3Ghz E6600 Overclock Problem

clarkey01

Diamond Member
Feb 4, 2004
3,419
1
0
I was running Orthos for 33 minutes, blend test, when I got errors, my temps are 51c @ full load, 3ghz on 1.3Vcore

Should I lower my overclock to 2.8Ghz???

I hadnt notice any problems, I simply ran Orthos to check up.

Any advice guys?

Cheers
 

DSF

Diamond Member
Oct 6, 2007
4,902
0
71
You could always run Memtest86+ overnight to rule out bad memory.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
MEMTEST86 or "+" would also find deficiencies in CPU configuration.

If the memory is good, your next step would be to bump up the VCORE.

1.3V VCORE on an E6600 is fairly limp. I can't remember if the retail-box "maximum" was 1.35V or 1.40V. I remember that the default value with my mobo and memory was 1.40 or higher.

Some people were pushing the VCORE above 1.46 or 1.47, but I think that may be excessive, or let me say -- I personally think it's too excessive for ME.

I think I had mine at 1.44V to get 3.33 Ghz or thereabouts.

Look again at the Anandtech article on overclocking the QX9650. The basic procedures still apply:

For 1,066 FSB -ready processors, try to start at 1,333 or 1,334. For 1,333 FSB processors, the article recommends starting at 1,600. By this, they mean knocking the processor multiplier down as far as it will go (6), raising the RAM voltage to just below recommended warranty maximum, and then pushing the memory as far as it will go by loosening the timings and raising the FSB.

ONce you've done that, go back to multiplier 9 and start OC'ing the processor. But if the memory in your case is rated at DDR2-800, you shouldn't have to loosen the stock timings. By the time you've found a comfortable trade-off between CPU-speed, temperatures and voltage, you should probably be able to loosen the timings. And you might be able to drop the VDIMM memory voltage.

You should also be able to just lower the speed to 2.8 or 2.88 Ghz, but 3 Ghz is only a 25% over-clock and therefore pretty easy for an E6600.

Doesn't the P5B motherboard use the 965 Intel chipset?
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
I said (hastily):

"By the time you've found a comfortable trade-off between CPU-speed, temperatures and voltage, you should probably be able to loosen the timings. "

What I meant was:

". . . . . you should probably be able to tighten the timings."
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
Are these the CoreTemp "TJunction" temperatures - one for each core? Or are they the TCASE temperature? If you are reporting an average for core temperatures, that's pretty good. If it's the TCAse temperature, it's about normal. My experience with the E6600, using a ThermalRight Ultra-120-Extreme cooler, was that you would get a TCase of around 52C pushing the processor to 3.3+ Ghz.

Like I said, without "endangering" your processor, you have lots of wiggle-room for VCORE adjustment. I'm guessing -based on experience -- that if you push the VCORE to 1.40V, the load-generated droop in the voltage will bring it back down to 1.37V.

Of course, the trick is to find the lowest voltage for any given over-clock setting. You choose your over-clock limit by attempting to stay within reason given some gut feeling about what might cause long-term damage to the processor.

Look at it this way. Intel sets the retail-box maximum voltage based on their lab tests. Intel doesn't want any processors returned under warranty RMA. Therefore, the probability of damage to the processor at the retail-box maximum voltage is practically 0%. If you exceed that voltage by 0.02, 0.03 or even 0.04V, you're exploring the low tail of a probability distribution for "failure after so many [hours, days, weeks, months, years]." What's 0.04 as a fraction of 1.375V? Less than 3%. Obviously, we don't know what the expected frequency of failure would be, as we follow the linear X-axis from 1.375 to 1.414V. But I doubt that it is significant. And we're talking about failure over some extended period of time. In a way, testing CPUs for reliability is probably not much different than testing light-bulbs at GE.

As I said, 3 Ghz should be easy for an E6600 . . . . 3.2 is still pretty easy. . . ..
 

clarkey01

Diamond Member
Feb 4, 2004
3,419
1
0
Originally posted by: BonzaiDuck
Are these the CoreTemp "TJunction" temperatures - one for each core? Or are they the TCASE temperature? If you are reporting an average for core temperatures, that's pretty good. If it's the TCAse temperature, it's about normal. My experience with the E6600, using a ThermalRight Ultra-120-Extreme cooler, was that you would get a TCase of around 52C pushing the processor to 3.3+ Ghz.

Like I said, without "endangering" your processor, you have lots of wiggle-room for VCORE adjustment. I'm guessing -based on experience -- that if you push the VCORE to 1.40V, the load-generated droop in the voltage will bring it back down to 1.37V.

Of course, the trick is to find the lowest voltage for any given over-clock setting. You choose your over-clock limit by attempting to stay within reason given some gut feeling about what might cause long-term damage to the processor.

Look at it this way. Intel sets the retail-box maximum voltage based on their lab tests. Intel doesn't want any processors returned under warranty RMA. Therefore, the probability of damage to the processor at the retail-box maximum voltage is practically 0%. If you exceed that voltage by 0.02, 0.03 or even 0.04V, you're exploring the low tail of a probability distribution for "failure after so many [hours, days, weeks, months, years]." What's 0.04 as a fraction of 1.375V? Less than 3%. Obviously, we don't know what the expected frequency of failure would be, as we follow the linear X-axis from 1.375 to 1.414V. But I doubt that it is significant. And we're talking about failure over some extended period of time. In a way, testing CPUs for reliability is probably not much different than testing light-bulbs at GE.

As I said, 3 Ghz should be easy for an E6600 . . . . 3.2 is still pretty easy. . . ..

Im going to be a loser and drop down to 2.8, what Vcore would you guess I could get away with???
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
Ya know what? I just checked my preliminary "Notebook" notebook June-July 07 entries that I began keeping with the E6600 with my 680i Striker Extreme mobo.

I don't have any entries where I ran the processor at a slower speed than 3.285 Ghz, nor lower voltage than 1.438V.

So all I can say here, is that 2.8 is so "easy" -- something like a 16% over-clock -- my best guess would put the voltage requirement somewhere between 1.30 and 1.35V.

The retail-box maximum is either 1.35 or 1.40V -- go to Intel and check which.

A 1.37V fixed setting should be no stress or strain. You'll just want to find the lowest -- or second-lowest once stable -- to run it.

I'm more in the habit now of finding a lowest setting that won't fail for 8 hours under PRIME95, and then I'll kick it up one notch or two. But the VCOREs I've reported for 3.35Ghz, 3.33 Ghz and 3.29 Ghz used the lowest stable VCore -- respectively 1.47V, 1.46V, and 1.44V.
 

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
Originally posted by: clarkey01
Im going to be a loser and drop down to 2.8, what Vcore would you guess I could get away with???

At 2.8, or at 3.0? At 3.0, you shouldn't require any more than 1.325v. At 2.8, 1.30 is probably about right, at least with your E6600. And Bonzai, I think you give really good advice, so don't offense, but I hereby vote you AT's "most likely to write a book, when 3 sentences would have been enough".;)
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
The newspapers publish my "letters to the editor" political commentary. Some of my best letters -- not just my opinion -- are longer than average, and some longer than that. They only publish my quickie 250-word remarks. So, I understand, don't take offense, and blame it all on coffee and my fingers.