My Q6600 overclock (yet another one)

[shamans]

Just recently got a Q6600 with G0 stepping, 1.325 VID and a Asus p5k-e/wifi-ap, Artic Freezer pro 7, Antec 900. (I've overclocked several AMD x2 systems before)

I can get 3.4 ghz stable with 1.4 vcore in bios (in cpu-z it's 1.36 vcore on load) with 1.4v for fsb strap (VTT) + load-line calibration enabled (decreases vdroop). Multiplier is 9, FSB at 378, Memory is 1:1. My limitation seems to be this slightly underperforming Q6600 chip I have. I can probably hit 3.6 ghz with 1.5 vcore at load but temperatures would start to hit low 60s at load.

So my questions are...
1) When 1.5v is the max "safe" vcore...is that in the bios setting or cpu-z at load?
2) Any other tips on increasing my overclock or lowering temps? Increasing my fsb strap to 1.4v seems to help. Would 1.5v be safe? What else can I do...
3) What's a safe operating long-term temp for Q6600?

 

[LOUISSSSS]

1.5vCore and you'll only get 60C @ full p95 SmallFFT load?

what cooler r u using & what r ur room temps in F?

and damn 1.325 is a HIGH VID for a g0 Q6600

 

[Idontcare]

Originally posted by: shamans
1) When 1.5v is the max "safe" vcore...is that in the bios setting or cpu-z at load?

BIOS...max safe voltage means the maximum the core will ever ever ever see. So you have to be careful that your board isn't miscalibrated and is overvolting.

Set your voltage to whatever in the BIOS (say 1.4V to be safe) and turn off EIST (speedstep) and C1E. Then in windows load CPU-Z and at idle check the Core Voltage. If this reads greater than your BIOS setting then your board is overvolting and you need to maintain room for this margin of error in the upper limits you set on your VCore in the BIOS.

Originally posted by: shamans
2) Any other tips on increasing my overclock or lowering temps? Increasing my fsb strap to 1.4v seems to help. Would 1.5v be safe? What else can I do...

Personally I wouldn't put more than 1.4V on my Q6600. I have five of them running at 3.3Ghz (9x367) with 1.33V or less and my loaded temps (small FFT) are in the high 50's to low 60's with lapped Tuniq. The only time I felt comfortable going to 1.5V on a quad was my B3 QX6700 when it was on vaporphase cooling (-40C).

There's just sooo much heat created by such high voltage and high clocks. Will you even notice 3.5GHz versus 3.3GHz performance-wise?

Originally posted by: shamans
3) What's a safe operating long-term temp for Q6600?

Anything <80C is "safe" provided the VCore is not greater than the VID. If you ratchet up the Vcore to 1.4 or even 1.5V as you are talking then your max safe temp is likely <60C.

 

[xx04201987xx]

Originally posted by: LOUISSSSS
1.5vCore and you'll only get 60C @ full p95 SmallFFT load?

what cooler r u using & what r ur room temps in F?

and damn 1.325 is a HIGH VID for a g0 Q6600

my G0 Q6600 VID is 1.3125 =
I'd say thats pretty high too...
well I'm running mine at 3.2(356x9) with 1.296V (in cpu-z)
and my ideal temp is around 30-35
I didn't try going higher but perhaps I should when I'm bored. =)

 

[DSF]

What are you using to measure your temps? When you say 'load', what is the processor doing? The fact that you're running at 1.4V and 3.4GHz and your temps are barely 60C at load strikes me as unusual.

By the way, you can hardly call a 2.4GHz chip that overclocks to 3.4GHz underachieving. Not every chip is going to hit 3.6Ghz.

 

[BoboKatt]

I have the same Q6600 G0 and same Asus P5Ke-wifi. I actually upped my oc yesterday from 3.2 to 3.4 (same FSB as you... using the 9x ad 1004 bios).

I know my CPU's VID is not as high as yours.. 1.2375 or something. To get a stable 3.4 GHz, I am at 1.325 or something in Bios... under load using Prime using CPU-z I see 1.304 or slightly higher. I could not imagine using 1.4 V... my temps would probably be insane.

With my current OC and 1.325v, my temps reach 60-62c on my TR UE... and my CPU is lapped. When I drop it down to 3.2 and volts accordingly... it's like 54c. I am still debating whether it's worth the extra 200Mhz to get all that extra heat/volts and power consumption.

