Please help me with this overclock (Q6600)

[iamgenius]

Okay. I decided to increase my overclock and get every MHz possible out of this Q6600. You can see my old overclock in my sig.


I'm now at 3.555 Fully stable according to intelburntest, and as you know, intelburntest is one hell of a stress program. I passed its high test 10 times, so that's stable to me at least.

See here:



Voltages in Bios are close to maximum safe voltages again according to me ! I'll just let you see it (better than saying it)





My G0 Q6600 is really greedy when it comes to Vcore. So, I can't really decrease its voltage. Ram is at 2.1 and MCH is +0.3 but I added a fan to it blowing inward.

Now to the big problem: Load temps are extremely high under intelburntest. I know this program stresses cpus more than any other and it can get them hotter by 15-20 degrees than any other program but still ! I reached 90 90 85 85. That's hot !

In prime95 small FFT I now reach 77 77 72 72 . Idle Temps are low though and you can see them in the first image I posted. I'm using realtemp to monitor temps. Coretemp reports the same temps, however, everest reports way cooler temps.

Are these temps normal with this Vcore? I'm using the infamous TRUE with AS5 as the thermal paste. I don't know if remouting will help but I'll do it anyway and try another thermal paste, which is the Shin-Etsu G571 because I heard good things about it. I'm waiting for it to arrive.

When I was at 3.320 you see in my sig, temps would never exceed 70 even with the AC off.

What do you think I should do? I feel like I go higher than 3.555 but I need to control overheating( I'm not hitting a FSB wall). is it my airflow? I think my airflow is good but I'll try to improve it as well. I'll post pics of my case if necessary.

 

[iamgenius]

Sorry, I noticed I didn't show you my idle and load temps

 

[Idontcare]

I reached 90 90 85 85. That's hot !

You could be very well throttling your CPU already. Do you know the specific TJmax value that your temperature monitoring program is using for your Go Q6600?

A lot of programs out there intentionally use the wrong TJmax value for Q6600's because they want the idle temps to appear more accurate at the expense of making the reported loaded temps less accurate. (core temp does this, on purpose per its author, for example)

Since you are intentionally playing with fire you should do your due diligence and make sure your reported temps mean what you think they mean.

 

[Shmee]

That's too bad you cant hit 3.6 with less than 1.5 vcore. I would say, stick with under 1.5, and just mind the max frequency at that vcore. My Q6600 with 1.25 VID can do 3.5 with 1.41 V, but I am temp limited with my xigamek dark knight, so I left it at that. This is on a evga 780i ftw btw.

What is the VID on yours?

 

[iamgenius]

You could be very well throttling your CPU already. Do you know the specific TJmax value that your temperature monitoring program is using for your Go Q6600?

A lot of programs out there intentionally use the wrong TJmax value for Q6600's because they want the idle temps to appear more accurate at the expense of making the reported loaded temps less accurate. (core temp does this, on purpose per its author, for example)

Since you are intentionally playing with fire you should do your due diligence and make sure your reported temps mean what you think they mean.

Tjmax is 100, isn't that the default with realtemp? Coretemp reports the exact same temps as realtemp........I don't think i'm throttling, I still have 10 degrees to go. I'll post a coretemp pic soon.

 

[iamgenius]

That's too bad you cant hit 3.6 with less than 1.5 vcore. I would say, stick with under 1.5, and just mind the max frequency at that vcore. My Q6600 with 1.25 VID can do 3.5 with 1.41 V, but I am temp limited with my xigamek dark knight, so I left it at that. This is on a evga 780i ftw btw.

What is the VID on yours?

VID is 1.2875 according to CoreTemp. My personal max Vcore is 1.55v. I'm a bit harsh on my goodies!

The PC runs perfectly stable at 3.555 and my idle temps are good. I know I will never ever get even close to what intelburntest does to my system. Add to that speed step which I will re-enable after I'm done overclocking.

I'm sick, I really want to reach that 3.6GHz ! I'll see what I can do about these temps.

Here is a coretemp screenie :



Uploaded with ImageShack.us

 

[Idontcare]

Tjmax is 100, isn't that the default with realtemp? Coretemp reports the exact same temps as realtemp........I don't think i'm throttling, I still have 10 degrees to go. I'll post a coretemp pic soon.

TJMax for G0 Q6600 is 90C, not 100C.

