Audio Decoding KILLS MacPro (2009) green factor

[Inconsequential]

Just to clarify with people

http://www.thunder-keep.co.uk/site/macproissue/itunes_playing.jpg

Here:

CPU Core 1/3/5/7 are the Tcore values taken directly from the core itself.

CPU A Temperature diode is Tcase I believe.

Tcore will ALWAYS be greater than Tcase.

iStat Menus, on my MacBook Pro at-least is showing Tcase.

Anyway, not really the point. We're arguing on random things which does not matter to get our problem fixed.

We are all agreed there is a temperature rise, we are not agreed to what extent it is on what machine. But IT DOES NOT MATTER.





Anyway:

What we need to do (and now i've got a useful response from apple that the performance drop is not normal) is to get consistent tests and results from programs.

My plan is when I get back to the machine:

Tests with and without audio playing:
1. Start VMWare Fusion with Windows 7 x64, restart it twice and shut it down and time how long it takes to do it.
2. Run the photoshop test (http://www.retouchartists.com/) three times.
3. Export 100 Lightroom images (RAW 10-15Mb 40D) with Lightroom 2.5.
4. Encode a DVD chapter of "Fawlty Towers".
5. Create a 1080p video in Quicktime 7 from 1540 4Mb JPEG images and time how long it takes.

Over all these separate tasks I will be running Temperature Monitor to graph it and the CPU usage.

Seeing as thats what I use my Mac Pro for I think that would prove the 25% performance loss theory

I suggest others do the same with the tasks they do, as if apple fixes the performance loss the other problems go away too.

CR

 

[cl-user]

cl-user: It's not in the hardware, but in OS X. That's been proven for some time, and smacman has just provided the latest confirmation.

We may come to that conclusion, but we don't seem to be in tune with whatever state Apple is in on this.

The glib brush offs with excuses like our machines roasting because our music was sourced other than from Apple, are crass beyond belief.

No matter how arrogant or blind Apple's defensive posture may seem to be, I assume it is in our interests to take their possible reasoning into account.

If (or when) we can cook up a fix for this, we'll put the whole issue to bed. But until then, we seem only to have Apple as a possible solution.

But their attitude seems to be to leave the AppleCare/Store bimbos to give us the run around.

 

[Max(IT)]

Just to clarify with people

http://www.thunder-keep.co.uk/site/macproissue/itunes_playing.jpg

Here:

CPU Core 1/3/5/7 are the Tcore values taken directly from the core itself.

CPU A Temperature diode is Tcase I believe.

Tcore will ALWAYS be greater than Tcase.

iStat Menus, on my MacBook Pro at-least is showing Tcase.

Anyway, not really the point. We're arguing on random things which does not matter to get our problem fixed.

We are all agreed there is a temperature rise, we are not agreed to what extent it is on what machine. But IT DOES NOT MATTER.

Perfectly clear , so he's probably speaking about monitoring Tcore and not Tcase. Not a huge difference anyway, and Tcase it's diode temperature, not the heatsink's temperature as he said

 

[nanofrog]

You said that the simple fact of CONNECTING an usb o fw peripheral cause a raise in cpu temperature.
I did, and the temperature increase was very limited. I don't know about audio peripheral because I have none.

It seems connecting an FW audio device is enough. For FW drives, writing is what's been posted on.

Sorry about any confusion, but it's buried in the thread, and not nearly as much to isolate the details yet.

I know, and in iStat Pro you have two reads: Cpu Heatsink and CPU.
The first one is the K-type and the second one is che cpu diode.
So why you (actually not you, but smacman ) are keep saying that iStat readout is inaccurate ? It's the same of Temperature Monitor ...
You can check both CPU diode and CPU heatsink

The thermocouple's reading would be acurate, but with the diode on the CPU die, there's a calibration requirement to use it for an acceptable real temp. It requires a measured temp of the IHS.

But even without a calibration, the rise in temp indicated by the CPU diode is accurate, as the variance subtracts out.

Anyone tested firewire + audio?
I assume highest common denominator, not sum.

IIRC, it has been, and the temps where high. FW drives (HDD's) were a bit different, as the post indicated it was in high write use.

We may come to that conclusion, but we don't seem to be in tune with whatever state Apple is in on this.

The glib brush offs with excuses like our machines roasting because our music was sourced other than from Apple, are crass beyond belief.

No matter how arrogant or blind Apple's defensive posture may seem to be, I assume it is in our interests to take their possible reasoning into account.

