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

massive playback test

I did a test on my 2 week old 2009 2.66 Quad / ATI4870 / 3GB Ram / OS 10.6.1... played a MP3 4min. song using:

- Quicktime 7
- Quicktime 10
- VLC
- iTunes
- Realplayer

Test also included Safari with a flash based live-streaming radio station, and TV stream with Livestation (dw-tv).

Also running the finder, skype, livequartz, textedit, and a bunch of pref panes (SMCfancontrol, soundsource, iStatMenus, cuteclips, caffeine, littlesnitch). The fans were set to the default speeds during all tests (no SMC).

CPU temperature is around 45ºc for internet browsing, with a reported ambient temp around 30-33ºc... thing is that the ambient sensor is wrong: it's not really that hot here!!! Two other thermometers report 21-22ºc... perhaps the sensor is wrong, or perhaps the fact that I have my MacPro under my desk has to do with the difference (I do feel some of the GPU's hot air there... where's the ambient temp sensor anyway???).

See attached stats.

CPU temperature rose 13º celsius during the test; power consumption rose 12-15 w. Extra activity from the fans did not kick in. After the test I closed all players and streamings. Temperature stayed at 54ºc (dropped 2ºc) , CPU workload back to 2% User, 3% system, and voltages decreased by 10-15 Watts.

Hope this helps. Cheers.
 

Attachments

  • Stats1.jpeg
    Stats1.jpeg
    63.5 KB · Views: 338
  • Stats2.jpeg
    Stats2.jpeg
    40.6 KB · Views: 270
Thx for showing this! Your basic temps are MUCH HIGHER than full LOAD temps on core i7 CPU's.

I wonder how the new iMac with this CPU will handle the heat?

1251726511aHky7sQBDR_5_1.png
 
handheldgames said:
Thx for showing this! Your basic temps are MUCH HIGHER than full LOAD temps on core i7 CPU's.
Click to expand...

Are you referring to my stats?

I think so too... 45-50 degrees celsius is too much in comparison with what others have reported in same ambient conditions, with NO load at all.

Does this warrant a call to apple and/or a trip to a certified service provider so they diagnose it?
 
Yeah.. Those temps are high.. Kyle's work at [H]ard OCP shows it.

Ive also noticed that some FLASH in a browser can also make the CPU heat on the 2009 start to quickly creek up into to the upper 50s...

What gives? Is the cooler design that bad? Is something that broken in OSX?

The 2008 I'm working on right now slowly increases to 41 degrees with heavy youtube/flash video usage.. whereas the 2009 skyrockets.

Guess I need to find someone with a Hackintosh running with a i7 CPU and see if it happens there.....
 
You got me curious so I decided to check too. 2009 2.66 Quad.

I have had VisualHub doing a re-encode of a video for a few hours, so I started out around 37C.

Added an MP3 in iTunes and watched it climb fairly quickly to 57C. Added a youtube video on top of that and it hit 61C where it has held.

I'm now remembering why I try and stay away from forums about products I own - it just makes me aware of problems I never would have known I had! :D


Edit: I closed iTunes and the temp dropped down to 47C within about 60 seconds.
 
Just got up to work on some email. Just noticed I'm idling at 60(c)!?! I'm not doing anything. Why is my CPU chugging down so much power? I have word, mail, and safari running. I'm just typing an email...

No audio is playing on the system.

WTF... It turns out that some web pages sitting on inactive tabs are the cause of the problem.... No video... No audio.. Just a web page with some flashing graphics.
 
handheldgames said:
Just got up to work on some email. Just noticed I'm idling at 60(c)!?! I'm not doing anything. Why is my CPU chugging down so much power? I have word, mail, and safari running. I'm just typing an email...

No audio is playing on the system.

WTF... It turns out that some web pages sitting on inactive tabs are the cause of the problem.... No video... No audio.. Just a web page with some flashing graphics.
Click to expand...

OS X hates Flash or Adobe Flash hates OS X. Either way Flash taxes any Apple Computer a bit*.

*Unless of course it is running Windows.
 
handheldgames said:
I wish that was the truth. My 2008 jumps 1 to 2(c) on flash. While the 2009 jumps 30(c). WTF?
Click to expand...

