Tjunction X6800

vpsaline

Distinguished
Dec 23, 2006
236
0
18,680
The temperature guide says that 60/65-65 is when the CPU is HOT. But Coretemp says that Tjunction Max is 85c.

My CPU idles in a 23C room at 45/46-46 with a Zalman 9700. I am confused as to where this stands, in other words, is there still room to overclock, have I gone too far, or are temperatures just right?
 
Solution
Don't feel like you're alone; many others are also confused by Intel's thermal specifications. To answer your questions, technically speaking, no and yes. Intel's papers go to great lengths explaining how they test CPU temperature in the lab to measure Tcase Max, which if you might have noticed, is defined to the tenths of a degree, while Tjunction Max is just a round figure such as 85c or 100c.

To put it into simplified "real world" terms, the practical difference is that Tcase Max applies to operating temperature, while Tjunction Max applies to shutdown temperature, which is strictly for overtemp protection, so you don't toast your transistors.

As soon as you exceed Tcase Max, you've reached an overtemp condition, in which you can...
vpsaline

Tcase and Tjunction are temperatures.
Tcase Max and Tjunction Max are specifications.

As I stated in the Guide, "Tcase Max is always the limiting thermal specification". Tcase Max is the specification which is shown in Intel's Processor Spec Finder - http://processorfinder.intel.com/details.aspx?sSpec=SL9S5 - NOT Tjunction Max. This is the most common misconception among users.

For your X6800, Tcase Max is 60.4c, which I've rounded off to 60c for simplicity. Since Tjunction (Core temperature) is 5c higher than Tcase (CPU temperature), this would corresponds to 65c. The other Specification, Tjunction Max, applies to overtemp protection such as Throttle and Shutdown. As such, any X6800 Core temperature (Tjunction) which exceed 65c should be considered "overtemp".

Further, as you haven't listed any load temperatures for Prim95 Small FFT's, it's not possible to see whether your rig is running within thermal specs.

Comp :sol:
 
*Tjunction Max or Tj Max is the temp your CPU will shutdown at. 85c Tj Max CPUs will shutdown if core temps reach 85c and 100c for 100c Tj Max CPUs. When core temps are 5c below Tj Max thermal throttling will occur which will drop vcore and CPU speed in attempt to cool down the CPU. Tj Max 85c will throttle at 80c and Tj Max 100c will throttle at 95c.
 
That quote is talking about Tjunction Max, not Tjunction. Tjunction is the core temperature, Tjunction Max is (approximately) the temperature where throttling occurs. It is not the shutdown temperature.

If you want to be more technical, throttling doesn't occur at a certain temperature, it occurs when the Digital Thermal Sensor gives a reading of at or just above 0. What Tjunction Max is used for the monitoring program and what temperature it displays is irrelevant to thermal protection mechanisms.
 
vpsaline,

There's a 5c gradient between Tcase and Tjunction, which means that "mean' or "average" Core temperatures are simply 5c higher than CPU temperature.
This is detailed on page 4 and illustrated in figure 5 of the following Intel document - http://arxiv.org/ftp/arxiv/papers/0709/0709.1861.pdf - and still applies to all variants of Core 2's, Core i7's and Core i5's.

There's also an "approximate" 5c threshold on "some" variants between Throttle activation and Shutdown activation. So as you can see, it's easy to get confused when you're trying to make sense of Intel's mess, while others are tossing numbers around without being specific or offering definitions and explanations.

To summarize, running your processor at Tjunction Max, regardless of variant, is just too damned hot for testing, stability and longevity, so once again, it falls back to Tcase Max as being Intel's limiting thermal specification.

For your X6800:

Vcore Max 1.5v
Tcase Max 60c
Tjunction 65c

Once again, just to help keep things in perspective, remember that Tcase and Tjunction are temperatures, while Tcase Max and Tjunction Max are specifications.

Comp :sol:
 
Ok then I guess Im confused about why Intel says on one hand Tcase max 60.4 and Tjunction Max 85C, isnt that a contradiction on their part? Or is what you are saying is that intel has two types of specifications bassicaly like them saying real world specification Tjunction max 65, physical limit of tjunction max 85C (but you should ignore this value)?

And are you saying that basically throttle will prevent Tcase from ever reaching 60.4 or surpassing it or it begins to throttle at 75.4C?
 
Don't feel like you're alone; many others are also confused by Intel's thermal specifications. To answer your questions, technically speaking, no and yes. Intel's papers go to great lengths explaining how they test CPU temperature in the lab to measure Tcase Max, which if you might have noticed, is defined to the tenths of a degree, while Tjunction Max is just a round figure such as 85c or 100c.

