I don't have a Mac Pro, but I have noticed that on my 2.93 GHz Core 2 Duo MBP, CPU usage jumps to 85% and temperatures to 151 ºF (66 ºC) when I convert an audio track to Apple Lossless. This does not happen when I rip audio CDs. Hmmm...
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Audio Decoding KILLS MacPro (2009) green factor
- Thread starter handheldgames
- Start date
- Sort by reaction score
I never had time to dig into this today... maybe tomorrow.
I did manage to load up Hardware monitor and have a look... it shows actual Tjunction core temps which are even higher than what must be the Tcase temp being reported by iStat. This is normal as Tjunction is typically 15-deg C higher than Tcase. I say the whole lot is still screwed up though. Some people like Handheldgames (?) show an idle temp only 6 deg. C above ambient which is extremely doubtful for any stock air cooling setup. My idle core temps are reported to be around 70-deg. C and there's no way they are actually that hot... my exhaust air is only mildly warmer than ambient. I'm guessing they are off by 30-deg. C. but I'll dig into this more tomorrow.
I also had some time on my MBA, so I went back through this thread and compiled what data people provided into the table below.
The numbers are all over the place with a couple people having high idle temps and only incremental increases when decoding audio to others with very low idle temps and dramatic increases when decoding audio.
It would be great if we could get some more consistent test information from various folks... to fill out the table. While it may not seem relevant, things like GPU and how your audio is being output may turn out to be relevant. Providing ambient temps (and even room temps with an external thermometer) with your idle and music decoding temps (I guess we should just stick with iStat) helps make them comparable with others. And finally, it would be useful for us to agree on a stress test that we can all use to load up all cores to determine our max temps under a stress test load.
BTW, the NC in the fans column indicates that those users noted that there was "No Change" in fan speed as a result in the audio decoding test. How can a 30-deg increase in temps by some not impact fan speeds?!
@smacman... I couldn't find your report on temps (although I could have easily missed it)... what is your situation?
I did manage to load up Hardware monitor and have a look... it shows actual Tjunction core temps which are even higher than what must be the Tcase temp being reported by iStat. This is normal as Tjunction is typically 15-deg C higher than Tcase. I say the whole lot is still screwed up though. Some people like Handheldgames (?) show an idle temp only 6 deg. C above ambient which is extremely doubtful for any stock air cooling setup. My idle core temps are reported to be around 70-deg. C and there's no way they are actually that hot... my exhaust air is only mildly warmer than ambient. I'm guessing they are off by 30-deg. C. but I'll dig into this more tomorrow.
I also had some time on my MBA, so I went back through this thread and compiled what data people provided into the table below.
The numbers are all over the place with a couple people having high idle temps and only incremental increases when decoding audio to others with very low idle temps and dramatic increases when decoding audio.
It would be great if we could get some more consistent test information from various folks... to fill out the table. While it may not seem relevant, things like GPU and how your audio is being output may turn out to be relevant. Providing ambient temps (and even room temps with an external thermometer) with your idle and music decoding temps (I guess we should just stick with iStat) helps make them comparable with others. And finally, it would be useful for us to agree on a stress test that we can all use to load up all cores to determine our max temps under a stress test load.
BTW, the NC in the fans column indicates that those users noted that there was "No Change" in fan speed as a result in the audio decoding test. How can a 30-deg increase in temps by some not impact fan speeds?!
@smacman... I couldn't find your report on temps (although I could have easily missed it)... what is your situation?
Attachments
not related.
converting audio means changing the format from one to another - e.g. AIFF to MP3.
ripping simply copies the file from one medium to another. e.g. copy AIFF from CD to AIFF on computer.
I dont want to mess up the progress on this - or add additional variables... But....
If you really want to see the temps fly HIGH! Try ripping a DVD with the new 64-bit version of Handbrake and listen to some music. My temps hit 76(c) before I stopped the test..
I could never imagine a scenario where people rip a DVD and listen to music at the same time! DOH!
If you really want to see the temps fly HIGH! Try ripping a DVD with the new 64-bit version of Handbrake and listen to some music. My temps hit 76(c) before I stopped the test..
I could never imagine a scenario where people rip a DVD and listen to music at the same time! DOH!
My test results
Ok, I was able to reproduce the temperature increase. And quantify what's happening with HW monitor, an IR thermometer on the intake/exhaust/HSF, and with an AC energy meter to measure the AC draw of the Mac Pro.
