Why TCase higher than Core temp?

D

davidm71

[H]ard|Gawd
Joined
Feb 11, 2004
Messages
1,568
Hi,

I have an Abit In9-32X with an E8600 on a TRUE. On Prime loads at 4.0ghz @1.25v (1.2125 stock) my Cpu temp is reported to be like 57 C and the cores are 50-52 degrees. Thing is I thought that was suppose to be the other way around. In the past I've done like three separate mounts to get the best temps but this is weird. Anyone have any ideas?

Thanks.

PS: On my i7 rig its the other way around though thats on water.
 
Try Real Temp, and see if its the same thing. I personally prefer Real Temp over Core Temp because of its easier use of its monitoring layout.
 
Pretty sure Core Temp and Real Temp always give the same values, maybe your chipset is really hot and your TRUE isn't circulating the air directly around the base of itself? You can always blame the sensors, they are rarely accurate.
 
Core Temp actually cripples my fps on some games, no idea how the hell that is related but its a one of its kind of problem. Core Temp always reports my temps 2-3 'C higher during LOAD then Realtemp and Speedfan and Etuner.
 
Kdot said:
Pretty sure Core Temp and Real Temp always give the same values, maybe your chipset is really hot and your TRUE isn't circulating the air directly around the base of itself? You can always blame the sensors, they are rarely accurate.
Click to expand...

I definitely get completely different temperatures between Coretemp and Realtemp. These are also different from the direct readings from the bios.

Coretemp gives me readings 10° C hotter than Realtemp and the Bios gives me readings about 20° cooler than Coretemp.

To try reconcile these, I have a trimmed thermistor under the socket cover touching the edge of the IHS and while this is not perfect and the edge of the CPU will be Core 4 and cooler than directly in the center (like [H]ardOCPs milling method), and gives me the feeling that Realtemp is the most accurate one. Also, if you have a lot of time on your hands, you can use the Realtemp calibration with Prime95.

Right now Realtemp says my CPU is 40° C (fan is on 1100 RPM only). Coretemp says it's 50° C which seems to hot to be realistic. Thermistor says 33.4° C. To me, there is no way the temperatures are 17° C hotter only 5mm over from where the thermistor is located. 7° C seems more likely.

-edit-

I rebooted that that instant so that temps would not drop t0o much (thermistor indicated no change) and went into the bios only seconds after I took my last realtemp reading an the Bios was reporting only 24° C which seems ridiculous.
 
Last edited:
I assume that by Tcase you are referring to the "CPU Temp" that you would see in the bios, or reported by a windows utility (like SpeedFan or HWMonitor) that pulls the value from the bios. Since you have a newer Core2Duo part, perhaps the calibration is not present in the BIOS for that chip.

I had this happen when i put a 45nm part into a p965 chipset a couple years ago. The motherboard was updated to "support" the new chip, but the temps were always way off. I had core temps consistantly 10 C lower than reported "CPU temp." When I wedged in a thermocouple as close as i could (without milling a channel in the CPU HS), it was a much lower reading, and below the reported core temp.

In general, assuming your Tj. Max is correct and the Bios is correctly reading the CPU temp, the core temp read from the CPU will always be higher than your CPU temp. It will also fluctuate more slowly. If you do a benchmark or burn-in, and your cores are pumped up quickly to 60C, you should see the CPU temp rise a little more slowly. The peaks will be close in this scenario, and could easily be equal. When the benchmark is stopped, the cores will drop quickly, but the CPU temp may stay high for a short time because of the amount of heat still in the base of the heatsink and the internal heat spreader. This may also explain the problem you are having, if it only happens right after a heavy CPU load is removed.
 
nayrhyno said:
I assume that by Tcase you are referring to the "CPU Temp" that you would see in the bios, or reported by a windows utility (like SpeedFan or HWMonitor) that pulls the value from the bios. Since you have a newer Core2Duo part, perhaps the calibration is not present in the BIOS for that chip.

I had this happen when i put a 45nm part into a p965 chipset a couple years ago. The motherboard was updated to "support" the new chip, but the temps were always way off. I had core temps consistantly 10 C lower than reported "CPU temp." When I wedged in a thermocouple as close as i could (without milling a channel in the CPU HS), it was a much lower reading, and below the reported core temp.

In general, assuming your Tj. Max is correct and the Bios is correctly reading the CPU temp, the core temp read from the CPU will always be higher than your CPU temp. It will also fluctuate more slowly. If you do a benchmark or burn-in, and your cores are pumped up quickly to 60C, you should see the CPU temp rise a little more slowly. The peaks will be close in this scenario, and could easily be equal. When the benchmark is stopped, the cores will drop quickly, but the CPU temp may stay high for a short time because of the amount of heat still in the base of the heatsink and the internal heat spreader. This may also explain the problem you are having, if it only happens right after a heavy CPU load is removed.
Click to expand...