I see people reporting temps like that on the 2009 Mac Pro's often enough to think that the factory is (yet again) getting the thermal paste wrong. Or at least inefficiently applied and maybe using a cheap (ineffective) grade too.

I've seen pics where someone removed their heatsinks and the paste appeared to have been applied by injection or flow mold (or maybe prestamped on a press and peal sheet). Anyway it was 5 or 6 times the thickness I use and see being used in pro-technician-tip-videos. It might have been the lighting in the photo but it looked 2.5 mm ~ 3 mm thick, perfect perimeter-inset geometry, and texture patterned on the exposed surface like it came off the CPU cleanly but still had a pattern in it.

If it were me and smcFanControl wasn't helping without ridiculous fan speeds I'd look into redoing their thermal paste job.
 
Tesselator said:
I see people reporting temps like that on the 2009 Mac Pro's often enough to think that the factory is (yet again) getting the thermal paste wrong. Or at least inefficiently applied and maybe using a cheap (ineffective) grade too.

I've seen pics where someone removed their heatsinks and the paste appeared to have been applied by injection or flow mold (or maybe prestamped on a press and peal sheet). Anyway it was 5 or 6 times the thickness I use and see being used in pro-technician-tip-videos. It might have been the lighting in the photo but it looked 2.5 mm ~ 3 mm thick, perfect perimeter-inset geometry, and texture patterned on the exposed surface like it came off the CPU cleanly but still had a pattern in it.

If it were me and smcFanControl wasn't helping without ridiculous fan speeds I'd look into redoing their thermal paste job.
Click to expand...
Assuming this is the case then, the temps would be high for everything users would attempt to run.

This seems far more specific, which is why I keep thinking it's a software issue.

Though if possible, it wouldn't be a bad idea to pull the heatsinks and re-paste if possible (skills/comfortable doing it + paste of course). ;)
 
nanofrog said:
Assuming this is the case then, the temps would be high for everything users would attempt to run.

This seems far more specific, which is why I keep thinking it's a software issue.

Though if possible, it wouldn't be a bad idea to pull the heatsinks and re-paste if possible (skills/comfortable doing it + paste of course). ;)
Click to expand...

Any chance we can get someone running boot camp to test this? I have a dedicated PC and don't have a Windows install to check this with. It would eliminate a hardware issue and point the problem directly at OS X. It not, it would expose a hardware design issue.
 
handheldgames said:
Any chance we can get someone running boot camp to test this? I have a dedicated PC and don't have a Windows install to check this with. It would eliminate a hardware issue and point the problem directly at OS X. It not, it would expose a hardware design issue.
Click to expand...

My copy of Win7 will be here sometime next week. If someone hasn't checked by then I'll look into it.

PM me next week if you still need someone to do it as I might forget.
 
nanofrog said:
Assuming this is the case then, the temps would be high for everything users would attempt to run.

This seems far more specific, which is why I keep thinking it's a software issue.

Though if possible, it wouldn't be a bad idea to pull the heatsinks and re-paste if possible (skills/comfortable doing it + paste of course). ;)
Click to expand...

Oh, it's more specific? I didn't read that part.. Too many pages. :D

Yeah, I guess there's something to this. I've seen a few posts where apps (like iTunes or etc.) on earlier systems are tame but on the 2009 boxes they eat CPU cycles like mad.
 
Tesselator said:
...but on the 2009 boxes they eat CPU cycles like mad.
Click to expand...
This is the impression I'm left with. The code's not efficient on the Nehalem parts, so it's using more cycles for the same application than a different machine.

Additionally, the coolers are working well enough at idle.

handheldgames said:
It's not just audio... Here is a clip of web video cooking the CPU to 65(c).

http://www.youtube.com/watch?v=AUvkbm7Vvlo&feature=youtube_gdata
Click to expand...
If at all possible, the creation of a list (comparitively speaking), of what features/applications are running hotter on Nehalem parts than the other MP models out there (CPU families Apple's used).

Also, comparing it to an equivalent software application under Windows (native, not VM) would help too, as it's the same cooler. Too much variation with PC's and their coolers, layouts,... to get an accurate gauge on the Windows side IMO.
 
here's another one. Having the Web Page sitting idle brought me up to 60(c)...

