Q6600 G0 TJmax, which one is right?

[Priv]

Ok I have been trying to find the right TJmax setting of the Q6600 G0 stepping but I just cant find a right one. Some people say 100C, intel says 90C, real temp says it's 95C. Now I'd generally believe intel if I wasnt getting 20C idle on one core in a 19C-20C room. TJmax of 95C and 100C seem more reasonable idle temperatures.

So my question is, is 90C the real TJmax? If not, which one is right?

 

[sportsfanboy]

100c for the go q6600, the b3 is 10c less at 90c.

I'm personally done using core temps as my temperature reading, just fed up with everything changing all the time. I use distance to tjmax in real temp. No matter what tjunction you set, distance to tjunction max will always be the same. I say keep the distance to tjunction max at or higher than 30c and enjoy.

 

[Priv]

Thanks for your answer. It's so annoying that intel is silent about stuff like this...

 

[CompuTronix]

The single best source of current Tjunction Max information available anywhere on the internet can be found on the Real Temp thread over at XtremeSystems - http://www.xtremesystems.org/forums/showthread.php?t=179044&page=114 - Since Intel's most recent Developers Forum (IDF) in October, the author of Real Temp, Kevin Glynn, (unclewebb), has changed Tjunction Max in Real Temp Beta releases to 100c for the Q6600 G0, according to Intel's updated Tjunction Max values for 65nm processors. Based upon my own research, as well as Kevin's, and the extremely well informed user, rge, we think that the actual Tjunction Max value for the majority of Q6600's fall between 95c and 100c.

Intel has stated that the DTS sensors are designed for Throttling and Shutdown protection, are more accurate at very high temperatures, become unreliable below 50c, and should be disregarded at Idle. Intel has also stated that Tjunction Max values vary between individual cores, and have a "range" or target value that can be as much as +/- 10c due to variables such as sensor "slope" error, which is especially pronounced on 45nm parts. This highlights the reason why Tjunction Max Values are rounded to the nearest 5c, such as 85c or 100c, while Tcase Max values are instead precise numbers to the tenths of a degree, such as the Q6600 B3 at 62.2c - http://processorfinder.intel.com/details.aspx?sSpec=SL9UM

Since everyone is so brainwashed on Core temperatures (Tjunction), they forget that there's a CPU temperature (Tcase)sensor designed for measuring temperatures from Idle thru Load, which is more accurate (when calibrated) than the often sloppy Tjunction sensors. Further, most users are not aware that the temperature shown in Intel's Processor Spec Finder is Tcase (CPU temperature), not Tjunction (Core temperature). Additionally, there's a known and constant relationship between Tcase and Tjunction at Load, which is 5c, and is shown on page 4, figure 5 of the following Intel document - http://arxiv.org/ftp/arxiv/papers/0709/0709.1861.pdf

This provides us with a simple 2 part procedure to accurately calibrate Tjunction Max values for each of your particular processor Cores. By using a standardized test setup and simple elementary school math, step 1 is to measure ambient temperature, then factor known Idle power dissipation values with CPU cooler efficiency values, which gives accurate Tcase (CPU temperature) for Idle calibration. By using the same standardized test setup, step 2 is to run Prime95 Small FFT's then add 5c to Tjunction (Core temperature) for accurate Load calibration.

It's just this simple:

Part 1: Ambient + Idle Power + Cooler Efficiency = Tcase Idle (CPU temperature)
Part 2: Tcase Load + 5c = Tjunction Load (Core Temperatures)

This is the calibration procedure in Section 9 of the Core 2 Quad and Duo Temperature Guide - http://www.tomshardware.com/forum/221745-29-core-quad-temperature-guide

When calibrated, your Tjunction Max will typically average about 97 to 98c, which is reasonable and in keeping with Intel's most recent statements concerning the confusion about Tjunction Max values. I developed the calibration procedure early this year before Real Temp was first released. My calibration procedure works, and is more accurate than just relying on uncalibrated sensors using Intel's default Tjunction Max values.

