Audio Decoding KILLS MacPro (2009) green factor

[Techhie]

It seems to me that there are two related problems regarding heat

1. The increased power draw, increased core temperatures, and decreased performance when running audio & video apps; and

2. The seemingly defective fan control.

Has ANYONE heard/seen their fan speeds increase on a 2009 MacPro as they increase processor load? No matter what I'm doing, I NEVER find my MacPro ramping up fan speed. This suggests to me defective fan control software. (I've had to use SMCFanControl to speed up my fans.)

So there may be two defective kexts, one for power control, one for fan control.

No, it's not any malfunctioning hardware. For some reason Apple decided to spike the fan threshold to a few degrees short of CPU failure in the '09 models, and that limit cannot be achieved by anything but a constant 800%(1600% octad) load in a very hot room due to the type of heatsinks and thermal compound being used.

 

[tonydickinson]

No, it's not any malfunctioning hardware. For some reason Apple decided to spike the fan threshold to a few degrees short of CPU failure in the '09 models, and that limit cannot be achieved by anything but a constant 800%(1600% octad) load in a very hot room due to the type of heatsinks and thermal compound being used.

This may be true, but I own a 2009 MacPro 2xquad Core Nahalem purchased in April 2009.

I live in Singapore where the ambient temperature is not usually less than 31 degrees and I notice that even playing iTunes the temperatures of the processors are in excess of 50 degrees, strangley enough there is usually a 5 degree difference in the temp of the two processors.

I will raise this again with Apple as it does seem to be totally excessive compared to a MacMini and an iMac. I am disappointed in this MacPro.

Suffice to say I also have a i7 hack which is far more efficient than the MacPro and the CPU cores run 10 degrees cooler!

 

[eemzah]

This may be true, but I own a 2009 MacPro 2xquad Core Nahalem purchased in April 2009.

I live in Singapore where the ambient temperature is not usually less than 31 degrees and I notice that even playing iTunes the temperatures of the processors are in excess of 50 degrees, strangley enough there is usually a 5 degree difference in the temp of the two processors.

I will raise this again with Apple as it does seem to be totally excessive compared to a MacMini and an iMac. I am disappointed in this MacPro.

Suffice to say I also have a i7 hack which is far more efficient than the MacPro and the CPU cores run 10 degrees cooler!

Don't bother wasting your time at Apple Ang Mo Kio, eServ or basically any Apple authorised repair centres. They know nuts about this issue. I probably saved you all the trouble of lugging your machine down and getting a 'everything is ok' reply. Save yourself the pain and anger listening to daft tech guys telling you to 'reinstall', reset SMC, don't listen to music when working blah blah blah...

Cheers.

 

[cl-user]

[...] strangley enough there is usually a 5 degree difference in the temp of the two processors.

The exhaust heat from the front CPU (A) cooler partially feeds into the heat sink/cooler of the rearmost CPU. Hence the rear tends to get hotter.

Also I find that sound apps seem to fry the rear CPU, whilst Cinebench's single CPU render test seems to work the front CPU. Puzzling that CPU allocation isn't random.

If the bug is not fixable, what's the betting that an OS update moves the audio bug to the first CPU - it is easiest to observe the bug presence through CPU-A temperature.

 

[cl-user]

Has ANYONE heard/seen their fan speeds increase on a 2009 MacPro as they increase processor load?

A couple of times, under both audio bug and saturation load (1600%), with no SMCfancontrol (hence vanilla setup), after a while I have seen a slight increase in fan speed. Fan response is very delayed, and Temperature Monitor (TM) CPU-A Temperature Diode readings were getting up into the 90C range (which must put the core temps over 100C).

I only did this after my CPU repair, as a stress test. The feeling that one is irrevocably destroying something grew too strong for me (I'm a G5 Quad overheating victim too).

How I'll be kicking myself if we find that TM has been wildly exaggerating.