If (or when) we can cook up a fix for this, we'll put the whole issue to bed. But until then, we seem only to have Apple as a possible solution.

But their attitude seems to be to leave the AppleCare/Store bimbos to give us the run around.

They won't admit to it officially, as that's potentially usable in a court, should a lawsuit ever happen (i.e. Class Action suit). So the official response is denial in order to cover their butts down the road, if ever needed.

 

[lbodnar]

The other important factor, is the ambient temp. That's the reason for before and after screen shots with each test. Ambient will have a significant effect on the reported temps.
...
There's 2x sensors. One is a K-type thermocouple in the heatsink (Tcase), and the other is a diode in the CPU die itself (Tcore). The accuracy of a K-type is +/- 1C, and the diode +/- 3C typically (manufacturer's data).

One of SMC main tasks is to control Tcase - it will activate forced cooling (fans) as soon as heatsink temperature approaches specified limit (63-67ºC for different Xeons.) Core temperatures are not controlled and are not even specified by Intel (apart from mobile CPUs where monitoring Tcase is problematic due to space restrictions.) SMC has no access to Tcore data.

Typically silicon functions fine for Tj below 100-120ºC.

Where did you get your information that Apple uses K-type thermocouple?
If so it is an odd engineering solution (requires high gain stabilised amplifier and much overspecified for simple +-1ºC accuracy.)

 

[AppleWorking]

It's those running over 81C that have a right to be concerned...

Are you serious about this? When I rip/encode a DVD in Vista I sometimes see my core temps @ 85C for hours. Sometimes for 12-14 hours. Is this something I should be concerned about? Do you think I could have potentially damaged my CPU (see sig)? (BTW, I ramped my fans pretty high in OSX and then booted into vista to do the encoding but the temps were still hovering around 85C.) Thanks

Edit: Used RealTemp to monitor...

 

[DoFoT9]

Are you serious about this? When I rip/encode a DVD in Vista I sometimes see my core temps @ 85C for hours. Sometimes for 12-14 hours. Is this something I should be concerned about? Do you think I could have potentially damaged my CPU (see sig)? (BTW, I ramped my fans pretty high in OSX and then booted into vista to do the encoding but the temps were still hovering around 85C.) Thanks

for a 12 hour period? no.. of course not! for days upon days (or months even) (especially with a sound card) of course!

 

[nanofrog]

One of SMC main tasks is to control Tcase - it will activate forced cooling (fans) as soon as heatsink temperature approaches specified limit (63-67ºC for different Xeons.) Core temperatures are not controlled and are not even specified by Intel (apart from mobile CPUs where monitoring Tcase is problematic due to space restrictions.) SMC has no access to Tcore data.

Typically silicon functions fine for Tj below 100-120ºC.

Where did you get your information that Apple uses K-type thermocouple?
If so it is an odd engineering solution (requires high gain stabilised amplifier and much overspecified for simple +-1ºC accuracy.)

Intel's Thermal Profiles are there for cooling purposes (gives load vs. temp in both a chart and graph) for the fans. The only control of Tcore per se, is the power management features in the core itself (Tcore hits a certain limit, and the multiplier is lowered).

Most of the Tj limit's I've seen are typically listed ~ 105 - 108C (presumably to allow for some safety margin; Intel doesn't actually list it), but it's not far off, as the materials are still doped semi.

As per the K types, most I've seen and used have a spec of +/1 1C (some are even 0.5C rated, and there's some that aren't that high, and rated at 2C <entire range>, such as an AN1226). A simple amp and thermocouple are all that's needed (i.e. 8-DIP/8-CDIP, TO-5 package for example). They're also not terribly expensive. IC's can be had for just over $3USD in quantity.

But most boards don't use it at all. Apple's do.

 

[lbodnar]

Are you serious about this? When I rip/encode a DVD in Vista I sometimes see my core temps @ 85C for hours. Sometimes for 12-14 hours. Is this something I should be concerned about? Do you think I could have potentially damaged my CPU (see sig)? (BTW, I ramped my fans pretty high in OSX and then booted into vista to do the encoding but the temps were still hovering around 85C.) Thanks

Tcore up to 100ºC is fine! Who started this panic?

 

[DoFoT9]

Tcore up to 100ºC is fine! Who started this panic?

for a few hours it may be fine. not for endless amounts of time. the temperature isnt the only worry though, heat/power/noise etc also come into play.