Comp

 

[sportsfanboy]

What he said^^^^

 

[soark]

100c for the go q6600, the b3 is 10c less at 90c.

I'm personally done using core temps as my temperature reading, just fed up with everything changing all the time. I use distance to tjmax in real temp. No matter what tjunction you set, distance to tjunction max will always be the same. I say keep the distance to tjunction max at or higher than 30c and enjoy.

I agree.. switch to distance to tjmax and you have the real thing..

 

[CompuTronix]

Not necessarily. Although distance to Tjunction Max is not an absolute temperature, it still involves Tjunction Max. You can still accurately calibrate Tjunction (Core Temperature).

Read my post above. If it's too much for you, then just use the "Easy Button" and rely on uncalibrated sensors and Intel's default Tjunction Max values.

 

[sportsfanboy]

Easy Button... What a great concept, I need one of those

 

[Zorg]

I was going to respond to the OP until I saw Comp posted.

Nothing to add here move along.

 

[CompuTronix]

Thanks, Tom.

 

[Zorg]

I'm personally done using core temps as my temperature reading, just fed up with everything changing all the time. I use distance to tjmax in real temp. No matter what tjunction you set, distance to tjunction max will always be the same. I say keep the distance to tjunction max at or higher than 30c and enjoy.

I agree.. switch to distance to tjmax and you have the real thing..

Not necessarily. Although distance to Tjunction Max is not an absolute temperature, it still involves Tjunction Max. You can still accurately calibrate Tjunction (Core Temperature).

That was a funny post, wasn't it.

Apparently they don't understand that core temp = Tjmax - distance to Tjmax. :lol:

 

[CompuTronix]

I'd be happy if I could just get everyone to understand that the temperature shown in Intel's Processor Spec Finder is CPU temperature not Core temperature, and that Core temperature is 5c higher.

I'd be even happier if Tom's would clearly label the Sticky's with a "Sticky" note icon with the word "Sticky" on it like it used to be a few years ago.

 

[Zorg]

They won't read it anyway. I told someone to read it and they said it was too much like work.

Everyone wants a shortcut, sad isn't it.

 

[CompuTronix]

"Easy Button"

 

[Zorg]

Yup :lol:

 

[Priv]

Very informative post. Explained a bit. Regarding Tcase I do not know what is wrong with the Tcase sensor of my Q6600, in everest ultimate it reads 22C with a room temperature of 20C. I doubt that is right. At load it is around 45C.

 

[CompuTronix]

There's nothing wrong with your Tcase sensor except that it needs to be calibrated. It's at least a few degrees low. Please read the Temp Guide Sticky and calibrate your temperatures. Use the link in my signature.

 

[sportsfanboy]

Zorg

I fully understand what tjuction is as well as tcase, distance to tjunction max and so on. The point I was trying to make was, using the core temperatures as a reading is pointless, as everyone has really been trying to figure that out since core2 came out. I personally have change my tjunction max setting 3 times (I have a b3 q6600). So I was saying, use distance to tjunction max and no more worrying about setting the proper tjunction max.

Of coarse CompuTronix way is the best....However, if your like me and getting close is enough, or you don't have a way to accurately measure ambient temperatures in the computer room, or simply don't care to go through the process, than distance to tjunction max is a good alternative.

 

[Zorg]

[Edited]

The big concern is exceeding the Tcase at high Tjunction numbers. The CPU won't throttle or shut down but damage or failure could still occur.

If the distance to Tjmax looks like a good number it may not be. Lets say that a distance to Tjmax of 15C looks good for example. That puts you at 82C given a Tjmax of 97C. Assuming a 7C load offset that will put the Tcase at 74c min. The Q6600 has a Tcase Max 71c.

So even though you are 15C away from a "damaging temp", which on it's face looks good, you are still exceeding the Tcase specification.

