Audio Decoding KILLS MacPro (2009) green factor

[nanofrog]

No... that's not at all what I'm suggesting. It could be a simple matter that there's a bug in the firmware or software used to report temps that's doing bad math and it could be doing any number of bizarre things to the final output.

The fact is, I don't believe anyone's established that the reported temps OR rise is actually real... have they?

Look at the related info:

- Temp's rise 20+ deg. C
- Less than 1% CPU utilization
- No ramp in fan speeds

It seems misguided to assume the temp. increase is actually happening. Until someone gets an IR thermometer on the heatsink to determine the actual temp. increase, the most logical explanation is that there is some bug in the firmware/software.

I actually understood this was part of the implication with your post.

But the heat is real. Stick your hand at the exhaust fan before and after playing an audio file, and you'll notice a significant rise. That shouldn't be happening.

I neglected to mention the PSU draw, performance drop,... involved, but there's more than just iSTAT numbers behind this. So it's not a buggy application reporting something that's not there (phantom data).

Now the issues could be in the firmware

There could also be other strange things at work as well. For example, I once had a PC who's temps shot up just like this then when I plugged in a particular USB Bluetooth dongle. It could be that the BT dongle was somehow mixing up register values that the firmware was using because the CPU certainly wasn't getting hot. I'm sure you've seen equally strange things that turn out to be somewhat different than initial evidence might indicate.

It's possible, but it seems some reporting this are using base systems (or very close to it, such as only a RAM or HDD upgrade).

Different video cards nor RAID seem to be the causality either.

Ok, perhaps I'm harping too strongly on the software being at fault. It's just that there's a lot of assumption making going on in my opinion and I don't think there's enough information to lay this problem at Apple's feet.

If all we had is OS X data, and nothing under Windows, then I'd agree. But the same MP running Windows isn't producing the same issues. That takes it out of Intel (and why I checked to see if there was yet a newer stepping that had released).

It's possible that some systems (more recent) have a newer firmware revision that's not yet been made part of an OS X update (testing on user systems, and it's reporting data back to Apple so they can determine if it's good or not).

I personally don't have this problem. See my screens shots of iStat with both no music playing and music playing. The temps just aren't changing and my Xbench scores are the same whether iTunes is playing music or not. Therefore, this problem must in some way be environmental and related to your configuration.

IIRC, you and Seisend are the only ones that have reported that there's no problem with temps and iTunes.

This is one of the reasons I'm wondering if Apple's released a newer firmware revision in more recent systems that's an attempt to resolve the issue. No announcement or addition of it to an update yet, as it's still in testing (unbeknownst to the users that they're test subjects, and their systems are sending bug reports).

Could it be other hardware in the system that's causing a bizarre quirk? Is it the audio interface people are using? Is there something in common between people that have the problem besides just a Nehalem Xeon? A standard trouble-shooting technique is to pull every non-essential piece of hardware from the system to see if that helps... maybe that's worth a try?

I recall a problem where people with an ATI 4870 in their system were having problems with iTunes crashing when the computer resumed from sleep... that's a good example of the kinds of bizarre interactions that can occur in a computer. Perhaps that's even related to this issue?

It's worth a try, but IIRC, the systems affected aren't exclusively 4870's, and some where base systems. Some posts don't give system details. So I'm not 100% sure of who has what exactly, so pulling items is a good idea.

Those with RAID (especially hardware cards), may be unable or unwilling to pull them though.

Utlimately to do this, we need a group of willing volunteers. Start by listing out the all of the details of the systems (hardware models, firmware and SMC revisions,...), look at what's been posted, and go from there (find comonality, and test those parts). OS installs need to be identical as well. It's likely a fair bit of work, but doable.

Just stating the facts. My system gets really hot when playing audio. It cools off substantially when I stop playing audio. When it is exhibiting this "hot" behavior, the power consumption rises dramatically, and the CPU scores substantially lower when benchmarked.

VirtualRain is apparently an exception, as is Seisend. The rest seem to be having the same results as you.

I'm really curious to find out what's different about their systems, as it could hold the key to figuring this out. And forwarding the results to Apple could help get a solution out much faster. We shouldn't have to do this, but it's worth it IMO to get a solution.

Your starting to sound like you work at Apple.

I dont' think this is the case, and he's been a solid contributor here on MR. Rather helpful, and quite knowledgable as well.

