Audio Decoding KILLS MacPro (2009) green factor

[AZREOSpecialist]

Unless a magical update happens, aka OSX gets fixed... WAKEUP APPLE!!! , The 2009 MP is turning out as the LEMON step child of the product line.

I know you were probably being dramatic, but you can't honestly believe that the current generation Mac Pro is a lemon by any means. While it seems mysterious as to why the temperature is rising more than expected, it doesn't affect the Mac Pro in any discernible way nor does it impact the computer's speed, your workflow, or what you can get out of the machine.

I certainly didn't buy the Mac Pro for it's "green" features, I bought it because it's a powerhouse and I expect it to throw out more hot air than your typical iMac or Mac Mini. There's something in OSX causing the temperature rise, but it will eventually be fixed. I haven't been back to this thread in a while, I'm curious about something... do we know that all of these programs, i.e. iStat Menus, etc., are reading the temperatures properly? I've heard discussions elsewhere that utilities are not properly reading the temperature sensors in the new 2009 Pros. Could that be a possibility?

 

[goodcow]

I certainly didn't buy the Mac Pro for it's "green" features, I bought it because it's a powerhouse and I expect it to throw out more hot air than your typical iMac or Mac Mini. There's something in OSX causing the temperature rise, but it will eventually be fixed. I haven't been back to this thread in a while, I'm curious about something... do we know that all of these programs, i.e. iStat Menus, etc., are reading the temperatures properly? I've heard discussions elsewhere that utilities are not properly reading the temperature sensors in the new 2009 Pros. Could that be a possibility?

I know this isn't exactly scientific, but at work we have a bunch of 3,1 octo 3.2's. My MacPro at home is a 4,1 octo 2.26.

The level of heat output near the PSU is noticeably much, much hotter on my 4,1 than the work 3,1's.

And yes, this does matter to me. Not only is it not "green," but it's sucking down more wattage, which is costing me more money, and it turns my small studio apartment into an oven, meaning I have to use the AC in months I really shouldn't have to, costing me even more money.

 

[dzam]

I know you were probably being dramatic, but you can't honestly believe that the current generation Mac Pro is a lemon by any means. While it seems mysterious as to why the temperature is rising more than expected, it doesn't affect the Mac Pro in any discernible way nor does it impact the computer's speed, your workflow, or what you can get out of the machine.

Yeah, but if your temperatures waves +/-30°C many times in day, it isn't good for components. It also speeds up fans, so there's much more noise without any reason.

Apple really should do something for this. If it doesn't, at least I'll require a new model of Mac Pro, when it arrives. This isn't normal.

 

[smacman]

Looks like this issue is starting to receive a bit more attention.

http://www.xlr8yourmac.com/

 

[AZREOSpecialist]

So... are we all pretty much in agreement that this is a Turbo Boost issue? This should be pretty easy to verify with the new quad core iMac.

 

[nanofrog]

So... are we all pretty much in agreement that this is a Turbo Boost issue? This should be pretty easy to verify with the new quad core iMac.

No. It's just with certain applications (audio/visual), not all instances were TB is engaged (i.e. no reports of it happening with things like calculator, word processing,... <low CPU% single threaded applications that exclude the underlying audio/video code in OS X>). Nor does it happen under Windows at all it seems, as there's been no reports of similar behavior on that side, and testing by MR members hasn't turned up anything either IIRC.

 

[gugucom]

IMO its directly connected to the Quicktime codec. The codec is the common denominator in all media apps that brings the bug out. Another factor may be a software fault in the power gating and turbo boosting software.

 

[nanofrog]

IMO its directly connected to the Quicktime codec. The codec is the common denominator in all media apps that brings the bug out. Another factor may be a software fault in the power gating and turbo boosting software.

This is how I see it as well.

Quicktime X, and Flash (not sure if it actually utilizes the Quicktime X code, but presume it does). Makes sense to me anyways.

 

[DoFoT9]

So... are we all pretty much in agreement that this is a Turbo Boost issue? This should be pretty easy to verify with the new quad core iMac.

hmm i think its an issue with turbo boost being activated or something of the like.

i know have my i7 imac i have been running audio in iTunes over the network (no audio on computer yet lol) for a good 10minutes and counting. i have also been on MR replying to messages etc. CPU temp is at 37°C/39°C with fan speeds all on stock basically.

soon i will copy some songs (lossless audio FWIW) across and try the test again.

for testing purposes, could any of your MP owners with the problems possibly try playing audio over the network and see if you still have the problem?

