Firewire Audio & IntelCPUPowerManagement causes CPU draw more power and heat increase

[EuphangeL]

I have found that by plugging in my Apogee Duet firewire audio interface and ONLY WHEN AppleFWAudio was loaded there was a 10c CPU raise that was making my MBP (late 2008) hot!

I wanted to monitor the power that was required for Duet to run and to see if the temperature rise was due to that, so:

• With Apogee drivers uninstalled (just to be sure its not their drivers related, which is not)
• MBP on battery power with System Profiler opened (to monitor the mA required by the computer from System Profiler -> Power)
• CoolBook opened to monitor the temps
• Terminal opened to load/unload the AppleFWAudio
• GeekBench on desktop to monitor the CPU performance

I did the following tests:

1st Test:

Idling machine. Duet is unplugged, AppleFWAudio.kext is not loaded

Results:

Idle CPU temps: ~40c :: Power Consumption: ~ -900mA :: GeekBench Score: ~3550

2nd Test:

Im plugging in Duet. AppleFWAudio gets loaded. Duet outputs audio

Results:

Idle CPU temps: ~50c :: Power Consumption: ~ -1850mA :: GeekBench Score: ~3450

3rd Test:

With Duet plugged in. I kextunload AppleFWAudio. Duet cannot longer output. The device is disappeared from System Preferences > Audio

Results:

Idle CPU temps: ~41c :: Power Consumption: ~ -1300mA :: GeekBench Score: ~3550 (like 1st Test)

4th Test:

I unplug the Duet. Wait for a minute

Results:

Idle CPU temps: ~41c :: Power Consumption: ~ -900mA :: GeekBench Score: ~3550

5th Test:

I plug the Duet again. The AppleFWAudio is not auto loaded since i unloaded previously. Duet cannot output

Results:

Idle CPU temps: ~42c :: Power Consumption: ~ -1300mA :: GeekBench Score: ~3550

6th Test:

With Duet plugged in, i kextload the ApppleFWAudio again. Duet can output again

Results:

Idle CPU temps: ~51c :: Power Consumption: ~ -1800mA :: GeekBench Score: ~3450 (100 points drop again)

I dont know if i'm paranoid but i think i bought a Firewire audio interface because it is considered for professional use. To my eyes there is nothing professional in rapid speedstep switching (bad BAD for AD/DA clocking) neither in the use of CPU power (even if this is just as little a.k.a. 100 points on geekbench). But most of all, CPU heat bothers me the most. I was working with a USB bus powered interface and my MBP was never getting that hot!

You can also watch a video i made that shows the rapid speedstep switch here: http://www.youtube.com/watch?v=JYHnX3CGUoo

I downloaded iStat Menus. I was reading the issue with the MacPros and they were saying that although no CPU usage is occuring, a CPU temp raise is there. Exactly my issue.

They found that this was because there was a power consumption increase by the CPU and iStat Menus has the ability to display the current power usage from the CPU and here's what i've found:

CPU Power with Firewire I/F unplugged: 0.95W - 1.33 W
CPU Power with Firewire I/F plugged: 3.91W - 4.10W

By disabling IntelCPUPowerManagement either by deleting the kext or by installing NullCPUPowerManagement i get a constant ~2.50W CPU power either with or without the interface plugged in.

That means the IntelCPUPowerManagement sets the CPU to draw more power for some reason when the interface is plugged. By disabling it the CPU power remains the same as when idle without the interface but because it's disabled, there is no power management for the CPU and the ~2.50W are above the required for powering an idling CPU

With IntelCPUPowerManagement enabled i get 0.95W to 1.30W when idling but this goes up to 4.10W when connecting the interface which is why the CPU produces more heat.

If this is not an AppleFWAudio & IntelCPUPowerManagement bug but its just the way it works, id like to have an official word for it.

I've filled a bug report at bugreport.apple.com using the same info as here. Will hope for an answer.

