On 11/5/2014 5:50 PM, Guenter Roeck wrote: > On Wed, Nov 05, 2014 at 06:34:20PM -0500, James Cloos wrote: >>>>>>> "GR" == Guenter Roeck <linux at roeck-us.net> writes: >> ... >> GR> AMD CPUs are known report wildly off temperatures especially at >> GR> low temperatures. >> >> Thanks. After some further digging, I wonder whether Aravind's recent >> patch might help? The apu does have both PCI_DEVICE_ID_AMD_16H_M30H_NB_F3 >> and PCI_DEVICE_ID_AMD_16H_M30H_NB_F4 and lspci reports no driver for the >> F4 device. >> > Aravind would have to comment; my understanding is that the F4 device > reports the power, not the temperature. You could try to apply the patch > and see if it makes a difference. Worst case there will be no change ;-).
Right, F4 is for reporting power values. The recent patch was a bug fix for fam15h_power, so I don't think that's what you are looking for.. >> GR> Does the temperature increase with load ? >> >> It seems to max out w/in a kelvin or two of the room's ambient temp. >> > Even under load ? Sounds weird. Yeah, I have not come across such odd behavior neither. So not able to tell you what might be wrong.. > Question of course is if the chip really > reports a useful temperature in the first place; this is something I can > not answer, since I have neither a datasheet nor a system with that CPU. Yes, the values reported are accurate and AFAIK there are folks who are interested to look into temperature/power values reported by hwmon, so I guess it's useful :) I shall try and ask around if someone knows anything about incorrect values being exposed in the registers and let you know. Thanks, -Aravind