Results: in the table below.
Table Clarifications:
Idle: is powering up the Mac Pro with no USB attachments (including my USB audio interface or the 24" LED ACD USB connector) and reaching steady state after 30 min of unattended operation.
Play Audio: As above but with iTunes playing music for 30+ min.
Delta: The difference between idle and play audio.
4 Core CPUTest: Stress test using CPUTest with Huge test type, 10 repetitions, and 4 instances (to load 4 physical cores).
Delta: The difference between idle and the CPUTest stress test
HW Monitor: The temps reported by Hardware Monitor.
Power Meter: The AC power draw of the Mac Pro as shown by my power meter. Calculated DC draw assuming 80% efficiency is also noted.
IR Thermometer: Black and Decker IR Thermometer which is accurate to +/- 3 deg F. Targets were the intake front cheese grate, rear exhaust vent, and the side of the HSF. Pics illustrate how these measurements were made.
Digital Thermometer: A digital thermometer placed near the Mac Pro throughout the testing to ensure room temps were stable.
Observations:
- Just plugging in my USB audio I/F causes the temp increase. Since this is always on, I was suffering from this temp increase all the time. Disconnecting it allowed my to reproduce the effects others were seeing.
- The temp increase when using my audio I/F or playing music in iTunes resulted in an average core temp increase of about 33 deg. C.
- The temp increase when running a full 4 core stress test was about 55 deg. C.
- The delta increase in temps for the audio test is about 60% of the stress test. The delta increase in AC power draw is 58%.
- The temp increase is real as the IR thermometer indicates a clear increase in exhaust and HSF temperatures.
- Under the stress test, my core temperatures reached a maximum of around 95-deg C before the fans ramped up, and even then, BOOSTA increased 150 RPM from 850 to 1000 and both Intake and Exhaust fan speeds increased 100 RPM from 600 to 700. Core temps then stabilized at around 94-deg. C.!
While I have no idea how accurate these temps are, I'm shocked the cooling system is so slow to ramp up.
Conclusions:
While I know I've been slow in coming to this conclusion and questioning the validity of the reported temps (which I still think was a good first step), I can definitely throw my support behind the fact that there is some bizarre load being put on the CPU that goes undetected by activity monitor but is equivalent to about 60% of a fully loaded quad core stress test... indicating that audio decoding, or in my case, just connecting my USB audio interface, is equivalent to fully loading approx. two cores.
I thank everyone for their patience with my previous posts as I know it must have been difficult... thinking "Who does this noob think he is!?"
I think there are three significant problems here:
1. Audio playback or even connecting a USB I/F causes this extraordinary load on the CPU
2. This load is somehow not being reported by activity monitor
3. The CPU cooling is very slow to ramp up
I'm not sure where to go next. The fact that simply connecting my USB audio interface causes this kind of load on the processor indicates there must be something seriously wrong with the CoreAudio subsystem on the Nehalem Xeon's.
I'll definitely log a call with Apple.
The following photos were captured during the audio playback test.
Ok, I was able to reproduce the temperature increase. And quantify what's happening with HW monitor, an IR thermometer on the intake/exhaust/HSF, and with an AC energy meter to measure the AC draw of the Mac Pro.
Results: in the table below.
Table Clarifications:
Idle: is powering up the Mac Pro with no USB attachments (including my USB audio interface or the 24" LED ACD USB connector) and reaching steady state after 30 min of unattended operation.
Play Audio: As above but with iTunes playing music for 30+ min.
Delta: The difference between idle and play audio.
4 Core CPUTest: Stress test using CPUTest with Huge test type, 10 repetitions, and 4 instances (to load 4 physical cores).
Delta: The difference between idle and the CPUTest stress test
HW Monitor: The temps reported by Hardware Monitor.
Power Meter: The AC power draw of the Mac Pro as shown by my power meter. Calculated DC draw assuming 80% efficiency is also noted.
IR Thermometer: Black and Decker IR Thermometer which is accurate to +/- 3 deg F. Targets were the intake front cheese grate, rear exhaust vent, and the side of the HSF. Pics illustrate how these measurements were made.
Digital Thermometer: A digital thermometer placed near the Mac Pro throughout the testing to ensure room temps were stable.
Observations:
- Just plugging in my USB audio I/F causes the temp increase. Since this is always on, I was suffering from this temp increase all the time. Disconnecting it allowed my to reproduce the effects others were seeing.