Does anyone else experiencing the same? If anyone has the time to do the same tests, it would be nice to have the results.
The commands for the terminal:

sudo kextstat -> Lists the currently loaded kexts
sudo kextunload -b com.apple.driver.AppleFWAudio -> Unloads the AppleFWAudio (dont be bothered about the classes report after unloaded) will disable output
sudo kextload "/System/Library/Extensions/AppleFWAudio.kext" -> Loads the AppleFWAudio and will enable output

For disabling IntelCPUPowerManagement i used NullCPUPowerManagement.kext

 

[Roman23]

Funny, I don't

suffer from this bug at all.. but then again my mac pro is a single-quad core and was a refurb..that came in this month and just purchased it. Looks brand new to me though.

good luck with that bug you have.

I have found that by plugging in my Apogee Duet firewire audio interface and ONLY WHEN AppleFWAudio was loaded there was a 10c CPU raise that was making my MBP (late 2008) hot!

I wanted to monitor the power that was required for Duet to run and to see if the temperature rise was due to that, so:

• With Apogee drivers uninstalled (just to be sure its not their drivers related, which is not)
• MBP on battery power with System Profiler opened (to monitor the mA required by the computer from System Profiler -> Power)
• CoolBook opened to monitor the temps
• Terminal opened to load/unload the AppleFWAudio
• GeekBench on desktop to monitor the CPU performance

I did the following tests:

1st Test:

Idling machine. Duet is unplugged, AppleFWAudio.kext is not loaded

Results:

Idle CPU temps: ~40c :: Power Consumption: ~ -900mA :: GeekBench Score: ~3550

2nd Test:

Im plugging in Duet. AppleFWAudio gets loaded. Duet outputs audio

Results:

Idle CPU temps: ~50c :: Power Consumption: ~ -1850mA :: GeekBench Score: ~3450

3rd Test:

With Duet plugged in. I kextunload AppleFWAudio. Duet cannot longer output. The device is disappeared from System Preferences > Audio

Results:

Idle CPU temps: ~41c :: Power Consumption: ~ -1300mA :: GeekBench Score: ~3550 (like 1st Test)

4th Test:

I unplug the Duet. Wait for a minute

Results:

Idle CPU temps: ~41c :: Power Consumption: ~ -900mA :: GeekBench Score: ~3550

5th Test:

I plug the Duet again. The AppleFWAudio is not auto loaded since i unloaded previously. Duet cannot output

Results:

Idle CPU temps: ~42c :: Power Consumption: ~ -1300mA :: GeekBench Score: ~3550

6th Test:

With Duet plugged in, i kextload the ApppleFWAudio again. Duet can output again

Results:

Idle CPU temps: ~51c :: Power Consumption: ~ -1800mA :: GeekBench Score: ~3450 (100 points drop again)

I dont know if i'm paranoid but i think i bought a Firewire audio interface because it is considered for professional use. To my eyes there is nothing professional in rapid speedstep switching (bad BAD for AD/DA clocking) neither in the use of CPU power (even if this is just as little a.k.a. 100 points on geekbench). But most of all, CPU heat bothers me the most. I was working with a USB bus powered interface and my MBP was never getting that hot!

You can also watch a video i made that shows the rapid speedstep switch here: http://www.youtube.com/watch?v=JYHnX3CGUoo

I downloaded iStat Menus. I was reading the issue with the MacPros and they were saying that although no CPU usage is occuring, a CPU temp raise is there. Exactly my issue.

They found that this was because there was a power consumption increase by the CPU and iStat Menus has the ability to display the current power usage from the CPU and here's what i've found:

CPU Power with Firewire I/F unplugged: 0.95W - 1.33 W
CPU Power with Firewire I/F plugged: 3.91W - 4.10W

By disabling IntelCPUPowerManagement either by deleting the kext or by installing NullCPUPowerManagement i get a constant ~2.50W CPU power either with or without the interface plugged in.

That means the IntelCPUPowerManagement sets the CPU to draw more power for some reason when the interface is plugged. By disabling it the CPU power remains the same as when idle without the interface but because it's disabled, there is no power management for the CPU and the ~2.50W are above the required for powering an idling CPU

With IntelCPUPowerManagement enabled i get 0.95W to 1.30W when idling but this goes up to 4.10W when connecting the interface which is why the CPU produces more heat.