 

[smacman]

A couple of times, under both audio bug and saturation load (1600%), with no SMCfancontrol (hence vanilla setup), after a while I have seen a slight increase in fan speed. Far response is very delayed, and Temperature Monitor (TM) CPU-A Temperature Diode readings were getting up into the 90C range (which must put the core temps over 100C).

I only did this after my CPU repair, as a stress test. The feeling that one is irrevocably destroying something grew too strong for me.

How I'll be kicking myself if we find that TM has been exaggerating.

It's been a while since I looked at them, but Intel puts out fairly detailed guidelines that should be followed in the system design phase. One of the things they talk about is CPU fan speed increasing linearly as the CPU draws more power. The scary thing is, they don't..

 

[cl-user]

Is this happening in both 32 and 64 bit kernel modes under SL?

Yes.
I find Cinebench results similar.
Single CPU render results were identical (with and without audio) - that is -26.6% (on 2.66x2x4).

 

[cl-user]

Same here.

It's a software flaw that's the culprit, as it's not occuring when running Windows on the same machine. That's FIXABLE.

Well is it? Could there be some reason that we haven't thought of yet. Yes on the face of it, anything in software can be fixed. But history shows many instances of bugs that couldn't be fixed.

Or the fix could be prohibitively expensive - or was seen as such before we started to make a noise about it.

Could this low level OS code be highly specialist. Perhaps key people have left Apple or Intel (they must be itching to tell us the scoop on all this, or are having a monster laugh).

Could there be some issue due to Apple's thrust to share a common core of OS code between such widely different systems as the Mac Pro and iPhone.

Could there be some devilish dependency from the code areas that cause this bug and other areas which are thought even more critical (eg. instability or performance).

I'm looking for excuses for Apple.

Its intriguing to see Apple's only public effort on this bug seems to be to avoid the issue, try to shove it under the carpet. There must be a reason for this. And it doesn't seem likely that Apple are so incompetent as not to have realized that this bug existed for at least half a year.

If Toyota treated their accelerator pedal bug like this - blanket cover up and no apparent attempt to fix ...

But what happens when an unstoppable missile impacts on an un-breachable wall? One was bluffing it (meaning Apple v customers, in the end all the lawyers and spin in the world can't change some things).

Meanwhile trust, brand image, long term consequences, ... will these customers roll over and forget. Maybe if we each get a bug-free 2010 replacement and apology. Somehow I can't see Apple embracing its customers so warmly. So we'll probably go forward with a sense that Apple might treat us like this on other issues in future, the next time might be more incapacitating.

 

[ncc1701d]

Download hardware monitor from
Bresink.de
Download Cinebench from
Cinebench for the Mac from apple servers
Run Hardware monitor and put close attention to Heat Sensor Diode of cpua and cpub in case you have an 8 core model and ambient temp
then let your mac pro idle for a few for the temps to stabilize
then fire up cinebench anb run the full set of tests take note of the temps
open iTunes and and play music for about 15 minutes. keep an eye on temps.
Without closing itunes or stoping the music run again the tests of cinebench
Look your performance drop....

Thank beto2k7

After each cinebench test, I just save to "clipboard" and paste it into pages to see the differences?

 

[ncc1701d]

I'm going to take the time and go through the facebook link. in the mean time, this is what I got from the cinebench. It's just a very rough one (still with USB things attached etc. First one is before music, second one is after 15 minutes of playing. Is this right?

I haven't changed ANYTHING (deleted names or whatever), this is exactly as it came after sending it to the clip board and then pasting to pages.


