RealTemp readings stuck

[BlackPearl]

I am trying to monitor my Celeron E3300 with RealTemp and temps always report 43C (Core 0)/40C (Core 1) whether the CPU is 0% or 100% load. Temp readings in BIOS (Biostar G41-M7) seem reasonable and change just as you would expect. Any idea what could be causing this? Thanks.

[burebista]

You did a Sensor Test?
Stuck sensors are common for 45nm CPU's.

[erocker]

If I'm not mistaken the temperature reading in your bios is taken from the CPU socket. Your on die CPU sensors are stuck.

[BlackPearl]

Thanks for your responses. The sensor test results in no change, so I guess they're stuck. Is there a way to unstick the sensors? If not, I guess I'll just live with the situation.

When running high CPU loads the smartfan functions normally, so maybe I shouldn't be concerned. It's a new CPU but it was so cheap that it's probably not worth the trouble of returning it.

[burebista]

Originally Posted by BlackPearl

Is there a way to unstick the sensors?

Yep, try to OC (upping Vcore mainly) and run LinX. If they don't move they are stuck badly.

[BlackPearl]

After OC (clock + Vcore) and LinX I now know that the sensors work above the 43/40, but that's still the lower limit. I'm using the HSF that came w/my Q6600 instead of the cute little mini-HSF that came with the E3300 so I guess it kept the non-OC CPU below that threshold even during the sensor test, SuperPi, etc.

Thanks very much for helping.

[brandonwh64]

have you also tested with coretemp or HWmonitor?

[unclewebb]

BlackPearl has a classic case of stuck sensors. They work fine above the sticking point but as the CPU cools down, they reach a point where they stop moving. As a CPU cools down, all Intel 45nm Core 2 sensors get stuck sooner or later. Intel admits that this can happen at any temperature below 50C which is why they don't recommend using these sensors for accurate temperature reporting.

Core Temp and HWMonitor read the exact same core temperature sensors. If they are stuck in RealTemp, they are stuck in those other two programs too.

[angelkiller]

@ Unclewebb

Do you know where the BIOS gets its temps?

[unclewebb]

There is a separate sensor embedded in the CPU. The bios can apply an offset to what this sensor reads so it can change when you update the bios. If you install the same CPU in two different boards from two different manufacturers, you can end up with a difference in reported temperatures in the bios.

The temperature displayed by this sensor is not the same as Tcase. The only correct way to measure the Tcase temperature is to cut a groove into the top of your CPU and mount a thermocouple at the geometric center of the IHS. The bios temp is more like a simulated, stab in the dark, approximation of Tcase.

The core sensors are designed to be more consistent. If they get stuck or read too high or too low in one board, they should do exactly the same in a different board. Too bad Intel didn't put a few more pennies into the quality of the core sensors that they use. The Core i7-900 series sensors are excellent, the Core 2 65nm sensors are pretty good when you use the correct TJMax and the Core 2 45nm sensors are horrible. They have some significant issues. They can get completely stuck at lower temperatures. They have slope error which means the further you move away from the calibration point, the more they can start reading either too low or too high and TJMax seems to be all over the place. TJMax seems to vary by as much as 10C or maybe more from core to core on the same Quad core CPU. Getting 100% accurate temperatures out of most of these sensors from idle to TJMax is impossible.