No need to get upset. He's admitted being "a little late to the party", and is trying to catch up as well as interject some objective opinions. Remember, it's easy to get so caught up in such issues, to lose sight of something simple.

BTW - Your iStat CPU reading of 55C is measured on the heat sink. In other words, your heat sink is approximately 55C. Add 5-10C to that for your CPU temp. If you want a much better temperature reading, use TEMPERATURE MONITOR which reports much more extensively. It sounds to me like not only are you suffering from the same problem, in your case, it does not cool off even when you stop playing audio. That would explain why your benchmark stays the same too. The question is, what else do you have running that may be keeping your CPU in this strange state? Make sure all your browsers are closed, and anything else that has anything to do with audio / flash rendering etc. is closed. 55C on the heat sink is way too hot for idling. My heat sink temp at idle is 32C. People with these chips overclocked to 4.0Ghz+ on Windows machines report those type of heat sink temps under heavy load.

I need to take a closer look at VR's screen shot.

I can confirm the OC temps. I didn't hit the upper 70'sC until I pushed it to over 4GHz (4.12 & 4.2GHz on air = stable, but hot). It's back on stock ATM. I'll likely push it back to 3.8GHz, and leave it alone (decent OC, and the temps were fine IIRC <need to go back and check my notes when I get around to it>).

EDIT:
I went back and took a closer look at the temps. Except for the CPU temp, the rest looks good. A couple of degrees rise in PSU and memory temps (rather decent overall here), yet the CPU stays the same (55C). But the CPU is quite warm for idle and 1% activity.

VR: When was it shut down last?

 

[handheldgames]

VirtualRain.....

Your temps are pegged at 55(c) which is high for idle or playing music. My Quad 2009 MacPro, idles at 32(c) As does my 2008 Octa @ 31(c) and my Quad Wolfdale Q6600 at 31(c) The only one that jumps up in temps is the 2009 MacPro.

Are you saying, after a reboot that your idle temp is normally at 55(c)? How long does it take to rise and level off? Its hard understand that there is no effect in playing music, even after 8 mins?

BTW folks... The same issue can be encountered with Photobooth. Not sure if it uses the Mic input. It's just not audio....

Ok, perhaps I'm harping too strongly on the software being at fault. It's just that there's a lot of assumption making going on in my opinion and I don't think there's enough information to lay this problem at Apple's feet.

I personally don't have this problem. See my screens shots of iStat with both no music playing and music playing. The temps just aren't changing and my Xbench scores are the same whether iTunes is playing music or not. Therefore, this problem must in some way be environmental and related to your configuration.

Are your cooling fans ramping up when your computer becomes a space heater?

Could it be other hardware in the system that's causing a bizarre quirk? Is it the audio interface people are using? Is there something in common between people that have the problem besides just a Nehalem Xeon? A standard trouble-shooting technique is to pull every non-essential piece of hardware from the system to see if that helps... maybe that's worth a try?

I recall a problem where people with an ATI 4870 in their system were having problems with iTunes crashing when the computer resumed from sleep... that's a good example of the kinds of bizarre interactions that can occur in a computer. Perhaps that's even related to this issue?

I'm as stumped as anyone, and I realize I'm seriously late to this party, so I understand if there's no appetite to continue to trouble-shoot this, but I can't imagine that we don't collectively have what it takes to get to the bottom of this.

 

[nanofrog]

Your temps are pegged at 55(c) which is high for idle or playing music. My Quad 2009 MacPro, idles at 32(c) As does my 2008 Octa @ 31(c) and my Quad Wolfdale Q6600 at 31(c) The only one that jumps up in temps is the 2009 MacPro.

Are you saying, after a reboot that your idle temp is normally at 55(c)? How long does it take to rise and level off? Its hard understand that there is no effect in playing music, even after 8 mins?

This is what I'm wondering.

BTW folks... The same issue can be encountered with Photobooth. Not sure if it uses the Mic input. It's just not audio....

Good to know about this app. Thanks.

 

[VirtualRain]

@smacman, please don't mis-interpret me... although I don't have the issue, I'm not discrediting anyone here who has. However, in reading this thread, I was led to believe that...

a) Temps are rising by 20-deg C
b) CPU utilization was 1%
c) Fans were not spinning up

I was just going back to basics here. Given the evidence, you have to start with questioning the validity of the temp. increase. Again, I apologize for going back to basics on this if that had been ruled out long ago, but I never saw anyone discussing this at length.

