Page 1 of 1

Core Temp's tjmax for x3470

PostPosted: Tue Dec 13, 2016 9:02 pm
by Arjay
Hi there;

I just installed core temp and have been walking through it. I'm running an LGA1156 board with a x3470 Xeon.

I'm very curious to determine what you have set the pre-programmed tjmax value to for this processor? :mrgreen:

Re: Core Temp's tjmax for x3470

PostPosted: Wed Dec 14, 2016 2:56 am
by The Coolest
All (non-Atom) Intel CPUs starting with the Nehalem architecture have the TjMax value programmed into the CPU at the factory.
Core Temp reads this value, displays it and uses it to calculate core temperature.

Re: Core Temp's tjmax for x3470

PostPosted: Wed Dec 14, 2016 8:48 pm
by Arjay
Makes sense. I wonder if it's possible that Intel messed up on what it reports for max temperature for the x3470 processor?

According to Intel specs max case temp is 72.7c I this case (pardon the expression) the case is the metal box on the CPU that the heat sink attaches to. What I'm finding is that anything near 80c (as reported by Intel) results in the chip shutting down. In fact, sustained 75c temperatures under load result in CPU temperature limiting.

Coretemp reports a max temperature of something like 99c or 100c as the figure it's using so this would be what the proc is reporting to Coretemp. Yet, it seems to be in direct conflict with Intel's published spec of 72.7 on the top of the CPU case.

Re: Core Temp's tjmax for x3470

PostPosted: Thu Dec 15, 2016 1:04 pm
by The Coolest
I think it would be much more likely that there is a BIOS setting responsible for the throttling and thermal shut down.
Most of Intel CPUs are rated around 100C for TjMax.
Tj stands for Tjunction, which is the place where the core is attached to the PCB of the CPU.
TcaseMax is the maximum rating for the top of the heatspreader, the delta in temperature between the two can be quite large. It also takes the TCaseMax a lot longer to stabilize than core temperatures, which can change momentarily depending on load.

Re: Core Temp's tjmax for x3470

PostPosted: Fri Dec 23, 2016 9:11 pm
by Arjay
There is a BIOS setting for CPU thermal management. What it does is allow/disallow thermal throttling by dropping the multiplier when the CPU thinks it's getting too hot.

So I disabled that and reduced the overclock a great deal, dropped the CPU voltage etc and then ran another test.

Due to the reduced clock/voltage the temps were low, yet, 3/4 of the way through a 15 minute test the CPU once again sent out a signal for thermal throttling. It's very odd because the temps were low (70c max) which should have meant that it could run all day long for the test without needing to thermal throttle.

I see similar behavior at increased clock speeds and voltage. It seems during testing it will run without sending the thermal throttling command even with spikes to 80c on the core temps in the early stages of a test. Yet, when the 15 minute test is 3/4 complete it will then send the thermal throttling command.

So at reduced clock speeds and voltage, with the core temps no higher then 70c, 3/4 of the way through the test it will thermal throttle. It's as if the decision to thermal throttle is not just dependent on core temperature but also on duration/length of time with the CPU @ 100% load.