- The temp increase when using my audio I/F or playing music in iTunes resulted in an average core temp increase of about 33 deg. C.
- The temp increase when running a full 4 core stress test was about 55 deg. C.
- The delta increase in temps for the audio test is about 60% of the stress test. The delta increase in AC power draw is 58%.
- The temp increase is real as the IR thermometer indicates a clear increase in exhaust and HSF temperatures.
- Under the stress test, my core temperatures reached a maximum of around 95-deg C before the fans ramped up, and even then, BOOSTA increased 150 RPM from 850 to 1000 and both Intake and Exhaust fan speeds increased 100 RPM from 600 to 700. Core temps then stabilized at around 94-deg. C.!
While I have no idea how accurate these temps are, I'm shocked the cooling system is so slow to ramp up.Conclusions:
While I know I've been slow in coming to this conclusion and questioning the validity of the reported temps (which I still think was a good first step), I can definitely throw my support behind the fact that there is some bizarre load being put on the CPU that goes undetected by activity monitor but is equivalent to about 60% of a fully loaded quad core stress test... indicating that audio decoding, or in my case, just connecting my USB audio interface, is equivalent to fully loading approx. two cores.
I thank everyone for their patience with my previous posts as I know it must have been difficult... thinking "Who does this noob think he is!?"
I think there are three significant problems here:
1. Audio playback or even connecting a USB I/F causes this extraordinary load on the CPU
2. This load is somehow not being reported by activity monitor
3. The CPU cooling is very slow to ramp up
I'm not sure where to go next. The fact that simply connecting my USB audio interface causes this kind of load on the processor indicates there must be something seriously wrong with the CoreAudio subsystem on the Nehalem Xeon's.
I'll definitely log a call with Apple.
The following photos were captured during the audio playback test.
Attachments
More Results
Here are the HW Monitor Screenshots.
The first is at idle.
The second is while audio is playing.
The third one is after disconnecting my USB Audio Interface... the temps drop very quickly back to idle.
The final one is under the 4-core stress test. Note the point at which the fans ramp up appears where the core temps hit around 96-deg. or Tcase is around 70-deg.
Here are the HW Monitor Screenshots.
The first is at idle.
The second is while audio is playing.
The third one is after disconnecting my USB Audio Interface... the temps drop very quickly back to idle.
The final one is under the 4-core stress test. Note the point at which the fans ramp up appears where the core temps hit around 96-deg. or Tcase is around 70-deg.
Attachments
VirtualRain,
Nice work and testing! I had a feeling something was causing your machine to be stuck in this strange state and it looks like you have confirmed that. In any case, your thorough testing and observations are very useful and will hopefully help build our case.
I am over 6000 miles away from my Mac Pro but when I'm back on Wednesday, I'll reply with my exact temp stats..
Thanks!
Nice work and testing! I had a feeling something was causing your machine to be stuck in this strange state and it looks like you have confirmed that. In any case, your thorough testing and observations are very useful and will hopefully help build our case.
I am over 6000 miles away from my Mac Pro but when I'm back on Wednesday, I'll reply with my exact temp stats..
Thanks!
Nice work VirtualRain. And at least you've a better idea of what's gong on. Any report to Apple is a good thing, and if enough come in, will push the priority level of the situation.
I'm concerned that Seisend is in the same situation, as the systems used for music editing, and think the situation is very similar to yours.
Hopefully, he'll see this, and be willing to take a look at his system's behavior.
And at least you've a better idea of what's gong on. Any report to Apple is a good thing, and if enough come in, will push the priority level of the situation.I'm concerned that Seisend is in the same situation, as the systems used for music editing, and think the situation is very similar to yours.
Hopefully, he'll see this, and be willing to take a look at his system's behavior.
For stress tests, you need to use an instance for each thread. if you have hyperthreading on, you need to run one for each virtual core, so 8 on a quad core, and 16 on an octo-core or you wont get true temps. These are the kind of stats I was looking for. if some of you could run your 2009 mac pro with 8 instances for 20 min, that would help greatly. thanks.
I hope we can get Apple to look into this... it is bizarre and alarming! Hopefully the more evidence we pile on... the better. HandHeldGames has spent a lot of time and effort on this issue, hopefully some added ammunition will help the cause.
Nice work VirtualRain.
Let's hope so.
