Thanks a lot for the tip. With it, I was able to identify my precise processor family (Family: Fh, Model: 2F, Stepping: 2) and perform some experiments here.
Fortunately, I still have XP in dual-boot with Windows 7 here in the same computer. So, I installed 32-bit TThrottler on XP and compared it with NextSensor's readings. Initially, I noticed there was a difference of about 35 degrees between them. So, I changed the "AMD Factor" from 49 to 14. TThrottler came close, but still 1-3 degrees below NextSensor's readings, in a non-linear, narrowly oscillating relationship.
The stress program provided with TThrottler had no effect at all, perhaps because it uses idle cycles as well, and then AMD Cool'n'Quiet throttles the processor by itself. I could tweak the parameters, but I found it easier to run SuperPI instead to stress the processor. Still an average of 2 degrees lower than NextSensor's readings under stress. So I changed the factor to 11, and TThrottler was consistently 2 degrees over NextSensor's readings. Too much correction. Then I tried 12. TThrottler was still an average 2 degrees above when the system was idle, but under stress the two programs matched almost exactly.
So, it seems 12 is the ideal value for XP in my PC, but I found it better to err on the side of caution and use 11 on Windows 7 (and use 66 rather than 68°C as the throttling threshold). Then I ran SuperPI again on Windows 7. Unfortunately, it didn't throttle. Temperature reached 69°C (as displayed by Throttler's floating bar) and I aborted SuperPI. After it cooled back down to the 40s, I ran Throttler again, this time as Administrator, and repeated the SuperPI test. Again, it reached 69°C, when there was a rule for Throttler to trigger at over 66°C. In such cases, NextSensor instantly cools my processor by some 5 or 6°C, albeit just for a few seconds before the processing heats the processor again. I was expecting something like that.
It seems Throttler is really incompatible with my processor. It accurately shows the change in temperature, but seems unable to do anything about it. Any suggestions? Thanks again.
Fortunately, I still have XP in dual-boot with Windows 7 here in the same computer. So, I installed 32-bit TThrottler on XP and compared it with NextSensor's readings. Initially, I noticed there was a difference of about 35 degrees between them. So, I changed the "AMD Factor" from 49 to 14. TThrottler came close, but still 1-3 degrees below NextSensor's readings, in a non-linear, narrowly oscillating relationship.
The stress program provided with TThrottler had no effect at all, perhaps because it uses idle cycles as well, and then AMD Cool'n'Quiet throttles the processor by itself. I could tweak the parameters, but I found it easier to run SuperPI instead to stress the processor. Still an average of 2 degrees lower than NextSensor's readings under stress. So I changed the factor to 11, and TThrottler was consistently 2 degrees over NextSensor's readings. Too much correction. Then I tried 12. TThrottler was still an average 2 degrees above when the system was idle, but under stress the two programs matched almost exactly.
So, it seems 12 is the ideal value for XP in my PC, but I found it better to err on the side of caution and use 11 on Windows 7 (and use 66 rather than 68°C as the throttling threshold). Then I ran SuperPI again on Windows 7. Unfortunately, it didn't throttle. Temperature reached 69°C (as displayed by Throttler's floating bar) and I aborted SuperPI. After it cooled back down to the 40s, I ran Throttler again, this time as Administrator, and repeated the SuperPI test. Again, it reached 69°C, when there was a rule for Throttler to trigger at over 66°C. In such cases, NextSensor instantly cools my processor by some 5 or 6°C, albeit just for a few seconds before the processing heats the processor again. I was expecting something like that.
It seems Throttler is really incompatible with my processor. It accurately shows the change in temperature, but seems unable to do anything about it. Any suggestions? Thanks again.