(see this post and the links to Intel's publications therein)

Read the posts near the end of this thread:What is the TJ MAX for the Q6600 G0?

Coretemp, and realtemp apparently, do not report the TJmax for the chip but instead are reporting to you the TJmax value that the thermal program itself is using when computing the temperature of the chip.

Because of the inaccuracies of the thermal probe (it is only calibrated to be correct when actual temps are right at TJmax setting) you can't have accurate temperature readings at both idle and load. The setup on the cpu is designed to be accurate at load but entirely inaccurate at idle.

Coretemp opted to make the reported temps at idle more realistic at the expense of being flat-out wrong at load as the temps approach the TJmax value.

I verified this directly with the author of coretemp.

If you want coretemp to correctly display your cpu's temperature as reported by the thermal diode on the cpu when the cpu is loaded then you must dicker with the TJmax offset value for coretemp so that it essentially starts using the correct TJmax value.

(I don't know how to do it for realtemp but I'm sure the documentation is out there on the web)

 

[iamgenius]

TJMax for G0 Q6600 is 90C, not 100C.

(see this post and the links to Intel's publications therein)

Read the posts near the end of this thread:What is the TJ MAX for the Q6600 G0?

Coretemp, and realtemp apparently, do not report the TJmax for the chip but instead are reporting to you the TJmax value that the thermal program itself is using when computing the temperature of the chip.

Because of the inaccuracies of the thermal probe (it is only calibrated to be correct when actual temps are right at TJmax setting) you can't have accurate temperature readings at both idle and load. The setup on the cpu is designed to be accurate at load but entirely inaccurate at idle.

Coretemp opted to make the reported temps at idle more realistic at the expense of being flat-out wrong at load as the temps approach the TJmax value.

I verified this directly with the author of coretemp.

If you want coretemp to correctly display your cpu's temperature as reported by the thermal diode on the cpu when the cpu is loaded then you must dicker with the TJmax offset value for coretemp so that it essentially starts using the correct TJmax value.

(I don't know how to do it for realtemp but I'm sure the documentation is out there on the web)

Interesting and valuable information, but to be frank I don't fully comprehend it as it is 1 am here and I'm sleepy as hell. I'll go sleep, do my homework by reading your links and then come back.

But, you can take the time to tell me exactly how to configure coretemp to read temps correctly.....I just go and change TJmax to 90 ??

Thanks a million.

 

[iamgenius]

I couldn't resist and I read some of the links. So my temps should actually be lower at load. I did adjust coretemp and realtemp. So my load temps were actually lower by 10 degrees !

All I did was to set TJmax to 90 in realtemp and set offset value of -10 in coretemp.

Now, that's good news to me actually! Everest was the program reporting the most accurate temps then.......coretemp and realtemp now match everest exactly at load !

Time to go to bed.

Thanks.

 

[Idontcare]

But, you can take the time to tell me exactly how to configure coretemp to read temps correctly.....I just go and change TJmax to 90 ??

Thanks a million.

See this post: http://forums.anandtech.com/showpost.php?p=29792708&postcount=29

Except I made a mistake at the time in step 3, you want to make that a -10 (negative 10) not a +10 for the correct offset.

 

[james1701]

Have you tried dropping the multi down to 8 and raising you fsb higher. I found I get my Q6600 higher on lower voltage this way. I don't know if it was the P45 chip set on my MSI board but it did let me get 3.6ghz on 1.45 volts.

 

[iamgenius]

Have you tried dropping the multi down to 8 and raising you fsb higher. I found I get my Q6600 higher on lower voltage this way. I don't know if it was the P45 chip set on my MSI board but it did let me get 3.6ghz on 1.45 volts.

Really? But that means my Ram will be at (3555/8)*2 = 888.75. Do you think my ram can take it? I don't have a multiplier less than 2.

 

[iamgenius]

Idontcare.........why does the TJmax has anything to do with the hottness of my cpu?? Why don't programs just give me the reading the diode is giving? I mean I'm not calculating the delta to TJmax, but the actual temp of my cpu. Even if TJmax is 1000, what does that do with my temps?

I don't get this !

 

[james1701]

You might need to loosen your timings just a little, but your memory probably will run at that speed. 8*425 give it a try.

 

[Idontcare]

Idontcare.........why does the TJmax has anything to do with the hottness of my cpu?? Why don't programs just give me the reading the diode is giving? I mean I'm not calculating the delta to TJmax, but the actual temp of my cpu. Even if TJmax is 1000, what does that do with my temps?