CINEBENCH R10
****************************************************

Tester :

Processor :
MHz :
Number of CPUs : 8
Operating System : OS X 32 BIT 10.6.2

Graphics Card : ATI Radeon HD 4870 OpenGL Engine
Resolution : <fill this out>
Color Depth : <fill this out>

****************************************************

Rendering (Single CPU): 3893 CB-CPU
Rendering (Multiple CPU): 15438 CB-CPU

Multiprocessor Speedup: 3.97

Shading (OpenGL Standard) : 6804 CB-GFX

****************************************************



CINEBENCH R10
****************************************************

Tester :

Processor :
MHz :
Number of CPUs : 8
Operating System : OS X 32 BIT 10.6.2

Graphics Card : ATI Radeon HD 4870 OpenGL Engine
Resolution : <fill this out>
Color Depth : <fill this out>

****************************************************

Rendering (Single CPU): 2875 CB-CPU
Rendering (Multiple CPU): 14505 CB-CPU

Multiprocessor Speedup: 5.05

Shading (OpenGL Standard) : 5432 CB-GFX

****************************************************

 

[Wyremunc]

Recently RMA'd a refurbished 2.93GHz MP (6GB RAM) for memory issues. Strangely enough my computer would only kernel panic while watching video or playing iTunes. I stumbled upon this thread, and can't help but wonder if this over-temp issue was actually the root of my problems too. However, all my error reports centered around one of the three memory modules. The "geniuses" at the Apple Store replaced all three modules, and the problem went away... for about two days. Is it possible memory (or the memory controller) is involved here as well?

Anyway, I was able to get an RMA (to my astonishment Apple was reluctant to take the thing back) because I claimed the computer was DOA, which may help those who have recently purchased.

For everyone else... perhaps the best way to get Apple's attention (and a bunch of other folks too) would be to claim this issue a virus!

BTW: The refurbished site has a 2.93 Quad w/3GB RAM now. If that's my old one, don't buy it!

 

[nanofrog]

Well is it? Could there be some reason that we haven't thought of yet. Yes on the face of it, anything in software can be fixed. But history shows many instances of bugs that couldn't be fixed.

Or the fix could be prohibitively expensive - or was seen as such before we started to make a noise about it.

This is what can't usually means. Unwilling to commit the necessary resources to solve the problem based on cost. There are cases where it's a valid situation, but I don't see OS X as such, no matter what machine it's run on (Apple labeled system that they did make a profit on).

In this case however, it doesn't appear the depth of the issue is that deep. Before the new iMacs dropped (LGA1156 based units), I conceded it could require a compiler re-write. But as those systems aren't having the issues, it doesn't seem to be the case, as the CPU uses the same architecture. The chipset is different, and there's no ICH chip at all, as those functions are on the P55 chipset used with that series of parts (socket).

Could there be some issue due to Apple's thrust to share a common core of OS code between such widely different systems as the Mac Pro and iPhone.

This issue is handled by the compiler/s used.

Could there be some devilish dependency from the code areas that cause this bug and other areas which are thought even more critical (eg. instability or performance).

It's possible, but I do see it as technically possible to solve. But Apple may not be willing to invest the resources, no matter how inexpensive.

If that ends up the case, I hope they get burnt, as it would be well deserved for such poor treatment of customers.

 

[Soulstorm]

This is not the first time a similar thing happens...

Wow. My father wanted to buy a Mac Pro for audio processing (he is a musician) and seeing this thread actually made him not to buy it.

I only hope that Apple's reluctancy to acknowledge the problem is just a means to silence the problem until a real fix comes up (which they should have been working on as we speak).

However, I should say my experience with the G5 (remember those beasts?):

The top line from Apple is totally untrustworthy. Before the Intel Macs came out, I had a dual G5 at 2.5 GHz, I live in Greece, and there's no Applecare here. One year and one month after purchase, my computer was fried because the liquid cooling system had a leak, and that destroyed the processor of the machine. I had to give 1000 Euros to fix it. Apple reseller here in Greece said that I had to pay the price. And Apple in the US didn't want to hear about it at all.