I'm willing to help try to isolate what's different.

I really think we need to determine what the idle and load temps are under a trusted utility in Windows so we can asses the validity of iStat and the effectiveness of the cooling system before we go much further though.

I would suggest that if someone has a dual boot Nehalem Xeon, let's use speedfan or coretemp to measure idle and load temps and see where those stand in relation to iStat under OSX.

Also, if there's any info at all I can provide about my system, I'd be happy to do so.

My 2009 Mac Pro Quad was purchased in April 2009. It's now running dual GT120's and I'm using a Lexicon USB audio interface. The temps reported in my screen shots were taken about 30 minutes after resuming from sleep when I got home this evening and the the 2nd screenshot with iTunes was after playing those 4 songs in the playlist.

EDIT: It's also much more useful to talk about idle/load temps in relation to ambient. Otherwise, they are incomparable.

Cheers,
-VR.

 

[handheldgames]

Virtual Rain...

Try the test without your external audio gear.... Ambient temps in my place are about 68(f).

@smacman, please don't mis-interpret me... although I don't have the issue, I'm not discrediting anyone here who has. However, in reading this thread, I was led to believe that...

a) Temps are rising by 20-deg C
b) CPU utilization was 1%
c) Fans were not spinning up

I was just going back to basics here. Given the evidence, you have to start with questioning the validity of the temp. increase. Again, I apologize for going back to basics on this if that had been ruled out long ago, but I never saw anyone discussing this at length.

I'm willing to help try to isolate what's different.

I really think we need to determine what the idle and load temps are under a trusted utility in Windows so we can asses the validity of iStat and the effectiveness of the cooling system before we go much further though.

I would suggest that if someone has a dual boot Nehalem Xeon, let's use speedfan or coretemp to measure idle and load temps and see where those stand in relation to iStat under OSX.

Also, if there's any info at all I can provide about my system, I'd be happy to do so.

My 2009 Mac Pro Quad was purchased in April 2009. It's now running dual GT120's and I'm using a Lexicon USB audio interface. The temps reported in my screen shots were taken about 30 minutes after resuming from sleep when I got home this evening and the the 2nd screenshot with iTunes was after playing those 4 songs in the playlist.

EDIT: It's also much more useful to talk about idle/load temps in relation to ambient. Otherwise, they are incomparable.

Cheers,
-VR.

 

[nanofrog]

I really think we need to determine what the idle and load temps are under a trusted utility in Windows so we can asses the validity of iStat and the effectiveness of the cooling system before we go much further though.

iSTAT is fine for reading the temp rise, just not the actual values for idle or working temp, unless its been calibrated.

But to do what you propose, we need someone with an '09 MP and a means of accurately measuring the Tcase temps to do a calibration. The latter is the hard part, as not everyone has K type thermocouples + copper plate to embed it in, or an IR thermometer.

Also, if there's any info at all I can provide about my system, I'd be happy to do so.

Highly appreciated.

You might regret it though... Question, post... wash, rinse, repeat.

My 2009 Mac Pro Quad was purchased in April 2009. It's now running dual GT120's and I'm using a Lexicon USB audio interface. The temps reported in my screen shots were taken about 30 minutes after resuming from sleep when I got home this evening and the the 2nd screenshot with iTunes was after playing those 4 songs in the playlist.

1. When was it given a cold start last?
2. Have you tried pulling the USB (& any FW devices that aren't necessary to boot) and check the temps after a cold boot?

EDIT: It's also much more useful to talk about idle/load temps in relation to ambient. Otherwise, they are incomparable.

Quite true for exact values, and figures into the test data as we go (needs to be part of the posts).

 

[VirtualRain]

I'll spend some more time on this tomorrow doing a variety of tests (cold/warm starts with/without various accessories). Stay tuned.

EDIT: Here's a good read that outlines how challenging it is to accurately calculate CPU temperatures. It's a bit of a dark art, and every CPU family is different and every software implementation for CPU temperature reporting may not consider all the nuances correctly.

http://www.techpowerup.com/realtemp/docs.php

While not directly referencing Nehalem, this could still be very relevant (I found this comment under the errata related to 45nm Xeons for Hardware Monitor)...

