Sidebar Sidebar
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

Guys you're not going to believe this... I started a thread over at ArsTechnica forums hoping someone might have some different insights on this issue.

A veteran poster named mckill responded with the following:

if it is the power management, you could simply try disabling it temporarily by remove AppleIntelCPUPowerManagement.kext in /System/Library/Extensions/ and rebooting.
Click to expand...

I finally found some time to do this, and conduct some testing. I am happy to report MAJOR PROGRESS.

Good News
  • Temperature no longer increases when audio is played
  • CPU appears to run ultra cool compared to before with max core temperatures under a stress test hovering around 80C
  • System seems stable

Bad News
  • There seems to be a performance loss of approximately 10% across all tests (Geekbench, Cinebench, Xbench) when compared to tests conducted with no audio playing prior to removal of the kext
  • Performance seem to be comparable to an unmodified Mac Pro playing audio. Multi-threaded benchmarks (ie. cinebench multiple cpu render test) are the exception and score slightly lower even than an unmodified system playing audio
  • I don't know what other features of the processor are not working with this kext removed
  • CPU consumes 21.5 watts at idle now as opposed to 3 watts before (I suspect Apple screwed up with this kext which should be switching in and out of CPU power saving state

EDIT:

This solution is not viable. See post below.
 
smacman said:
Guys you're not going to believe this... I started a thread over at ArsTechnica forums hoping someone might have some different insights on this issue.

A veteran poster named mckill responded with the following:



I finally found some time to do this, and conduct some testing. I am happy to report MAJOR PROGRESS.

Good News
  • Temperature no longer increases when audio is played
  • CPU appears to run ultra cool compared to before with max core temperatures under a stress test hovering around 80C
  • System seems stable

Bad News
  • There seems to be a performance loss of approximately 10% across all tests (Geekbench, Cinebench, Xbench) when compared to tests conducted with no audio playing prior to removal of the kext
  • Performance seem to be comparable to an unmodified Mac Pro playing audio. Multi-threaded benchmarks (ie. cinebench multiple cpu render test) are the exception and score slightly lower even than an unmodified system playing audio
  • I don't know what other features of the processor are not working with this kext removed
  • CPU consumes 21.5 watts at idle now as opposed to 3 watts before (I suspect Apple screwed up with this kext which should be switching in and out of CPU power saving state

It is clear that this is not an official sustainable solution, but what I do know is that until Apple issues a fix, I will be running without the kext. Whatever power management features this kext was intended to deliver are clearly not working properly anyway, so I would rather have a system that benchmarks slightly slower that it's true capability, but appears to operate normally in every other regard. Thoughts??
Click to expand...

hmmm. I wonder if it works with external FW devices such as HD's and Audio soundcards ? Can you test transfer files between a external FW harddrive ?

edit. thanks for the post by the way, maybe that will bring us a lot closer to the solution of the hole thing...
 
AZREOSpecialist said:
Running iTunes at 3% CPU utilization spikes the CPU 20C, and using the remaining 397% available to the 4 physical cores apparently causes only a 15C rise in temperature above iTunes levels.
Click to expand...

What about other non-music apps that use approx 3% CPU:
do they increase the CPU temperature by 20C?
 
Just found a major deal breaker. System can not wake from sleep, and instead, after a forced shutdown and restart, returns the following error:

panic(cpu 0 caller 0xffffff80002d3f95): "Error 4 trying to transition CPU 1 to OFF"@/SourceCache/xnu/xnu-1486.2.11/osfmk/i386/acpi.c:168

So I guess I will put the kext back, and be back to square 1. Damn it!!!!
 
smacman said:
Just found a major deal breaker. System can not wake from sleep, and instead, after a forced shutdown and restart, returns the following error:

panic(cpu 0 caller 0xffffff80002d3f95): "Error 4 trying to transition CPU 1 to OFF"@/SourceCache/xnu/xnu-1486.2.11/osfmk/i386/acpi.c:168

So I guess I will put the kext back, and be back to square 1. Damn it!!!!
Click to expand...

damn man that sucks :( at least we somewhat know the problem now though..
 
smacman said:
Just found a major deal breaker. System can not wake from sleep, and instead, after a forced shutdown and restart, returns the following error:

panic(cpu 0 caller 0xffffff80002d3f95): "Error 4 trying to transition CPU 1 to OFF"@/SourceCache/xnu/xnu-1486.2.11/osfmk/i386/acpi.c:168

So I guess I will put the kext back, and be back to square 1. Damn it!!!!
Click to expand...

Interesting find.

Maybe start looking inside the file? Can you post the contents?
 
DoFoT9 said:
i wonder, if it is just one setting somewhere, if its possible for us to change it ourselves?
Click to expand...

It's a compiled file so no fiddling by us.