 

[nanofrog]

hmm i think its an issue with turbo boost being activated or something of the like.

i know have my i7 imac i have been running audio in iTunes over the network (no audio on computer yet lol) for a good 10minutes and counting. i have also been on MR replying to messages etc. CPU temp is at 37°C/39°C with fan speeds all on stock basically.

soon i will copy some songs (lossless audio FWIW) across and try the test again.

for testing purposes, could any of your MP owners with the problems possibly try playing audio over the network and see if you still have the problem?

It's not related to Turbo Boost or Hyper Threading. If it were, we'd be seeing differences with the new iMacs (i5 & i7 models with different behavior from each other (HT), and the MP's (same if it were TB)).

Architecturally, DMI separates them from the MP's (LGA1156 vs. LGA1366). But if it were the CPU's, the problems would be under Windows as well.

So it's definitely in the software (OS X).

 

[DoFoT9]

It's not related to Turbo Boost or Hyper Threading. If it were, we'd be seeing differences with the new iMacs (i5 & i7 models with different behavior from each other (HT), and the MP's (same if it were TB)).

Architecturally, DMI separates them from the MP's (LGA1156 vs. LGA1366). But if it were the CPU's, the problems would be under Windows as well.

So it's definitely in the software (OS X).

so far the problem hasnt appeared on either i7 or i5 imac AFAIK. do you know anybody that has the problem with an imac?

so if it is in the software, what components are different from a MP to an iMac?

 

[nanofrog]

so far the problem hasnt appeared on either i7 or i5 imac AFAIK. do you know anybody that has the problem with an imac?

I've not seen anything.

so if it is in the software, what components are different from a MP to an iMac?

The difference would come down to the QPI most likely, as OS X isn't optimized for the Nehalems (compiler actually). The iMac CPU's are working well, as they're DMI to the chipset, which was used in all the previous Intel parts (FSB is gone, but DMI remains in the LGA1156 parts).

 

[DoFoT9]

I've not seen anything.

thats a good thing

The difference would come down to the QPI most likely, as OS X isn't optimized for the Nehalems (compiler actually). The iMac CPU's are working well, as they're DMI to the chipset, which was used in all the previous Intel parts (FSB is gone, but DMI remains in the LGA1156 parts).

that would make sense then. the question is what could possibly be happening to the QPI to make the CPU get so hot? are there multitudes of messages being sent because its not optimised? or would it be something else?

 

[nanofrog]

thats a good thing

Sort of. I've not been looking that hard lately (not spotted a thread similar to this one in the iMac section). So I could easily have missed something the size of a proverbial Mack Truck.

that would make sense then. the question is what could possibly be happening to the QPI to make the CPU get so hot? are there multitudes of messages being sent because its not optimised? or would it be something else?

I can't help but think Unterminated Loops are the most likely culprit.

 

[DoFoT9]

Sort of. I've not been looking that hard lately (not spotted a thread similar to this one in the iMac section). So I could easily have missed something the size of a proverbial Mack Truck.

ive been looking nothing at all

CPU is at 47°C now. not sure why. fans are at same speed, CPU usage is still around 2%.

edit; now 48°C
edit edit: nm i forgot that i opened parallels

I can't help but think Unterminated Loops are the most likely culprit.

hmm would make sense. they work in the same manner as over the oldschool bus networks if there is no terminating device at each end i take it?

 

[nanofrog]

hmm would make sense. they work in the same manner as over the oldschool bus networks if there is no terminating device at each end i take it?

No.

Granted, the methodology used in these CPU's is simplified, it's got multiple layers to it <skips the routing layer> (QPI wiki).

I'm thinking more of UMA vs. NUMA is the causality within it. Assuming the compiler wasn't optimized for the LGA1366 chips, then this could be were we'd be getting into the nuts-and-bolts of it.

 

[DoFoT9]

No.

Granted, the methodology used in these CPU's is simplified, it's got multiple layers to it <skips the routing layer> (QPI wiki).