When Intel introduced the third generation of Core and Core-based Xeon processors, manufactured in 45 nm technology ("Penryn" series), they also changed specific aspects how the Digital Thermal Sensors (DTS) on the chip cores are working. To check if your system is using a third generation Core processor, open the System Info window in the application and press the button More info… inside the window. A processor of this series will be identified as x86 Family: 6, x86 Model: 23.

If a processor of this series additionally identifies as being Xeon(R), it won't be possible to get an accurate per-core temperature reading. Those processors are equipped with Digital Thermal Sensors which do not work linear enough under normal operating conditions. They will only show accurate readings near the maximum allowed temperature limit. For this reason, Temperature Monitor won't display core temperature readings in that case, Hardware Monitor will display a "Hot" status information only.

Do we know for a fact that Nehalem Xeon's do not use the same DTS scheme?

From another Overclocking forum...

Sensors are not calibrated to report correct temps unless at TjMax (Throttling temp). Beyond 35C difference from TjMax the reported temperature starts to skew more and each chip seems to skew a different amount

And who knows how knowledgeable the authors of the various OSX temp reporting apps are. Are they doing the calculations correctly? Are the accounting for slope skew? Are they even reading the right registers?

Here's a 4000 post thread on XS frequented by the author of RealTemp for Windows that further demonstrates how challenging it is to keep up with reporting accurate CPU temperatures. (The Core i7 stuff starts to appear around page 100)...
http://www.xtremesystems.org/forums/showthread.php?t=179044

The bottom line is that CPU temps should not be trusted easily at the best of times. And do we really understand the relationships between Turbo Mode, C1E states, and core shutdown when it comes to impact on temperatures and how the OSX task schedular can impact core activity and transitions between states? Yikes!

 

[Transporteur]

IIRC, you and Seisend are the only ones that have reported that there's no problem with temps and iTunes.

This is one of the reasons I'm wondering if Apple's released a newer firmware revision in more recent systems that's an attempt to resolve the issue. No announcement or addition of it to an update yet, as it's still in testing (unbeknownst to the users that they're test subjects, and their systems are sending bug reports).

That should be easy to figure out. A comparison of the age of the Pros should do that.
Coconut Identity Card tells the approximate age of the machine.

 

[Inconsequential]

See my thread here for screenshots and benchmarks:

https://forums.macrumors.com/threads/833485/

 

[smacman]

I'll spend some more time on this tomorrow doing a variety of tests (cold/warm starts with/without various accessories). Stay tuned.

EDIT: Here's a good read that outlines how challenging it is to accurately calculate CPU temperatures. It's a bit of a dark art, and every CPU family is different and every software implementation for CPU temperature reporting may not consider all the nuances correctly.

http://www.techpowerup.com/realtemp/docs.php

While not directly referencing Nehalem, this could still be very relevant (I found this comment under the errata related to 45nm Xeons for Hardware Monitor)...



Do we know for a fact that Nehalem Xeon's do not use the same DTS scheme?

From another Overclocking forum...


And who knows how knowledgeable the authors of the various OSX temp reporting apps are. Are they doing the calculations correctly? Are the accounting for slope skew? Are they even reading the right registers?

Here's a 4000 post thread on XS frequented by the author of RealTemp for Windows that further demonstrates how challenging it is to keep up with reporting accurate CPU temperatures. (The Core i7 stuff starts to appear around page 100)...
http://www.xtremesystems.org/forums/showthread.php?t=179044

The bottom line is that CPU temps should not be trusted easily at the best of times. And do we really understand the relationships between Turbo Mode, C1E states, and core shutdown when it comes to impact on temperatures and how the OSX task schedular can impact core activity and transitions between states? Yikes!

Interesting articles and comments, but in my opinion, we need to make sure we don't spin our wheels here. I personally don't think the accuracy of the temps are that important. The fact that I and many others have noted the physical presence of heat coming from the machine while merely playing back audio is proof enough that something is amiss. I have no doubt that the max temperatures reached are probably still within Intel's specificaitons, it's just totally unnecessary thats all. What other architectural features of the chip are being compromised in this state? The actual temperature reached is not too important. What is important in my opinion is this:

1) The CPU gets substantially warmer when decoding any audio [Confirmed Physically]. This is not the case with other Macs, or the Mac Pro in Windows.

