ALCPU Forums

I am using Core Temp 1.0 RC7 on an AMD A10-5700 running Windows 8.1. Compared to my Core Temp experience with Windows 7 on an AMD Phenom and XP on an AMD Athlon, I am seeing some things that surprise me.

The temperature reading is very erratic. It can jump up more than 10degC in a single 1-second sample. Then it typically lowers itself over the next several seconds back down to hovering around 50. (This is when nothing big is going on.) TJMAX is listed at 70; but some of the jumps are well over that. Indeed, after Core Temp has been running for a few hours, I will generally see a reported max as high as 84. With TJMAX at 70; that tends to worry me. Should it?

It never behaved this way with the other systems. With them, temperature changes were always gradual. I am using a beta version because the release version did not work at all. Could this erratic behavior be misbehavior? Or could it be that there is a lot less thermal lag between the sensor and the source of the heat with these new APUs? Could it be that the power regulation is agile enough that the brief high temperatures can be ignored? Would it make sense to introduce some low pass filtering on the reported temperature to avoid anxiety? E.g., a configurable "average temperature over [parameter] seconds".

Sadly the thermal sensors on AMD processors are not very reliable as it seems. Especially on the new APUs.
There is not much I can do from my side since Core Temp can only display the value it reads from the processor.

FWIW the newer FX processors also jump around in temp a lot per-second because of AMD's more sophisticated power states used in these processors and the sampling rate.

I don't believe 70C is the correct TjMax for the APUs as the graphic section can drive the processor temp higher than a normal discrete CPU. The APUs are closer to a notebook type of processor which is designed to run at higher temps because they have limited cooling. Unfortunately AMD sucks at providing technical information on their products and temps are some thing of a national security secret - to AMD. Their advanced power states on the latest Beema/Mullins is also different so expect lots of changes going forward.

You can indirectly find the TjMax if your temp gets too high as the APU will start to throttle the CPU frequency to lower the temp. Under sustained max load the frequency will oscillate back and forth at whatever the true TjMax temp is to prevent the APU from reaching the safety shut down temp. There is no need to worry about it burning up because it will shut-off before that can happen.