I'm thinking more of UMA vs. NUMA is the causality within it. Assuming the compiler wasn't optimized for the LGA1366 chips, then this could be were we'd be getting into the nuts-and-bolts of it.

now that is very cool stuff!!

interesting::

As of 9th September 2009, Intel altered their implementation on mainstream versions of the Core i5/i7 chips (notably the i5 750, i7 860 and 870) to exclude a QPI, replacing....

but, the i5s/i7s still have a QPI dont they? it says so in System Profiler

 

[nanofrog]

but, the i5s/i7s still have a QPI dont they? it says so in System Profiler

No.

LGA1156 (iMac models)
i5-750
i7-860

LGA1366 (MP's) <all the Xeon Nehalem and i7-9xx parts>
W3520
E5520

You'll notice DMI for the LGA1156 parts, and QPI for the LGA1366 parts. The system profiler is wrong (not distinguishing the DMI based Nehalem from QPI Nehalem P/N's, and labeling all Nehalems as QPI I'd guess).

 

[gugucom]

My W5590 has exactly the same behavior. So it isn't isolated to the models that Apple used. But the errata of the chips are all the same typically for one design. It could be a chip erratum that works in conjunction with certain software functions. Perhaps Windows is aware of this or it is a fault which requires a certain use case which only appears in OS X apps. If it is related to a 5500 erratum it will probably not repeat itself in another chip.

 

[smacman]

Has anyone else noticed the following entry in the console log after a fresh boot?

WARNING: ACPI_SMC_CtrlLoop::initCPUCtrlLoop - turbo enabled but no turbo P-state found

I have it 8 times in a row, so I assume that dual cpu models will have 16 instances of this message. I wonder if it is somehow related to our problem...

 

[nanofrog]

My W5590 has exactly the same behavior. So it isn't isolated to the models that Apple used. But the errata of the chips are all the same typically for one design. It could be a chip erratum that works in conjunction with certain software functions. Perhaps Windows is aware of this or it is a fault which requires a certain use case which only appears in OS X apps. If it is related to a 5500 erratum it will probably not repeat itself in another chip.

I only used a couple of parts to illustrate the differences, but I'd expect it not only on the Xeon parts, but all the LGA1366, as they're developed from the same architecture, and in even the same errata in the newer parts (D0, as the first parts shipped are C0 in the i7-9xx line).

I can't help but think Apple's compilers weren't optimized for Nehalem before completing 10.6's master, and still hasn't been addressed.

Intel worked with MS to optimize Win7 for use on their chips (source). So I'm interpreting this as inclusive of Nehalem architecture optimizations into the compilers and resulting OS.

 

[smacman]

For anyone else that was using iStat to figure out CPU temp during audio playback, iStat's CPU reading is actually from the CPU heat sink sensor. After installing "Temperature Monitor" I discovered that the CPU is running even hotter than I thought. After 5 mins of playing audio, all 4 cores were reporting temps of 87-89C with the CPU heat sink in the 60s... Ouch!

To make matters worse, it's not just a thermal issue. I decided to test available CPU load by using XBench. Doing a XBench CPU and Thread test (uncheck all other boxes), the score is 20% lower if tested while playing audio. The same test on my 17" Macbook Pro yields the exact same result whether playing audio or not.

So basically, for all of us that tend to listen to music while we are working, you are not only cooking your CPU, but also losing aprox. 20% of your CPU capability.

 

[Macinposh]

To make matters worse, it's not just a thermal issue. I decided to test available CPU load by using XBench. Doing a XBench CPU and Thread test (uncheck all other boxes), the score is 20% lower if tested while playing audio. The same test on my 17" Macbook Pro yields the exact same result whether playing audio or not.

So basically, for all of us that tend to listen to music while we are working, you are not only cooking your CPU, but also losing aprox. 20% of your CPU capability.

Auch! Loosing 20% of the performance when listenig to music... Windowsboys are going to have slugfest on this one...

 

[nanofrog]

So basically, for all of us that tend to listen to music while we are working, you are not only cooking your CPU, but also losing aprox. 20% of your CPU capability.

OUCH.

Auch! Loosing 20% of the performance when listenig to music... Windowsboys are going to have slugfest on this one...

Muhahaha...

But it's not like MS's OS development has ever been devoid of bugs either...

 

[DoFoT9]

oh my! This is terrible and pathetic.