2) Benchmark tests of the CPU come in 25% lower when the machine is decoding audio. This is not the case on my Macbook Pro, Mac Mini, or the Mac Pro in Windows. [Confirmed]

3) Power Consumption rises markedly when decoding any audio [Unconfirmed] (Trusting iStat for this info)

As nanofrog stated earlier, there is definitely a rise in temperature happening that shouldn't be. Apple should fix this ASAP.

VirtualRain,

I'll try not to misinterpret what your saying, but I still get the impression that you don't necessarily believe that temperatures are actually rising here. Doesn't the fact that the machine outputs significantly hotter air within a minute or two of playing back audio provide enough evidence? I still think that your machine is not immune to this problem. If anything, it sounds like your machine might be stuck in this condition even when you stop audio playback.

 

[nanofrog]

While not directly referencing Nehalem, this could still be very relevant (I found this comment under the errata related to 45nm Xeons for Hardware Monitor)...

Do we know for a fact that Nehalem Xeon's do not use the same DTS scheme?

They are still diode thermal sensors. But what you may not realize, is that their typically accuracy = +/- 3C (across it's entire thermal range). It's not dead accurate, but close enough, especially as they're designed to kick in fans, not report temps to utilities. There's K type thermocouples that are accurate to +/- 1C.

In either case, even with the variance, you can still get an accurate rise in temps. That's what I was getting at in the earlier post.

It's certainly accurate enough to demonstrate that for a low CPU utilization, during audio decoding, that the thermal output of the CPU significantly increases.

For simple proof that the utility isn't totally wonked, is stick your hand at the exhaust vent on the back of the system. You'll feel it. You may not be able to measure it with your fingers, but you know its not completely false.

And who knows how knowledgeable the authors of the various OSX temp reporting apps are. Are they doing the calculations correctly? Are the accounting for slope skew? Are they even reading the right registers?

The data's reported directly off the CPU, and isn't up to OS X (the chip reports the rise as a negative number). So the utility just sees a temp, and does simple math. The profiles are designed into the chip.

The profile that has to be developed is for the fans in order to maintain the temp vs. power profiles Intel developed for the specific chip (i.e. 5500 series, and it's posted in the series datasheet; Table 6-2 IIRC).

The bottom line is that CPU temps should not be trusted easily at the best of times. And do we really understand the relationships between Turbo Mode, C1E states, and core shutdown when it comes to impact on temperatures and how the OSX task schedular can impact core activity and transitions between states? Yikes!

Actually, for what we need, they are accurate enough, as we're interested in the rise, not high accuracy of static temps (since the variance subtracts out. That's why Intel uses diodes. It works, and is a low cost solution.

I have no doubt that the max temperatures reached are probably still within Intel's specificaitons...

Intel's thermal profile lists 67C at TDP max across the LGA1366 parts.

They can go higher, but the actual temp isn't published. Tjmax is ~103 - 108C range, given the material the die is made of.

VirtualRain,

I'll try not to misinterpret what your saying, but I still get the impression that you don't necessarily believe that temperatures are actually rising here. Doesn't the fact that the machine outputs significantly hotter air within a minute or two of playing back audio provide enough evidence? I still think that your machine is not immune to this problem. If anything, it sounds like your machine might be stuck in this condition even when you stop audio playback.

I'm under the impression he doesn't realize that even if the accuracy of the sensors isn't the most accurate temp sensor ever developed, that the rise reported is still accurate.

So the comments seem to follow the belief that to get an accurate rise, the reported temps must be accurate. It makes sense, but it ignores one simple fact (and Intel relies on this), that the rise is accurate since the variances subtract out (same system = unchanged hardware; coolers, case, ambient air temp,... didn't change). Keep in mind, the CPU reports the data, and it's the temp rise, not individual temps (idle, working). It does the math. The utility then repeats it.

 

[SatyMahajan]

Core Audio

Curious if anyone has tried dropping an mp3 into a GarageBand and/or Logic session and done the playback test for 20 minutes to see if this problem still happens.

 

[smacman]

Curious if anyone has tried dropping an mp3 into a GarageBand and/or Logic session and done the playback test for 20 minutes to see if this problem still happens.

Yes. The problem still happens. In fact, anything that creates audio (even continuously pressing the keyboard volume keys will cause this).