I was in the opinion that it was a calibration bios problem as like you said they left minimal support for the Wolfdales never mind the E8600 and bios updates were limited year prior the release of that chip so I'm lucky it even works....

Thanks.

PS: Anyone else with an Abit IN9 32X out there can confirm this please reply.
 
Spare-Flair said:
I definitely get completely different temperatures between Coretemp and Realtemp. These are also different from the direct readings from the bios.

Coretemp gives me readings 10° C hotter than Realtemp and the Bios gives me readings about 20° cooler than Coretemp.
Click to expand...
See my below explanation of TCase. As for Real Temp versus Core Temp, the only differences in temperature are due to the fact that they may have different default Tjmax values for older CPUs. However, both programs obtain their readings directly from the DTSes embedded in the CPU, in the form of distance to Tjmax (what those sensors actually report). If you check the distance to Tjmax reported by each program, you'll find that they are the same.

You have a Q6600, and there is a bit of contention regarding what the actual Tjmax is for pre-G0 steppings, so that's why the Tjmax values are different between the two apps. I believe Real Temp uses a more correct Tjmax value.
davidm71 said:
I was in the opinion that it was a calibration bios problem as like you said they left minimal support for the Wolfdales never mind the E8600 and bios updates were limited year prior the release of that chip so I'm lucky it even works....

Thanks.

PS: Anyone else with an Abit IN9 32X out there can confirm this please reply.
Click to expand...
It has nothing to do with calibration. The fact is that the "CPU temp" sensor is actually mounted on the motherboard, and is useless for measuring the CPU temperature. The only way to measure TCase is to use the Intel-specifed method of milling a channel into the center of the heatspreader and mounting a calibrated temperature probe. You can completely ignore the CPU temp since it is entirely inaccurate.

Use either Core Temp or Real Temp to check your temperatures. With newer CPUs (45nm Core 2 and up), they will report the same temperatures, and they will be the most accurate values you can possibly obtain short of mounting your own thermal probe according to Intel's specs.
 
The CPU Temp sensor is adjusted depending on the CPU installed on certain motherboards. I tested this extensively on a GIGABYTE p965 board. With the board out in the open, on a workbench, with an E8400 installed, the CPU temp was 10 C higher than with a E4300 installed. The temp was measured with a thermocouple located as near as possible and the probe read just a degree or two low compared to the CPU Temp for the E4300 and about 10 C low for the E8400. Additionally, the idle "Core Temp" (in Windows) for each processor was nearly identical (and for the E8400 lower than the "CPU Temp"). For good measure, all these measurements were done with the chips clocked the same, both clocked to 2.83 Ghz.

Obviously, this one scenario with two chips in a specific board doesn't prove anything conclusively, but it does indicate that the BIOS is capable of adjusting the CPU Temp output depending on the type of CPU installed.
 
I don't understand why my bios is reporting such a ridiculous temperature for my CPU. EP45-UDP3 just doesn't recognize the B3 Q6600 properly?

Right now both realtemp and physical thermistor touching the CPU IHS is indicating a temperature around 40° C

My bios is reading 24° C for the CPU and 30° C for the case!
 
nayrhyno said:
The CPU Temp sensor is adjusted depending on the CPU installed on certain motherboards. I tested this extensively on a GIGABYTE p965 board. With the board out in the open, on a workbench, with an E8400 installed, the CPU temp was 10 C higher than with a E4300 installed. The temp was measured with a thermocouple located as near as possible and the probe read just a degree or two low compared to the CPU Temp for the E4300 and about 10 C low for the E8400. Additionally, the idle "Core Temp" (in Windows) for each processor was nearly identical (and for the E8400 lower than the "CPU Temp"). For good measure, all these measurements were done with the chips clocked the same, both clocked to 2.83 Ghz.

Obviously, this one scenario with two chips in a specific board doesn't prove anything conclusively, but it does indicate that the BIOS is capable of adjusting the CPU Temp output depending on the type of CPU installed.
Click to expand...
And how do you know that the measurements weren't affected by the location of the thermocouple you used as well? The fact is, putting one near the edge of the heatspreader is not a useful way of measuring the temperature since the actual source of heat is located in the center of the packaging (which is why Intel's spec requires you to drill a channel in the center and mount a thermocouple there).
Spare-Flair said:
I don't understand why my bios is reporting such a ridiculous temperature for my CPU.
Click to expand...
Did you read my last post? It's all in there.
 
You must log in or register to reply here.
Back
Top