We'd need someone with kext knowledge to help us out :/

Netkas could help, but I doubt even with the knowledge from the ATi flashing he knows how to make a kext for Intel Power saving :/
 
smogsy said:
Hey ive made a formalized letter to send to Anandtech,Wired,cnet,tech world
news,try for yahoo & for google news,twitter (lockergnome)

can you please verify ive not missed anything out?
Click to expand...

Thats good, few pointers.
"The above simple tasks results in the following problems:
Higher temperatures on the CPU(s) (From 25c ambient up to 90c)
Performance loss up to 25% (1-2 Cores)
Power consumption increase to 50w"

Should be:
"The above simple tasks results in the following problems:
Higher temperatures on the CPU(s), heatsink up from 30 to 50+, core temps up to 65+ from 40C.
Performance loss up to 25% (1-2 Cores)
Power consumption increases by 50W."

Otherwise fine!

Edit: Oops, double post, how embarrassing!
 
True enough, but to prevent damage to the system I have to address the heat issue. She uses her system for writing, browsing and music. The performance part of this issue is somewhat lost for her.

Would you have me ignore the heat issue and suffer a premature failure of the CPU?

Concorde Rules said:
People are on the wrong track.

Something to do with the CPU is being used, hence and foremost the performance drop, then the power usage and finally temperatures.

Fix the performance drop, fix the rest...
Click to expand...

My wife's system is a single quad core chip (1x2.93) and went from 35c to 59c when I stopped it after about 10 mins.

Installing smcFanControl, setting the fans to higher RPMs and re-running the test for about 20 mins resulted in a top temp of 44C.

So it also hits the single processor systems pretty hard as well.

cl-user said:
...

My conclusion, this audio bug hits 2.66 and 2.93 09 dualies worst (meaning bad enough to damage the CPUs). Not that any should waste power on audio.

...
Click to expand...
 
i emailed my modified story to Wired/Anandtech/tech world news/CNET/Macworld/engadget, after your suggestion cord.
if you can think of any others please state the url & ill get it posted. each story has been posted to the SEnior Editor of each company.

i got an email back from Wired

ylan Tweney - Wired.com to me, Brian
show details 3:58 PM (36 minutes ago)
Thanks for the tip. We've got our hands full with CES this week, but
we'll look into it next week.

dylan.

--
Dylan Tweney | Sr. Editor, Wired.com | dylan@tweney.com | wired@tweney.com

http://www.wired.com/gadgetlab/ | http://tweney.com
520 Third Street, Suite 305, San Francisco, CA 94107 USA
650-605-3348
Click to expand...
 
BillB50 said:
I question this logic - I read Anand's article on upgrading CPUs on a Nehalem MP. He found that 8 core MPs use special CPUs without the normal outer shield/heat spreader. In fact, he ruined his processor board during the tightening portion of the upgrade. I can't believe our issue boils down to improper assembly from the factory? I have been following this thread and don't recall anyone trying this and reporting success - am I incorrect.

Not intending a flame with this post, just questioning reseating. It seems that Apple has done enough research/work with Intel to order specific processors assembled without outer heat spreaders so that heat could more quickly be pulled away from the cores (here is a link to the article again:http://www.anandtech.com/mac/showdoc.aspx?i=3597&p=10 )...
Click to expand...

This only applies to 8-core systems. Quad systems from the factory utilize Intel Nehalem CPUs with integrated heat spreaders. That's why a CPU transplant on a Quad is much simpler and can utilize a retail CPU. My suggestion of checking the thermal paste was prompted after my results seemed much cooler than Smacman's. Other than software variables, the only other real variable would be the thermal paste application on the CPU. It's so much easier to check the thermal paste before taking it into the shop, and if that's all it takes to fix the problem why not give it a shot?

To each his own. Just trying to help.
 
AZREOSpecialist said:
This only applies to 8-core systems. Quad systems from the factory utilize Intel Nehalem CPUs with integrated heat spreaders. That's why a CPU transplant on a Quad is much simpler and can utilize a retail CPU. My suggestion of checking the thermal paste was prompted after my results seemed much cooler than Smacman's. Other than software variables, the only other real variable would be the thermal paste application on the CPU. It's so much easier to check the thermal paste before taking it into the shop, and if that's all it takes to fix the problem why not give it a shot?

To each his own. Just trying to help.
Click to expand...
Thermal Paste may certainly be a factor in terms of the exact temps reported (for an identical test on different '09 machines of the same specs), but there's the bigger problem here that needs to be solved: OS X doesn't run properly on the '09's, Quad or Octads in terms of heat, power, and performance reduction under certain conditions.
 
Concorde Rules said:
It's a compiled file so no fiddling by us.

We'd need someone with kext knowledge to help us out :/

Netkas could help, but I doubt even with the knowledge from the ATi flashing he knows how to make a kext for Intel Power saving :/
Click to expand...

Does the kext vary from system to system? Could we try using the kext from a quad core iMac?
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back