There is no activity on screen. No video - or anything..... Just search results from YouTube....
 

Attachments

  • another_web.png
    another_web.png
    112.4 KB · Views: 267
nanofrog said:
This is the impression I'm left with. The code's not efficient on the Nehalem parts, so it's using more cycles for the same application than a different machine.

Additionally, the coolers are working well enough at idle.


If at all possible, the creation of a list (comparitively speaking), of what features/applications are running hotter on Nehalem parts than the other MP models out there (CPU families Apple's used).

Also, comparing it to an equivalent software application under Windows (native, not VM) would help too, as it's the same cooler. Too much variation with PC's and their coolers, layouts,... to get an accurate gauge on the Windows side IMO.
Click to expand...

I don't think this is a CPU cycles issue. On my 2009 Quad I see a 10-15C jump when playing music on iTunes yet I see only a 1% CPU utilization. This cannot be a CPU cycles issue, since 1% utilization should not cause a 10-15C rise in temps. Something else is going on here, but what?
 
Well, you can't get CPU heat without CPU activity - all other things equal. I mean I don't thing streaming video is going to increase voltage to the CPU or lower the fan speeds. So that just leaves "cycles".

I dunno, just trying to look at it logically. :p
 
AZREOSpecialist said:
I don't think this is a CPU cycles issue. On my 2009 Quad I see a 10-15C jump when playing music on iTunes yet I see only a 1% CPU utilization. This cannot be a CPU cycles issue, since 1% utilization should not cause a 10-15C rise in temps. Something else is going on here, but what?
Click to expand...

Tesselator said:
Well, you can't get CPU heat without CPU activity - all other things equal. I mean I don't thing streaming video is going to increase voltage to the CPU or lower the fan speeds. So that just leaves "cycles".

I dunno, just trying to look at it logically. :p
Click to expand...
I'm thinking the issues lie in the branch prediction aspect of the processor. It's not done by the actual cores, but by the controller, as it has to do it's "thing" before the data and instructions are sent to them. IIRC, it doesn't report as part of the core utilization %, nor does the IMC or even QPI channels.

If the code isn't optimized (even if it was written for Nehalem, but not quite right), it could be running into dead-ends, loops,... in the controller. Thus it's utilization is high (not the cores), but no where near efficient, generating heat.

I'm not 100% positive, and that's why I was hoping a list of applications/features could be created, then attempt to replicate it in a native installation of Windows on the same cooler (takes out variances due to the case and cooler used). It should be more conclusive as information is collected to determine if sofware is the causality or not.
 
Yeah, one of those would be a good bet. ;) Especially is they don't register as CPU activity. Sounds like we need to rewrite the activity monitors as well!
 
AZREOSpecialist said:
There's also a lot more headroom on the W5500 series used in the 8-core. All W5500 series used by Apple have 95W TDP and 2/2/3/3 turbo boost speed bins vs. 130W TDP and 1/1/1/2 turbo boost speed bins on the W3500 series used in the Quads. This means the 8-cores will have lower temps and higher turbo boost speeds than the Quad, which will run hotter.

The W5580 and W5590 are the two exceptions with 130W TDP. I can't find any information on the speed bins available for turbo boost on those steppings.
Click to expand...

He may be on to something here. Your extreme i7 Processors might not be as friendly as the original Xeons in terms of Turbo boost utilization. The scheduler will significantly speed up your few used cores if it can fit all of the work within them. ITunes sounds like a strong candidate for this.
 
Tesselator said:
Yeah, one of those would be a good bet. ;) Especially is they don't register as CPU activity. Sounds like we need to rewrite the activity monitors as well!
Click to expand...
It's a little bit of work, but I would have thought that there were enough interested users with a native installation of Windows (preferably a newer version, such as Vista or Win7) to compare the same tasks with OS X.

Obviously not... :eek: :p
 
Well... I picked up a copy of xp from the IT room I was boxing up today. Guess I should have grabbed vista? Changed out a HD and I'm installing a boot camp xp installation.

I never boot camped as Ive always had an abundance of wintel hardware laying around.

Well I installed and wound up with a disk error. Fun.
 
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