If this is not an AppleFWAudio & IntelCPUPowerManagement bug but its just the way it works, id like to have an official word for it.

I've filled a bug report at bugreport.apple.com using the same info as here. Will hope for an answer.

Does anyone else experiencing the same? If anyone has the time to do the same tests, it would be nice to have the results.
The commands for the terminal:

sudo kextstat -> Lists the currently loaded kexts
sudo kextunload -b com.apple.driver.AppleFWAudio -> Unloads the AppleFWAudio (dont be bothered about the classes report after unloaded) will disable output
sudo kextload "/System/Library/Extensions/AppleFWAudio.kext" -> Loads the AppleFWAudio and will enable output

For disabling IntelCPUPowerManagement i used NullCPUPowerManagement.kext

 

[seisend]

Hi EuphangeL. Thanks for your post and welcome to the forum !

I have a MP09 and MBP07 with the Apogee Ensemble. I remember back in the days with the MP09 problem with the "audio/firewire/heat" bug that I experienced CPU heat increase by 10°C when I connected my Apogee Ensemble to my MBP!

I don't know much about power consumption but you know that your Apogee Duet needs power and it has 48V Phantom power built in.

Also the 100 point drop with GeekScore doesn't look like that much. It looks normal to me. Correct me if I am wrong.

I worked two years with this setup (MBP+Apogee Ensemble). I don't believe that you recognize a difference in performance in practice except for the CPU heat increase.

Maybe there is more behind this problem, but I don't think so.

Would be interesting what other users have to say about this.

As for now, I wouldn't get too deep into this and create some nice music instead

If there will be no attention to this thread I would recommend posting this in the MBP section of this forum, too.

 

[EuphangeL]

Hi EuphangeL. Thanks for your post and welcome to the forum !

I have a MP09 and MBP07 with the Apogee Ensemble. I remember back in the days with the MP09 problem with the "audio/firewire/heat" bug that I experienced CPU heat increase by 10°C when I connected my Apogee Ensemble to my MBP!

I don't know much about power consumption but you know that your Apogee Duet needs power and it has 48V Phantom power built in.

Also the 100 point drop with GeekScore doesn't look like that much. It looks normal to me. Correct me if I am wrong.

I worked two years with this setup (MBP+Apogee Ensemble). I don't believe that you recognize a difference in performance in practice except for the CPU heat increase.

Maybe there is more behind this problem, but I don't think so.

Would be interesting what other users have to say about this.

As for now, I wouldn't get too deep into this and create some nice music instead

If there will be no attention to this thread I would recommend posting this in the MBP section of this forum, too.

Hi seisend and thanks for your answer

Now this is not about the bus power Duet takes. It's clearly a kernel extensions related issue

No there is not a noticable CPU performance reduction but still, Firewire is know for NOT using the CPU at all as everyone suggests firewire for audio interfaces.

Actually i'm just waiting for an official Apple's answer. I have opened a bug report and i'm waiting

 

[EuphangeL]

Apple Engineering responded to my Bug Report:

25-Jun-2010 12:01 AM STEVEN ANDRIELLA :
Engineering has provided the following information:

This is a maxbusstall issue. FireWire isoch services call requiremaxbusstall(). This is expected.

We are now closing this bug since our engineers are aware of the issue and will continue to track it offline. A different resolution to this issue may be considered in the future.

-----

So anyone can explain exactly what that means? What's maxbusstall exactly? He says that FireWire Isochronous services call requiremaxbusstall(). But what is maxbusstall exactly? Does that means that the stream between the bus and the audio interface never idles?

 

[l3xh2k]

I can confirm that I see this CPU temp change using a Focusrite Saffire connected to a late '07 MBP.

It's pretty annoying because for me it puts the fan into a higher gear.

No I don't know what maxbusstall is (I have guesses based on experience in previous programming gigs but that's not really helpful).