TSK's Crazy Theory on TJmax and the DTS - Overclock.net

TheSubtleKnife · 03-10-08

Okay, this is JUST a theory, on what I have observed, and it is more wrong than right, but it kinda makes sense in my head. Here goes.

The tests were done on an E2180, M0 stepping.

Before I start, I need to point this out. Intel's website states the "Tcase Max" as 73.2, and Core Temp states the "Tjunction Max" as 85. The difference between those two is 11.8C, or basically 12C apart. Remember this number.

For the temperatures, I rigged a thermometer with a probe so the probe is in contact with fins of the stock intel heatsink. No fan on the heatsink, so we can better assume that the core temperatures are close to the temperatures of the heatsink.

Just fired up orthos, and you can see that the difference between cores is close to 12C, which is not too surprising, because the core heats up faster than the heatsink.

Now a couple mins later, temps are higher, but the difference is still 12C..hmmm

about 10 mins later...difference is STILL 12C???

not much later*sigh* still 12C

Left to go play brawl for 2 hours and guess what...even after the temperature stabilized, the HS is still 12C cooler than the "core"

Notice that if you assume that the Tjunction is around 73, instead of 85, you'd get readings much more similar to the readings on the thermometer.

Next I decided to try something different. Intel has included a bit in their CPU's that are activated when the CPU reaches a certain threshold. This could either be the Tcasemax of around 73C, or it could be the debated Tjunctionmax of 85, or 95C. To put this to the test, I upped the core voltage, so the CPU temps will hopefully cross the threshold.

What I found out was interesting. The PROCHOT bit was enabled BEFORE the temps hit the "Tjunctionmax" temperature. Also, the temperature of the heatsink was close to the Tcasemax of 73 ( i know its off by about 5C, but I think if I gave it more time, it would have balanced out more, and read closer to 73C, but I didnt want to fry my chip for too long)

The "highest temp" that Realtemp recorded was 93C, below the tjunction of 95C. You can also see that the PROCHOT bit has been tripped, meaning it had passed the threshold.

From what I have observed, I think that the "TJunctionmax" needed for the DTS calibration is equivalent to the "Tcasemax" that intel has already given us. That could explain why they havent released the TJmax info, because we already have it. Remeber that this just some crazy idea that I just thought of, so yeah, dont quote me on any of this.

UPDATE:
I found this info from intel

Quote:

In the event of a catastrophic cooling failure, the processor will
automatically shut down when the silicon has reached a temperature
approximately 20 °C above the maximum TC.

which means that is should shut down at 93.2C (73.2+20) actual temps. Take the average of the last readings of 111C and 123C that i got before it shut down, and you get 118C. subtract 118 by 21.8(Realtemp Tjunc - Tcasemax)= 96.2C, which is pretty darn close to the temp that its supposed to shut down at.

With the above as evidence, I can say that the TJmax for my E2180 is NOT 85C (Coretemp) or 95C (realtemp), but closer to 75C. Note that the DTS is calibrated for higher temperatures, so this TJmax is not accurate at idle temps

*CyberDruid* · 03-10-08

Interesting. Thanks for the test

GigaByte · 03-10-08

Interesting, very, VERY. As I said in a different thread I am starting to lean towards 95c Tj max for M0 stepping E2xxx/E4xxx as the delta has changed to 5c for these steppings according to the temp guide by Computronix. Having your CPU throttle below Tj max is correct. Tj max which stands for Tjunction max is the maximum allowable temp for the cores, another word for core temps is just "Tjunction". Now adding the "max" to the word mean the maximum, or highest, as in cannot go any higher. This is how Tj max gets it's description as being the shutdown temp. If its the Tjunction MAX why would it shut down beyond it? Max is max. It is said the PROCHOT# signal will activate about 5c before Tj max is reached. Once Tj max is reached THERMTRIP# will cut power to the cores, hense shutdown.

You said the max temp RealTemp read was 93c, if its ok with you can you do the same test again, but let it go up to 95c (the odd degree or two above 95c reccomended) and say if it shut down or not?