 

[VirtualRain]

I haven't been home all day... just borrowing my GF's laptop here to say that I'll get back on this later tonight.

I'm not doubting that some people have temp increases (including me!). The intent of my prior posts was to point out it's very unwise to blindly trust CPU temperature reporting software. Every family of CPU's introduces quirks, requires tweaking of algorithms, and may have different non-linear characteristics away from TjMax... it really is not a simple problem and therefore, whenever you see erratic temps, the first thing to question is the reporting tool.

At any rate, there's no arguing with a noticeable increase in exhaust air temps.

I guess the first step for me is to determine if I have this issue all the time, some of the time, or not at all.

Another thing to think about is the complicated nature of the Nehalem power management...

Consider that Nehalems can not only utilize C1E states to lower the multi under low load conditions, but they can actually shut down entire cores. So consider that in an absolute idle state, a Nehalem may have half its cores shut-down and the other two running at a dramattically reduced (clock) multiplier. Temps might look really good in this state. Now, let's say you suddenly fire up even a moderately multi-threaded app... well all of a sudden, your two inactive cores come back online and your clock multi jumps to full rated speed and one or two cores might even go into turbo mode. It's not hard to see that the power consumption of the CPU might easily double rather suddenly. With the cooling system still in a lazy state, temps will also jump dramatically.

I'm not saying this is what's happening or that firing up iTunes and decoding music should have this dramatic effect on the processor's power state, but it very well could... either because OSX is not handling it well, the app itself is poorly coded, or this is how Nehalem is actually designed to respond to this kind of software... we may never really know the answer.

At any rate, I'll refrain from further speculation until we can do some more testing. Unfortunately, the tools in OSX are probably going to limit our ability to really understand what's going on. Is there any tool that can report Nehalem's core status in real time (C1E states, cores active/inactive, and clock multipliers)?

 

[smacman]

The intent of my prior posts was to point out it's very unwise to blindly trust CPU temperature reporting software. Every family of CPU's introduces quirks, requires tweaking of algorithms, and may have different non-linear characteristics away from TjMax... it really is not a simple problem and therefore, whenever you see erratic temps, the first thing to question is the reporting tool.

I think you are still missing the point. Who cares what the temperature is? The point is, there is a substantial temperature rise happening unnecessarily.

Consider that Nehalems can not only utilize C1E states to lower the multi under low load conditions, but they can actually shut down entire cores. So consider that in an absolute idle state, a Nehalem may have half its cores shut-down and the other two running at a dramattically reduced (clock) multiplier. Temps might look really good in this state. Now, let's say you suddenly fire up even a moderately multi-threaded app... well all of a sudden, your two inactive cores come back online and your clock multi jumps to full rated speed and one or two cores might even go into turbo mode. It's not hard to see that the power consumption of the CPU might easily double rather suddenly. With the cooling system still in a lazy state, temps will also jump dramatically.

Power consumption does not just double. It goes from 3W to 50W (1600% increase). Any audio causes this. Not just iTunes. Simply holding down F12 on your keyboard (volume adjustment) is enough to put the chip into this state. Remember also that this does not happen in Windows. I'm not buying your explanation. Something is grossly wrong here.

 

[SatyMahajan]

I haven't been home all day... just borrowing my GF's laptop here to say that I'll get back on this later tonight.

I'm not doubting that some people have temp increases (including me!). The intent of my prior posts was to point out it's very unwise to blindly trust CPU temperature reporting software. Every family of CPU's introduces quirks, requires tweaking of algorithms, and may have different non-linear characteristics away from TjMax... it really is not a simple problem and therefore, whenever you see erratic temps, the first thing to question is the reporting tool.

At any rate, there's no arguing with a noticeable increase in exhaust air temps.

I guess the first step for me is to determine if I have this issue all the time, some of the time, or not at all.

Another thing to think about is the complicated nature of the Nehalem power management...

Consider that Nehalems can not only utilize C1E states to lower the multi under low load conditions, but they can actually shut down entire cores. So consider that in an absolute idle state, a Nehalem may have half its cores shut-down and the other two running at a dramattically reduced (clock) multiplier. Temps might look really good in this state. Now, let's say you suddenly fire up even a moderately multi-threaded app... well all of a sudden, your two inactive cores come back online and your clock multi jumps to full rated speed and one or two cores might even go into turbo mode. It's not hard to see that the power consumption of the CPU might easily double rather suddenly. With the cooling system still in a lazy state, temps will also jump dramatically.