However, I stumbled across the internet in many sites that said the same thing. I give you 2 links of the hundreds:

http://www.macintouch.com/readerreports/powermacg5/topic4243.html
http://www.macsimumnews.com/index.php/archive/macosgliquid_cooled_power_mac_g5s_leaking
http://www.everymac.com/systems/app...rmac-g5-liquid-cooling-info-leaks-issues.html

So, to sum up:

This is the first time I say I won't be buying any high-end machine from Apple. Especially the ones that come in into the grey aluminum tower, like the Mac Pros. Apple has long proved to have problems with the cooling systems and CPU fans/heat temperatures.

 

[smacman]

Wow. My father wanted to buy a Mac Pro for audio processing (he is a musician) and seeing this thread actually made him not to buy it.

I only hope that Apple's reluctancy to acknowledge the problem is just a means to silence the problem until a real fix comes up (which they should have been working on as we speak).

However, I should say my experience with the G5 (remember those beasts?):

The top line from Apple is totally untrustworthy. Before the Intel Macs came out, I had a dual G5 at 2.5 GHz, I live in Greece, and there's no Applecare here. One year and one month after purchase, my computer was fried because the liquid cooling system had a leak, and that destroyed the processor of the machine. I had to give 1000 Euros to fix it. Apple reseller here in Greece said that I had to pay the price. And Apple in the US didn't want to hear about it at all.

However, I stumbled across the internet in many sites that said the same thing. I give you 2 links of the hundreds:

http://www.macintouch.com/readerreports/powermacg5/topic4243.html
http://www.macsimumnews.com/index.php/archive/macosgliquid_cooled_power_mac_g5s_leaking
http://www.everymac.com/systems/app...rmac-g5-liquid-cooling-info-leaks-issues.html

So, to sum up:

This is the first time I say I won't be buying any high-end machine from Apple. Especially the ones that come in into the grey aluminum tower, like the Mac Pros. Apple has long proved to have problems with the cooling systems and CPU fans/heat temperatures.

I can't say I blame you for feeling that way. Once bitten, twice shy. When you look at how Apple deals with issues like this (or fails to deal with, to be more accurate), it's hard to believe they have any long term customers at all..

 

[TonyK]

...

If Toyota treated their accelerator pedal bug like this - blanket cover up and no apparent attempt to fix ....

But they did. Remember the first fix of removing the floor mats? They explanation was the mats were causing the peddle issues.

 

[flubber]

Recently RMA'd a refurbished 2.93GHz MP (6GB RAM) for memory issues. Strangely enough my computer would only kernel panic while watching video or playing iTunes. I stumbled upon this thread, and can't help but wonder if this over-temp issue was actually the root of my problems too. However, all my error reports centered around one of the three memory modules. The "geniuses" at the Apple Store replaced all three modules, and the problem went away... for about two days. Is it possible memory (or the memory controller) is involved here as well?

Anyway, I was able to get an RMA (to my astonishment Apple was reluctant to take the thing back) because I claimed the computer was DOA, which may help those who have recently purchased.

For everyone else... perhaps the best way to get Apple's attention (and a bunch of other folks too) would be to claim this issue a virus!

BTW: The refurbished site has a 2.93 Quad w/3GB RAM now. If that's my old one, don't buy it!

I also got this problem on my quad core 2,66GHz Nehalem. When encoding or watching a movie, I got thousands of ECC errors on memory bank 3, and eventually a kernel panic. They first replaced the memory module, which didn't resolve the issue. Then they replaced my processor board. This also didn't resolve anything, so they replaced my CPU (which has an integrated memory controller). I've been free of ECC errors for 4 months now.

I also suggested that the processor might have gone defective because the fans never seemed to speed up. Some Apple engineers discussed this with me. I also showed them the W3520 (nehalem) datasheet, explaining the processor's temperatures. In the end, everything seemed to be fine. The datasheet mentions maximum temperatures for each CPU load (in Watts). For a maximum of 130W (800% CPU load), the processor may get as hot as 67,9°C. If it gets any hotter, fans should speed up. If they don't, you've got a problem. In my case, the fans never speeded up, but the maximum temperature I could reach was 68°C so it wasn't necessary either.

