Ambient CPU Temperature readings don't change until reboot

Question

Hi everyone,
It's been a good while since I last posted on this forum and much has changed since...
I have recently bought a second hand "Early 2005" DP 2.7 Ghz G5 Powermac and everything has been working perfectly for the last couple of months - no traces of liquid coolant spills or odd fan behaviour so far (fingers crossed). I was just hoping to compare experiences with other users concerning the following behaviour though:
When I boot the machine from cold, both "iStat" or "Temperature Monitor" agree and correctly show the ambient temperature readings for both CPUs at actual room temperature as expected, which is usually 23ºC. This value doesn't get updated after hours of use until I reboot no matter into which OS. It will then show an updated value - usually ~40ºC which seems reasonably accurate.
I can replicate this in Tiger (10.4.11) and Leopard (10.5.8) using the same versions of both apps so it would seem as though the ambient temperature sensors only get read once at boot time rather than being polled at intervals like the the remaining sensors (HDD, CPU core, Backside and Drive Bay).
Has anyone else ever noticed this behaviour?
These are my usual temperature readings and fan speeds in "Reduced" processor performance, for future reference:
!http://i195.photobucket.com/albums/z126/thearchit/Apple%20Forum/iStatscreen01.png!
Cheers!

Answer 1

Hello The hatter,
(WARNING: long post!)
As previously mentioned, when I set this machine up I decided to have all three Operating Systems on it, since some apps may be compatible with one but not the other, and with 2TB of available space I couldn't see a reason why not to, so I proceeded with a fresh install of Panther, Tiger and Leopard.
I use Leopard for web and email, and Panther for audio recording. Tiger remains largely unused. Ok, I confess to have played Call of Duty 2 once - for testing purposes, of course
As the described behaviour occurs for both CPU Ambient Sensors, I'm ruling out any hardware defects. It could be something at a lower hardware level, Open Firmware, or limited OS API support. (i.e - AppleAD741x .kext)
Quite possibly, fan/pump thermal management doesn't care about readings from CPU ambient sensors. They may be there merely for informational purposes. For example, there is also a 3rd room temperature ambient sensor placed at the front of the tower, right beneath the front speaker/fan assembly that I've not seen it listed in either "Temperature Monitor" or "iStat", hence my reasoning that it is possible that the OS is actually polling these sensors on wake and bootup, but there may be no way of polling them manually on runtime due to the aforementioned API limitations or else.
I could contact the developers of iStat and Temperature Monitor to find out what they may know about it, as they are probably best placed to answer queries like these apart from Apple themselves, of course. But I hesitate about bothering them, as I haven't made any contributions and they have already given so much to the community for free.
I really wish that someone could test this to confirm it is expected behaviour. i.e. CPU ambient sensors only get polled at system boot or wake from sleep.
I've run iostat -l using the terminal and something caught my eye.
I'm guessing this should be the main object definition for all CPU B's hardware sensors. There's an interesting "platform-do-reset-on-wake" parameter:
(output clipped to 80 lines width)
| | | +-o supply-monitor@5a <class PPCI2CInterface, registered, matche$
| | | | | {
| | | | | "#address-cells" = <00000001>
| | | | | "hwsensor-zone" = <0000000000000000000000000000000000000000$
| | | | | "hwsensor-type" = <","adc","adc","adc">
*| | | | | "platform-do-reset-on-wake" = <0000000010000000000000000000$*
| | | | | "hwsensor-location" = <"hwsensor-id" = <0000000f00000010000000110000001200000013>
| | | | | "AAPL,phandle" = <ff989230>
| | | | | "hwsensor-polling-period" = <0000000a0000000a0000000a000000$
| | | | | "built-in" =
| | | | | "hwsensor-params-version" = <00000001>
| | | | | "name" = <" = <">
| | | | | "compatible" = <" = <0000005a>
| | | | | "#size-cells" = <00000000>
| | | | | }
Presumably, AMB stands for "Ambient" and ADC for "Analog Devices" Controller, the sensor manufacturer, most likely.
AD1 looks like this, and AD2,3,4 look the same:
| | | | +-o adc-sensor <class IOService, registered, matched, active$
| | | | | | {
| | | | | | "zone" = <00000000>
| | | | | | "location" = "CPU B AD7417 AD1"
| | | | | | "name" = <" = <">
| | | | | | "polling-period" = <0000000a>
| | | | | | "compatible" = <" = <00000001>
| | | | | | "type" = "adc"
| | | | | | "sensor-id" = <00000010>
| | | | | | }
| | | | | |
| | | | | +-o IOHWSensor <class IOHWSensor, registered, matched, act$
| | | | | {
| | | | | "IOProbeScore" = 0
| | | | | "current-value" = 575
| | | | | "CFBundleIdentifier" = "com.apple.driver.AppleHWSenso$
| | | | | "polling-period-ns" = 18446744073709551615
| | | | | "IOPropertyMatch" = {"device_type"="adc-sensor"}
| | | | | "IOMatchCategory" = "IODefaultMatchCategory"
| | | | | "version" = 1
| | | | | "type" = "adc"
| | | | | "IOPowerManagement" = {"CurrentPowerState"=1}
| | | | | "IOProviderClass" = "IOService"
| | | | | "zone" = <00000000>
| | | | | "location" = "CPU B AD7417 AD1"
| | | | | "polling-period" = 18446744073709551615
| | | | | "IOClass" = "IOHWSensor"
| | | | | "sensor-id" = 16
| | | | | }
| | | | |
Interestingly, after parsing "ioreg -l" output further, I can find objects for every sensor available to "iStat" when looking for its exact name, for example "Backside", "U3 Heatsink", "Drive Bay":
| | | +-o temp-monitor@98 <class PPCI2CInterface, registered, matched,$
| | | | | {
| | | | | "platform-do-set-conversion-rate" = <0000000090000000000000$
| | | | | "#address-cells" = <00000001>
| | | | | "hwsensor-zone" = <0000000400000004>
| | | | | "hwsensor-type" = <">
*| | | | | "hwsensor-location" = <">*
| | | | | "external-high-limit" = <00000055>
| | | | | "hwsensor-id" = <0000000600000007>
| | | | | "AAPL,phandle" = <ff9794f8>
| | | | | "internal-high-limit" = <00000055>
| | | | | "hwsensor-polling-period" = <0000000a0000000a>
| | | | | "built-in" =
| | | | | "hwsensor-params-version" = <00000001>
| | | | | "internal-low-limit" = <0000000a>
| | | | | "platform-do-set-limits" = <0000000010000000000000000000000$
| | | | | "name" = <" = <">
| | | | | "compatible" = <" = <00000098>
| | | | | "external-low-limit" = <0000000a>
| | | | | "#size-cells" = <00000000>
| | | | | }
| | | | |
| | | +-o temp-monitor@94 <class PPCI2CInterface, registered, matched,$
| | | | | {
| | | | | "#address-cells" = <00000001>
| | | | | "hwsensor-zone" = <00000001>
| | | | | "hwsensor-type" = <" = <">*
| | | | | "hwsensor-id" = <00000004>
| | | | | "AAPL,phandle" = <ff978b90>
| | | | | "hwsensor-polling-period" = <0000000a>
| | | | | "built-in" =
| | | | | "hwsensor-params-version" = <00000001>
| | | | | "name" = <" = <">
| | | | | "compatible" = <" = <00000094>
| | | | | "#size-cells" = <00000000>
| | | | | }
| | | | |
These PowerPC machines are amazing! Looks like there's fun to be had studying the IOKit framework. I'll post any findings.
Cheers