I'm not saying this is what's happening or that firing up iTunes and decoding music should have this dramatic effect on the processor's power state, but it very well could... either because OSX is not handling it well, the app itself is poorly coded, or this is how Nehalem is actually designed to respond to this kind of software... we may never really know the answer.

At any rate, I'll refrain from further speculation until we can do some more testing. Unfortunately, the tools in OSX are probably going to limit our ability to really understand what's going on. Is there any tool that can report Nehalem's core status in real time (C1E states, cores active/inactive, and clock multipliers)?

While I agree that tools (especially software ones, since I run a software team) are never 100% bug free, I think what others are pointing out is that the empirical, statistical data (measuring tool) directly correlates to the observed data (vent temp) for only the 2009 Mac Pros.

2009 Mac Pros seem to have evidence of a rise both empirically/statistically as well as observed whereas 2008 Mac Pros, 2009 i7 iMacs, etc. don't.

 

[DoFoT9]

I think you are still missing the point. Who cares what the temperature is? The point is, there is a substantial temperature rise happening unnecessarily.

VirtualRain is simply saying that you should not just trust the temperature readings that you are given. based on his observations and views, it is not wise for us to completely rule that out - i agree with him on that. the fact that the fans do not 'rev up' could possibly point to this. arent the speed of the fans directly linked to temperatures?

the point VR is making is that dont trust the temperature readouts at all. maybe they jump up ~20°C because of a software/firmware issue, and there is no temperature rise at all!? maybe apple uses diagnostic software that is not effected by these firmware/software issues? that could possibly explain why they dont see it as a "problem".

2009 Mac Pros seem to have evidence of a rise both empirically/statistically as well as observed whereas 2008 Mac Pros, 2009 i7 iMacs, etc. don't.

assuming that the readouts are correct that is

 

[smacman]

VirtualRain is simply saying that you should not just trust the temperature readings that you are given. based on his observations and views, it is not wise for us to completely rule that out - i agree with him on that. the fact that the fans do not 'rev up' could possibly point to this. arent the speed of the fans directly linked to temperatures?

the point VR is making is that dont trust the temperature readouts at all. maybe they jump up ~20°C because of a software/firmware issue, and there is no temperature rise at all!? maybe apple uses diagnostic software that is not effected by these firmware/software issues? that could possibly explain why they dont see it as a "problem".



assuming that the readouts are correct that is

And I am simply saying that the fact that HOT AIR blows out the back of my machine within minutes of playing audio, and becomes cool again when I stop playing audio is proof enough that there is a rise happening. The rise in the temp sensors of the adjacent memory banks is also supporting evidence that something is happening. I don't understand why this is so difficult to understand???

 

[DoFoT9]

And I am simply saying that the fact that HOT AIR blows out the back of my machine within minutes of playing audio, and becomes cool again when I stop playing audio is proof enough that there is a rise happening. The rise in the temp sensors of the adjacent memory banks is also supporting evidence that something is happening. I don't understand why this is so difficult to understand???

missed that post

 

[nanofrog]

Another thing to think about is the complicated nature of the Nehalem power management...

This is definitely more complicated, but the LGA1156 parts used in the Nehalem iMacs also have the same power management features as the LGA1366 parts. So that wouldn't be the cause.

The only architectural difference is the connection between the CPU and chipset (DMI vs. QPI)

Remember also that this does not happen in Windows. I'm not buying your explanation. Something is grossly wrong here.

To me, I'm thinking OS X isn't optimized for Nehalem, while Win7 is. And it's causing the QPI section to work properly or not, depending on the OS loaded.

VirtualRain is simply saying that you should not just trust the temperature readings that you are given. based on his observations and views, it is not wise for us to completely rule that out - i agree with him on that. the fact that the fans do not 'rev up' could possibly point to this. arent the speed of the fans directly linked to temperatures?

They're off, and potentially in the ~15C range, unless you calibrate it with a temp probe of some sort. So in cases that this isn't possible, you don't trust the temp readings at all. But if its been calibrated, you enter an offset in the utility. Accuracy is +/- 3C at this point (actual diode accuracy).

They aren't perfectly linear in their reporting response, but the worse case of it out of whack, is 3C. That's not horrible.

