

billiardicus

Distinguished
Mar 2, 2008
186
0
18,680
I tightened the memory to 4-4-4-12
3.3ghz@1.325v
Ran Prime95 small FFT's for 2hr26min last night before one core reported an error. Hottest temps were 63, 61, 59, 63.

I'm almost there. I'll bump the vcore one more notch and run again tonight.

I've read that changing the multipliers doesn't effect max OC much, but if I lower the multiplier, I can keep 1:1 synchronous mode for the memory while still over clocking it.

example: 9x367 or 8x413 both equal 3.3 ghz, but for the memory 2x367=734mhz and 2x413=826mhz. Thus the memory is OC'd w/ an 8x multiplier but underC'd w/ a 9x multiplier.

Thoughts?
 

Thanatos421

Distinguished
Mar 26, 2007
549
0
18,990
Keep the multi at 9x. The less you stress your northbridge the better. You can get away with lower voltages for NB keeping the multi high and that leads to lower NB, and ambient system temps. Don't worry about having your RAM underclocked a bit, you won't see a difference.
 

billiardicus

Distinguished
Mar 2, 2008
186
0
18,680
Thanks for the advice. Admittedly I don't understand how the multipliers and NB voltages work other than the simple multiplication I was talking about.

Anyway, it seems that buying RAM that's faster than 800mhz is pretty much wasteful then?
- You want 1:1 synchronous mode, so DDR2 multiplier is 2.0
- You'd need to set the CPU host frequency >400mhz to get the ram to 800mhz, and that's quite a OC for a CPU.
 

Thanatos421

Distinguished
Mar 26, 2007
549
0
18,990
Well, DDR2-800 is so cheap right now, that there's not a reason NOT to buy it. I went with 1066 simply because I'm pushing a 480 FSB, and I didn't want to worry about OC'ing RAM and killing it.