 

[smacman]

In the end, everything seemed to be fine. The datasheet mentions maximum temperatures for each CPU load (in Watts). For a maximum of 130W (800% CPU load), the processor may get as hot as 67,9°C. If it gets any hotter, fans should speed up. If they don't, you've got a problem. In my case, the fans never speeded up, but the maximum temperature I could reach was 68°C so it wasn't necessary either.

Just curious, which temperature reading were you monitoring when you observed temps reaching 68C?

 

[DeepCobalt]

By the way... as Gainestown and Gulftown have the same chipset, and as it is likely this is a software issue, won't any upcoming 2010 Mac Pros have the same problem? If so, this would be a big incentive for Apple to fix this problem pronto.

 

[Inconsequential]

By the way... as Gainestown and Gulftown have the same chipset, and as it is likely this is a software issue, won't any upcoming 2010 Mac Pros have the same problem? If so, this would be a big incentive for Apple to fix this problem pronto.

Yes, it will.

Good eh?

 

[cl-user]

[...]perhaps the best way to get Apple's attention (and a bunch of other folks too) would be to claim this issue a virus!

My vote is that we keep hammering away at truth and clarity, work to nail down as much information as possible so Apple and all new comers can see exactly what is going on.

Try to present things so we cut defensive attitudes off before they gain momentum. And try not to irritate any egos.

Some humor, exquisite wordsmithery, relevant pictures, captivating stories, etc., may help get the message over. But in general less obfuscation, more on message. Merge all constructive contributions. That's the target.

AFAICS the message is: there's a bug, what is it, how to see it, what to do about it (temporary remedies), consequences, and above all try to get through to Apple in hope of them working on a complete fix.

I'm putting together material to that end for a web site (will notify when something running).

 

[beto2k7]

Maybe this is a little extreme but.... what about us organizing a class action lawsuit against apple?

 

[smacman]

Maybe this is a little extreme but.... what about us organizing a class action lawsuit agains apple?

I think it is a great idea, and a real possibility. The challenge would be the lack of participation. There are not that many Mac Pro (Early 2009) users out there, and within this small group, only a small percentage are aware of the issue. On the positive side, even if they fixed it all tomorrow, there is still a reasonable case given the unnecessary heat these machines were all exposed to. Combine that with the fact that performance has been adversely affected, and we could have a chance.

 

[DeepCobalt]

Maybe this is a little extreme but.... what about us organizing a class action lawsuit against apple?

The idea sounds appealing on the surface, but I think the result would not be what most of us want... that's the problem with class action law suits--they just benefit the lawyers.

The law firm would easily get the amount of people necessary to file the suit--anyone who bought a 2009 Mac Pro would be automatically eligible, and signing up would be trivial (these days done online or through a mailing). They would get plenty of folks. The problem is that the damages would take years to sort out, and the lawyers would collect 40%+ of the proceeds resulting in a minimal compensation for each Mac Pro buyer. Sure, it would give Apple a bit of bad press and perhaps even a small financial pinch, but it would in no way get us--the end user--a properly functioning Mac Pro. All we'd get is a $20 check, if we're lucky. Not to mention, Apple would settle this, so no prolonged media drama.

I think we should keep on the track of getting Apple to fix this. If all else fails, we can look at a retaliatory measure, but being aware, we're just making lawyers rich.

 

[flubber]

Just curious, which temperature reading were you monitoring when you observed temps reaching 68C?

I used Hardware Monitor and observed CPU A temperature diode.
The heatsink gets less hot, and the cores reached temperatures of about 85°C.
Here you can find the datasheet. Page 81 includes data on the temperatures.

http://www.intel.com/Assets/en_US/PDF/datasheet/321332.pdf

 

[jlozier]

My Mac Pro also shows a 5W to 55W increase in CPU Wattage, and a reported temp increase from 30˚C to 60˚C when I play music (iTunes and Spotify)- I play musically constantly and never realised I'm baking my processor!