ABIT KT7 CPU Temp Accuracy

SUMMARY: ABIT’s KT7 adjusts its socket thermistor readings to approximate CPU temps.

I thought it would be interesting to see just what in-socket thermistors are reading, so I snaked a thermocouple under the socket of a KT7 to read CPU back temps in the same location as the in-socket thermistor. I did this placing the thermocouple next to the thermistor’s little blue bulb and proceeded to measure motherboard and thermocouple temps.

THE TEST SETUP

I used an ABIT KT7 running a Duron 800 @ 1000, 1.9 volts, Prime 95. Cooling was with an Alpha PEP66, rotated shroud, with a Delta 38 fan. To get a range of readings, I used a variable power supply to power the fan from 10 to 13 volts, recording temps with both the thermocouples and thermistor. The heatsink was never removed as voltages were varied, nor was the board shut down during the test. CPU temps were recorded with a thermocouple inserted into the base of the PEP66. Ambient temps did not change at 22.3 C. The Omega HH23 digital thermometer was used to record temps.

CPU Temps

TEST RESULTS

Voltage

CPU Temp

Thermocouple Temp

Thermistor Temp

10 Volts

41.2 C

48.1 C

43 C

11 Volts

40.6 C

47.5 C

42 C

12 Volts

40.2 C

46.8 C

41 C

13 Volts

39.5 C

45.6 C

40 C

As you can see, CPU and the in-socket thermistor temps are close; however, note that the actual CPU back temp, as measured by the thermistor, is much higher, indicating some interpretation in BIOS. As long as it’s linear, then the relative relationship is OK. Note, however, that this does NOT mean that using the in-socket thermistor to compare different heatsinks is OK.

Because thermistors measure CPU Back temps, readings are influenced by the heatsink’s effect on the secondary heat path; these effects differ by heatsink. If you’re comparing different fans on the SAME heatsink, for example, the relative readings are valid. As soon as you switch heatsinks, different interaction with secondary heatpaths renders any comparison questionable at best, and at worst, invalid.

Let me also clearly state that I can’t tell you that a ThermoEngine, for example, will show the same relative pattern; every heatsink will vary in their differences between thermistor, actual and CPU temps – especially for different designs. Use the in-socket thermistor as your fire alarm – don’t expect much more than that.

Further, I can’t tell you definitively that other motherboards do the same thing, although it appears that most do.

Email Joe


Be the first to comment

Leave a Reply