I don't get this !

The way the thermal diode is setup to work is that it is designed to simply provide a reference temperature relative to the TJmax setpoint.

If TJmax is set to 90C (it is set at the factory before the chip is shipped) and your CPU is loaded and running at 80C then the thermal diode is actually reporting a value of "-10". Meaning the chip is 10C below the factory set TJmax.

Thermal diode doesn't report that the CPU is at 80C. What we consumers do is we get a program like coretemp which will read the thermal diode's reporting (-10 in this example) and converts that reported value into a temperature value by adding TJmax to it.

Since TJmax is both CPU and stepping specific (Q6600 G0 vs. B3, etc) this is why the temperature reporting programs actually need to have a TJmax lookup table built-in so it can cross-reference the reported thermal diode reading and come up with a temperature value that seems believable.

For a long time, years actually, Intel refused to publicly release the TJmax values so everyone just guessed what they were (including the guys who made the programs like coretemp). Then when 45nm came along Intel decided they would publish all the TJmax values including the ones they set for their 65nm processors.

That is how we came to find out that the Q6600 G0 TJmax is actually 90C. (kinda...the initial publication actually contained an error which was later corrected but unfortunately a lot of people committed to memory the first number - the wrong number - and a lot of time on these forums was expended on eradicating the misinformation)

So that is why TJmax matters. But these free temperature programs are kinda like a popularity contest and the authors want their products to be popular and even though they know it is wrong they use a TJmax value that makes the reported temps appear accurate (which we already discussed above) because nobody wants to use a temp program that seems wrong to the novice's uneducated eye.

 

[VirtualLarry]

Now to the big problem: Load temps are extremely high under intelburntest. I know this program stresses cpus more than any other and it can get them hotter by 15-20 degrees than any other program but still ! I reached 90 90 85 85. That's hot !

In prime95 small FFT I now reach 77 77 72 72 . Idle Temps are low though and you can see them in the first image I posted. I'm using realtemp to monitor temps. Coretemp reports the same temps, however, everest reports way cooler temps.

Are these temps normal with this Vcore? I'm using the infamous TRUE with AS5 as the thermal paste. I don't know if remouting will help but I'll do it anyway and try another thermal paste, which is the Shin-Etsu G571 because I heard good things about it. I'm waiting for it to arrive.

I get temps inline with that with my G0 @ 1.325v in BIOS, LLC enabled. CPU @ 3.6Ghz even. 9 * 400. I wouldn't worry about it too much. G0s, when overclocked to stupendous levels, get pretty warm.

 

[iamgenius]

The way the thermal diode is setup to work is that it is designed to simply provide a reference temperature relative to the TJmax setpoint.

If TJmax is set to 90C (it is set at the factory before the chip is shipped) and your CPU is loaded and running at 80C then the thermal diode is actually reporting a value of "-10". Meaning the chip is 10C below the factory set TJmax.

Thermal diode doesn't report that the CPU is at 80C. What we consumers do is we get a program like coretemp which will read the thermal diode's reporting (-10 in this example) and converts that reported value into a temperature value by adding TJmax to it.

Since TJmax is both CPU and stepping specific (Q6600 G0 vs. B3, etc) this is why the temperature reporting programs actually need to have a TJmax lookup table built-in so it can cross-reference the reported thermal diode reading and come up with a temperature value that seems believable.

For a long time, years actually, Intel refused to publicly release the TJmax values so everyone just guessed what they were (including the guys who made the programs like coretemp). Then when 45nm came along Intel decided they would publish all the TJmax values including the ones they set for their 65nm processors.

That is how we came to find out that the Q6600 G0 TJmax is actually 90C. (kinda...the initial publication actually contained an error which was later corrected but unfortunately a lot of people committed to memory the first number - the wrong number - and a lot of time on these forums was expended on eradicating the misinformation)

So that is why TJmax matters. But these free temperature programs are kinda like a popularity contest and the authors want their products to be popular and even though they know it is wrong they use a TJmax value that makes the reported temps appear accurate (which we already discussed above) because nobody wants to use a temp program that seems wrong to the novice's uneducated eye.

Can't be said better. Thanks. So, delta is the thing being actually measured. BUT, isn't that an awkward way of doing it? or is it the only way to be accurate?