 

[sportsfanboy]

I know what your saying... I read on the Xtreme forums about using 30c or higher for distance to tjunction max, which is what I recommended to the OP. UncleWeb even agreed with it, so....

In just about every case if not all cases, a distance to tjunction max of 30+ Celsius is safe.

Also define damaging temp... None know for sure if running the chips that hot damages them. Again on the Xtreme real temp thread, UncleWeb ,RGE, others that work with him, have said why why would Intel set the throttling point so high if they thought it would harm the chip. UncleWeb also wrote " run your chip as hot as you want as long as your computer runs great". I'm not saying UncleWeb is the all knowing computer Genie, but he seems to know quite a bit about this stuff.

That being said, I still prefer to take a more cautious road, so I use 30+ distance to tjuntion max, as I'm no longer obsessed with my core temps, I just want to keep them within reason.

 

[Zorg]

That's like saying why would ATI allow the fans to run so slow that their cards run extremely hot if it damages them. If a chip fails prematurely due to increased electromigration it is not their concern, as long as it is not in warranty, which excludes all OCing. Come to think of it, that still doesn't explain ATIs insanely low fan speeds. They don't want their chips throttling prematurely because they would catch grief. Additionally the prochot# and thermtrip# are there to protect the CPU from catastrophic failure, not necessarily to be used as a ceiling to be operated just underneath. If you hit prochot# for a short time and fix the problem, you will probably be fine. If you OC, crank up the Vcore and sit just below the prochot# you more than likely won't. I don't think that Intel ever expects Tjmax to be reached under normal operating conditions, let alone for sustained periods of time.

In either case, I wasn't saying that you should be obsessed with the core temp especially if you are 30+ south of Tjmax. I was just saying that using distance to Tjmax as a rule of thumb for someone that doesn't have a handle on the relative values could lead to trouble. And that does not take DTS errors into account, which can only be addressed by the calibrations in Comps guide.

Reworking the numbers using the "run it unless it throttles" approach we get Q6600 Tjmax ~97 - distance to Tjmax of 2C - 7C max Tcase to Tjmax load offset = a Tcase of 90C. Running a Tcase of 90C for extended periods of time with a Tcase max of 71C specification from Intel is probably a very bad idea. And will probably also result in catastrophic failure.

Maybe someone with some spare chips should run a test and put the issue to bed. I would like to know how long one would last myself.

 

[sportsfanboy]

I hear you Zorg and your right... I never said chips should be run near their tjunction maxes. I also agree that high temps usually go together with high voltage, where electromigration can be a problem.

What I have been reading on the Realtemp thread has made me rethink the over all importance of temperatures. Uncleweb pretty much said that you are safe to run your chip as hot as you want as long as it's stable. That were his words not mine. So I used the logic that a modern Intel chip should be able to run up to what Intel says is max temperature, for the length of the processor warranty. I mean think about it... If Intel says max tcase is 62c, and you run it around that hot, and it dies, can't you use the warranty? I think you should be able to. Every one goes by Intel's max voltage numbers, so why is the temperature any different?

Unclewebb has brought just about every processor he has to within a couple degrees of tjunction max and they haven't died yet. But I don't know how long he test them and if there was any damage. I'm sure he will post if that happens.

 

[Zorg]

I want to keep my chip longer than the warranty period, plus I really have no warranty due to my OC. I would like to see a long term test at Tjmax though. It would be interesting to see how they fare.

 

[sportsfanboy]

That would be interesting, it would give us an idea of how hard we can actually push these chips.

It is weird that Intel would specify a particular tcase max, then release documents that give tjunction max. When it's now clear that if you do reach tjunction max (which according to them is a safety feature that triggers "throttling"), you would be well over the specified tcase max. Do they do this s--t on purpose, lol.

Maybe I'll post something similar in the Realtemp thread and see what feedback I get.

 

[Zorg]

Let me know what you find out.