Anyhow you can try taking your RAM out of the equation... raise the Strap or play with the options till you know your RAM is way underclocked - I set my PC26400 Patriot ram to 5-5-5-15 and all others on auto and V to 2.0 for the 4 sticks just to make sure and it was running at like 756mhz just when I was testing.

Drop your CPU to 1.35 or something... up the next two options in your bios under the CPU one by one level... just one. Leave all others on auto

 

[Syzygies]

Hi. I'm new to this forum, and I believe the implicit rules are this: Pledges like me have to goal-tend the load-line calibration issue for a couple of days, till they get sick of it and someone else joins.

Read everything on Anandtech that Kris Boughton has written. If you understood what you read, it correlates with any previous understanding of electronics or engineering (his Vdroop curves scream "Gibbs effect" to me), and you still want to enable load-line calibration, go ahead.

Before overclocking, my second Q6600 build ran cooler at full load than my first Q6600 build ran at idle. Luck, case, fans, mobo, stepping, but I also had to unlearn worshipping false gods. The copper Ninja heatsink inspires object lust in me, but the Thermalright Ultra-120 eXtreme bests it in reviews, before one attaches a pair of fans at full speed for push-pull, perhaps the best one can possibly do on air.

It seems there are many false gods once one gets to overclocking, too. The bulk of what's written online makes me think of gifted rock musicians who can't read sheet music. Kris Boughton clearly has a "classical training" and can read sheet music. I come from a scientific background but still winced at the information overload on the first few tries through his recent articles. Then I printed out a handful of articles from this site 2-up for an airplane flight, worked through them with a highlighter word-for-word, and overclocking really started to make sense. I was getting good results before because of my ample hardware margins, but I was making classic mistakes.

Worrying what someone else is going to think of your voltage readings, how you compare to their efforts, is worshipping a false god. These are just proxies for what really matters: How does your rig benchmark, and how are you doing on thermals? I have a third need: Stability for months at a time while running computations that can take weeks, while I'm on the wrong coast to press reset if the rig freezes up. My goal is to make 3.2 Ghz this stable on air while never exceeding 60C core temps, on my own stress tests that are more abusive than mprime (Prime95 for Linux). I'm almost there. My favorite test, having each core iteratively build GHC Haskell from scratch, blew the power supply on my friend's $10K 8 core 64 GB compute server, which had passed every other test!

Say "nomimal" a few hundred times in a row, until the word no longer seems from the English language. Then disable load-line calibration, up your nominal voltage a bit to compensate, and see if your stability and thermals improve.

 

[Tullphan]

:shocked:

 

[shamans]

A lot of good questions here. I'll try to answer them:

I'm using arctic freezer 7 pro. I didn't install the cpu heatsink/fan myself (the parts supplier I got it from put it on). I told them I wanted them to install the stock intel cooler and I'll install the arctic freezer 7 pro myself but they messed up the order. Anyway, I gotta go find a tube of thermal grease before I can redo it myself to make sure it's done right.

My temperatures are highly dependent of the cpu fan speed (obviously). I'm trying to get it to run at around 800 to 1300 rpms on load. It can go as high as 2700 rpms and that's where I get about 60C - 63C on load at 1.5 vcore (at bios). On cpu-z, that translates to about 1.46 vcore on load. If I have it running on lower rpms, obviously the temps rocket beyond that to probably high 60s, low 70s. My antec 900 case helps tremendously - all fans are set to low except for the 200 mm fan (it's on medium).

I've resigned myself to running at around 3.2 to 3.3 ghz with about 1.3 to 1.4 vcore. Turns out the initial 3.4 ghz stable claim I made wasn't really stable (computer would restart about 2 hours after prime95).

Right now, I'm trying 3.2 ghz (8 x 400fsb) + 1.3625 vcore + load-line calibration disabled + most other voltages are lowered...Getting about 58 C on load with 1300 rpm on cpu fan.

My VID is just way too high...

Temps are measured using CoreTemp (and Everest)
Definition of load is "Prime95 running on all 4 cores".

I don't think my ram/motherboard is the limitation since I've tried higher fsbs and memory speeds + tighter timings and they worked out okay.

What benchmark appplications should I use?