To put it into simplified "real world" terms, the practical difference is that Tcase Max applies to operating temperature, while Tjunction Max applies to shutdown temperature, which is strictly for overtemp protection, so you don't toast your transistors.

As soon as you exceed Tcase Max, you've reached an overtemp condition, in which you can still push the temperatures even hotter, until the processor throttles, and then finally shuts down at Tjunction Max.

Think of it like the temperature guage in a car, except the guage has three needles; one CPU needle which is alwyas 5c lower than the two Core needles. We all know the red zone, which begins at 60c-65c-65c means hot, but if you keep driving until the needles are pegged at 80c-85c-85c, then it's going to shut down.
 
Solution
You're very welcome. I saw an opportunity to bitch about the new ranking system, and just couldn't pass it up.

If I may make a recommendation; the very best source of ongoing information and discussions on this topic that's available anywhere on the Internet is on the Real Temp thread at XtremeSystems - http://www.xtremesystems.org/forums/showthread.php?t=179044&page=149 - which has reached 149 pages of 3,707 posts, as of this writing.

I've followed the thread since Real Temp was first developed, and occasionally correspond and share findings with the author, Kevin Glynn, (username unclewebb). He's very interactive, is quite personable, is always helpful, and provides excellent explanations.

Comp :sol:
 
I drew down your analogy, to help explain it to friends who also have X6800 processors and who are just as confused as I am. Did I get it right? Maybe you can include something similar in your temperature guide, great analogy by the way.

inteltachometer.jpg
 

Actually, yes. Although I've explained the finer shades of gray to you in different terms, the information is in the Guide, in several Sections. In previous versions, I had included more information specifically concerning Tjunction Max, and had provided Tjunction Max values in the Scales, however, since Intel wasn't forthcoming with the values for all varaints, I found that it caused more confusion than it alleviated, so I removed Tjunction Max values from the Guide.

Although I've tried my best to strike a balance between detail and simplicity, the topic is far from simple, so the Guide is not an easy read. I'm sometimes body slammed by the occasional drive-by critic for the Guide being too complex, which is why it syas right up front: "This Guide is intended for intermediate to advanced users. Although certain strict definitions have been relaxed to simplify concepts, this Guide contains detailed technical information."

Further, since the Calibrations Section in the Guide uses Ambient and Tcase idle to find Tjunction load, it doesn't need Tjunction Max values to achieve accurate, custom SpeedFan temperature calibrations, provided that the Test Setup is observed to the letter.


Nice illustration. I would make the following changes so it looks like:

Tcase: CPU
Tjunction:
Core 1
Core 2

Tcase Max 60c
Tjunction 65c

Processor Throttle:
Tcase Overtemp 75c
Tjunction 80c

Shutdown Protection:
Tcase Overtemp 80c
Tjunction Max 85c
 
wow, I used CoreTemp, RealTemp & Prime95 and got this...
x6800.jpg

They really have this down to a science. I calibrated Realtemp, from Tjunction max 90C (default) down to 85C, I presume that is correct, and got those results. Do I have headroom to OC or am I pretty much at the limit? My room was 22C at the time, and my Airconditioner is broke, usually I do intensive cpu tests with AC on and room temp of 14-20C or less. ( I know 14 is cold :ouch: )
 

The author of Core Temp, Authur Liberman, uses Tjunction Max 85c for the X6800, while the author of Real Temp, Kevin Glynn, uses 90c, which is the latest information that Intel has released.

However, understand that Tjunction Max values differ from individual Core to individual Core, and from Stepping to Stepping, and from variant to variant, and from Model to Model, as well as deviations between DTS sensor arrays among 65 nanometer and 45 nanometer technologies such as slpoe error and linearity, which means that Tjunction Max values are just approximate, canned, default values that are not etched in stone! This means that of the 27 processor variants which I've personally tested extensively, those which were (for example) Tjunction Max 100c, the actual mean value was 97 to 98c, not 100c!

Thus, your X6800 is most likely Tjunction Max 87 to 88c, not 85c or 90c. Among all the high and low order variables, the only constants that we know somewhat for certain, or which can be measured are ambinet, idle power dissipation, CPU cooler efficiency, and the known 5c gradient between Tcase and Tjunction. If you want to find the most accurate customized temperatures for your particuar X6800 sample, then do the SpeedFan calibrations in my Guide.

Also, 22c is the industry Standard for ambient temperature, and is the temperature at which I perform all thermal testing, so as to have valid, reproducable and consistent results.