Maybe wrong TJ_max for T6400 mobile cpu

[CoreTemp-User295]

Using your latest 0.99.5 beta build 20 results in TJ_max = 90°C on a Dell Studio 1555 with T6400 CPU:



However intel specifies TJ_max = 105°C for this CPU:
http://processorfinder.intel.com/detail ... Spec=SLGJ4

Is there another source for the correct TJ_max? Unfortunately the processor datasheet for the 45nm cpus:
http://download.intel.com/design/mobile ... 012001.pdf
in section 5 page 101 only contains the TJ_max of some 45nm CPUs, but not this particular one above.

[The Coolest]

Mobile CPUs may come in two varieties, 105C and 90C. The TjMax flag is used to tell whether a process is of the 90C variety or not.

[CoreTemp-User295]

I extracted a CoreTemp-Dump.txt from the T6400 CPU.

It's MSR 0xEE flag evalutes to 1:
eax : 0xC6B90400
=> eax & 0x40000000 == true

Which makes CoreTemp assume TjMax == 90°C.

The 0xEE flag was discovered by Rudolf Marek (click), which you collaborate with on a regular basis and as far as I know is the ONLY person who ever reverse engineered this code. The everest authors also consumed the MSR 0xEE (4th posting: click).

I e-mailed with Rudolf this weekend that his coretmp.c driver doesn't support Penryn CPUs the correct way. Meanwhile I think he got in touch with you and released an updated driver on Monday:
http://lists.lm-sensors.org/pipermail/l ... 26265.html

I see that you and Rudolf as well extrapolate MSR 0xEE Tj_max reporting to Penryn architectures. Is there any evidence that this is true? I haven't seen any intel documentation regarding this (and do know the CPU spec, the Architecture spec, the IDF SF08, Taipei 08 documentation). Did somebody actually reverse engineer code or have documentation regarding this matter?

The way I see it: If the above intel documentation (processorfinder) is right (T6400 Tj_max = 105°C) , then CoreTemp is worng (TJ_max = 90°C).

Unfortuantely I don't own the T6400. But since the difference between 90 and 105 is 15°C I would expect to easily verify that at idle temps (I know DTS is not calibrated for idle temps, but should be sufficient). I will ask somebody to report his idle temps. If they are around 35-40°C, then CoreTemps assumption of 90°C is likely to be correct.

Any suggestions?

[CoreTemp-User295]

Intel's ark database also mentions TJ_max = 105°C for the afore mentioned CPU:


http://ark.intel.com/Product.aspx?id=40479&code=T6400

However with TJ_max = 90°C this cpu is idling at 40-45°C, which sounds quite reasonable.

[The Coolest]

I'm pretty sure the 0xEE register is still valid for these processors. But I can't remember the source I got this information from originally (been over 3 years now).

[CoreTemp-User295]

I mentioned already the source where you got the information regarding 0xEE flag from. It's a hyperlink and I guess you'll remember if you click it.

At that time you first heard about the 0xEE flag, there simply was no Penryn available. Penryn Cores have a different TJ_max then Merom and any Core before. The T6400 is Penryn based. The first Penryn Cores were launched 6th January 2008. What you say is: You heard in 2006 about a flag that will be valid on a chip released two years later.

At the time where no information about TJ_max for desktops were available, people used the mobile TJ_max - and they were wrong.