Go for it dude! I don't see a need personally. The 4 threaded test was already enough to illustrate that the Mac Pro cooling system has issues.
Hardly matters in this case.
He has proved there is a MASSIVE issue in the Mac Pro Audio Subsystem
well, when I use prime95 on my i7 PC at 4GHz, I get much lower temps when I only use 4 instances of prime 95. Around 70 degrees with 8 instances, and under 60 degrees with 4. but yah, I can see why you wouldnt want to go higher with the temps you are already getting. if more people could do the 4 instance one, that would be great.
That is really high. There is no reason for a premium computer to go that high in temps. They seem to really have screwed up the cooling in the Mac pros.
I'm taking into account that the temp reporting isn't calibrated, and could be 15C lower ( = 75C). Still high, but not as bad as 90C+ (especially in terms of electromigration).
Now if the 90C+ is more accurate than that, or worse, it is calibrated, then that's just NUTZ.
I'd erased this bit from the previous post, and it seems that was a mistake. I'm listening music with iTunes for about 2 hours now, with no other CPU heavy apps running (just a little coding in Eclipse but that doesn't need any CPU at all).
istat says 52°C for CPU A and 30.25W.
CPU B has 40°C and uses 27.81W.
I really can't image that a difference of less than 3 Watts can justify a temperature increase of 12°.
My report to Apple concerning this audio issue has been send last week.
istat says 52°C for CPU A and 30.25W.
CPU B has 40°C and uses 27.81W.
I really can't image that a difference of less than 3 Watts can justify a temperature increase of 12°.
My report to Apple concerning this audio issue has been send last week.
We don't know what TjMax value is being used on the OSX temp utilities... and therefore no idea whether the reported temps are accurate... so there's no point comparing between OSX and Windows. However, I'm willing to bet your cooling solution is much more effective than that in the Mac Pro which only ramps up the HSF fan to meager 1000 RPM at around 95-deg C core temp.
We don't know what TjMax value is being used on the OSX temp utilities... and therefore no idea whether the reported temps are accurate... so there's no point comparing between OSX and Windows. However, I'm willing to bet your cooling solution is much more effective than that in the Mac Pro which only ramps up the HSF fan to meager 1000 RPM at around 95-deg C core temp.
I know, i know. I wasnt comparing it to windows, I was just saying that using an instance for each thread is a better test to get the max temperature of your system and was giving an example of my computer using 4 and 8. And yes my sealed liquid cooling is probably better, but I wont get into that
I know, i know. I wasnt comparing it to windows, I was just saying that using an instance for each thread is a better test to get the max temperature of your system and was giving an example of my computer using 4 and 8. And yes my sealed liquid cooling is probably better, but I wont get into that
Yeah, I hear you... I'm intimately familiar with Prime95, water cooling, and overclocking. I intentionally chose to run a 4 threaded stress test to see what impact it would have just loading the physical cores, realizing I wouldn't be fully stressing the HT aspects... but it turns out that until we know more, the temps seam dangerously high and the cooling inadequate for a full balls-out stress test!
a 486 dx2-66 can decode mp3's in real time. big software bug if a new mac pro is taking 100% cpu to perform the same.
I thought I already posted this, but my message seems to have disappeared. I suggest we forget sending complaints to Apple and go directly to the media. Let's start writing letters and sending emails to the editors of the major Mac publications and news outlets. Apple needs the smack-down of bad press before it does anything to remedy this issue.
As long as the complaints are made directly to Apple, Apple will do nothing to fix the issue because it is not receiving any public scrutiny or attention. A thread in a Mac forum is not going to do anything to make Apple fix the problem. Perhaps some discussion of a class action is in order here as well. I don't think temperature is as big of an issue as the 20% performance penalty when playing music that should take less than 1-2% of CPU power. THAT is a very big deal, if true, and cause for claims that Apple misrepresents the capabilities of their system.
As long as the complaints are made directly to Apple, Apple will do nothing to fix the issue because it is not receiving any public scrutiny or attention. A thread in a Mac forum is not going to do anything to make Apple fix the problem. Perhaps some discussion of a class action is in order here as well. I don't think temperature is as big of an issue as the 20% performance penalty when playing music that should take less than 1-2% of CPU power. THAT is a very big deal, if true, and cause for claims that Apple misrepresents the capabilities of their system.
Are you volunteering to contact the major media? I'm all for throwing my support behind this. Feel free to PM me if you want my email or something further.