As per the fan ramping or not, it depends on where it hits the Thermal Profile developed by Intel (can be found in the datasheets for the CPU series). Apple has to get the fans to fit this, or run the risk of damaging the CPU's, and shortening their lifespan. If it's over the curve, then the fans will ramp. Otherwise, they'll stay at the minimum RPM setting.

the point VR is making is that dont trust the temperature readouts at all. maybe they jump up ~20°C because of a software/firmware issue, and there is no temperature rise at all!? maybe apple uses diagnostic software that is not effected by these firmware/software issues? that could possibly explain why they dont see it as a "problem".



assuming that the readouts are correct that is

The rise is real though. You only need to stick you hand at the exhaust vent to prove it's not a phantom reading by the software.

 

[SatyMahajan]

Obviously people have empirical evidence (however inaccurate) that some how correlates to observed evidence (since we know the rises/falls somewhat line up).

I think the point of contention is not whether THERE IS a problem but rather HOW BIG of a problem is it.

Only way we'll know is if we have measurement tools of which we understand/trust quality.

 

[nanofrog]

Obviously people have empirical evidence (however inaccurate) that some how correlates to observed evidence (since we know the rises/falls somewhat line up).

I think the point of contention is not whether THERE IS a problem but rather HOW BIG of a problem is it.

Only way we'll know is if we have measurement tools of which we understand/trust quality.

I want to know the cause, as though the exact extent isn't known (i.e. precise temps), it's significant enough for the concern to be warranted.

And the use of Windows can confirm it's not a fault of Intel (same process = functions as expected on the same system under Windows, not OS X). That lends the causality to be either software (OS X), or possibly firmware.

 

[SatyMahajan]

I want to know the cause, as though the exact extent isn't known (i.e. precise temps), it's significant enough for the concern to be warranted.

And the use of Windows can confirm it's not a fault of Intel (same process = functions as expected on the same system under Windows, not OS X). That lends the causality to be either software (OS X), or possibly firmware.

As do I my friend.

I don't think it's HW, as you said that was ruled out by holding it constant and using two different OS with same software (iTunes).

I can tell you this. I have After Effects CS3 and after moving to the 2009 Nehalem all of my comps that required the Cycore FX (that ship with AE) crashed every time I opened them. I narrowed it down to being a problem with Hyperthreading, which I tested by turning HT off through the proc pref pane. When I purchased the updated version of the Cycore FX HD (which was annoying but fine since they were better anyway), problem went away, I confirmed with one of the sales reps that it was a problem with Hyperthreading that they resolved in software fix.

So....I think quite frankly it's just a bug in the CoreAudio portion of the SW stack. iTunes causes it, Volume up down causes it, Quicktime causes it, Garageband/Logic causes it. No one in pro audio would have noticed this (or cared for that matter) as they assume that the system is going to be taxed, so they won't be reporting this problem, but the rest of us can and hope that Apple addresses it.

That being said, I just sent an email to a buddy in the CoreAudio group. Let's see what happens.

 

[smacman]

As do I my friend.

I don't think it's HW, as you said that was ruled out by holding it constant and using two different OS with same software (iTunes).

I can tell you this. I have After Effects CS3 and after moving to the 2009 Nehalem all of my comps that required the Cycore FX (that ship with AE) crashed every time I opened them. I narrowed it down to being a problem with Hyperthreading, which I tested by turning HT off through the proc pref pane. When I purchased the updated version of the Cycore FX HD (which was annoying but fine since they were better anyway), problem went away, I confirmed with one of the sales reps that it was a problem with Hyperthreading that they resolved in software fix.

So....I think quite frankly it's just a bug in the CoreAudio portion of the SW stack. iTunes causes it, Volume up down causes it, Quicktime causes it, Garageband/Logic causes it. No one in pro audio would have noticed this (or cared for that matter) as they assume that the system is going to be taxed, so they won't be reporting this problem, but the rest of us can and hope that Apple addresses it.

That being said, I just sent an email to a buddy in the CoreAudio group. Let's see what happens.

Very interesting points and I totally agree with your analysis. Let's hope your friend might be able to help or point us in the right direction.

 

[VirtualRain]

Is there any OSX tool that can report Nehalem's core status in real time (C1E states, cores active/inactive, and clock multipliers)?