 

[lbodnar]

Intel's Thermal Profiles are there for cooling purposes (gives load vs. temp in both a chart and graph) for the fans. The only control of Tcore per se, is the power management features in the core itself (Tcore hits a certain limit, and the multiplier is lowered).

Most of the Tj limit's I've seen are typically listed ~ 105 - 108C (presumably to allow for some safety margin; Intel doesn't actually list it), but it's not far off, as the materials are still doped semi.

As per the K types, most I've seen and used have a spec of +/1 1C (some are even 0.5C rated, and there's some that aren't that high, and rated at 2C <entire range>, such as an AN1226). A simple amp and thermocouple are all that's needed (i.e. 8-DIP/8-CDIP, TO-5 package for example). They're also not terribly expensive. IC's can be had for just over $3USD in quantity.

But most boards don't use it at all. Apple's do.

CPU diodes are typically +/- 3C from what I've used. Even from years ago (National Semi LN32).

Thanks, I am an electronics engineer and I know how to interface to thermocouples and how they are used in the lab environment. I just asked the source of your information that Apple uses thermocouple in production MacPro.

 

[VirtualRain]

Here's a brief primer on CPU temps...

Each CPU has a value called TjMax (Tjunction Max) which is the temperature at which the CPU will begin throttling to avoid damage. In the past, TjMax was not published data, so temp reporting software authors had to guess at the value of TjMax. With Nehalem, this changed, TjMax (or rather TjTarget) for each processor can now be read by software directly from the CPU. It seems that most Nehalem CPU's have a TjMax value of 100-deg. (source - from the author of RealTemp)

(BTW, Nanofrog, you say TjMax is 105-108... can you provide a link to some research? That's not what the folks at XS have found)

There are two types of temperature diodes in an Intel processor...
Tcase is under the heat spreader
Tjunction is at the junction of the core and substrate and there is one for each physical core in the processor. It's actually reported by the processor as the distance from TjMax. Thus if a value of 35 is reported, the software must subtract this from TjMax to get an absolute core temperature (if TjMax is 100, in this example, a core temp of 65-deg would be reported). In the past, without published TjMax values, this meant that reported core temps, were simply the software author's best guess. Now, with Nehalem, and a register containing the actual CPU's TjMax (or more correctly TjTarget) value, temp monitoring software authors can actually report the real core temps.

I highly recommend Hardware Monitor for monitoring your temps...
http://www.bresink.com/osx/HardwareMonitor.html
- It reports core temps (Tjunction)
- It's been proven to be accurate (thanks to smacman)
- It will log changes in a graph

We don't know what TjMax value Hardware Monitor is using, but smacman comparing it to Real Temp (which reads the actual TjMax from the CPU) in Windows indicates it's reporting the same values, leading me to believe it's using the correct TjMax and can be trusted.

If you aren't familiar with Real Temp for Windows or want to learn more about temperature monitoring of CPU's, I highly suggest you read this brief primer on the program.

Alternatively, you can go through this 4000 post thread on XtremeSystems where the author of Real Temp provides some great insights into the challenges of cpu temperature monitoring.

At any rate, iStat simply reports Tcase which does not map linearly to Tjunction, and I would not rely on this.

Cheers!

 

[lbodnar]

for a few hours it may be fine. not for endless amounts of time...

thank you for your valuable input

MTBF is either tabulated or specified for fixed temperature e.g. 60ºC and then derated. I have not seen Intel's reliability data reports but Motorola/Freescale used to publish them for G4 series. Generally speaking lifetime of the silicon is reduced by the factor of 2 with each 7-10ºC increase in junction temperature. However it starts with something like 10^6 hours (>100 years.)

 

[DoFoT9]

thank you for your valuable input

thats ok

i was merely stating that this problem is not like those that other users/computer are used to.

yes, the CPUs are rated to last x years at y temperatures but that isnt my point. my point is that it also effects other things and this particular issue is clearly constant. users of the 2009 MP + FW audio device who never put their computer to sleep could be severely degrading the quality/longevity of their computers, increasing their power usage bill by a considerable amount, decreasing the performances of the machine and adding extra stresses onto their machines that simply should not be there! we can get into the technical aspects of it all, but thats besides the point - arguing over whether it will make a difference is stupid IMO.

in the end - it shouldnt be happening in the first place and it needs to be addressed by apple.

 

[lbodnar]

in the end - it shouldnt be happening in the first place and it needs to be addressed by apple.

I agree absolutely. But the arguments should be chosen carefully. If someone claims that this reduces the longevity of the MacPro hardware Apple engineers will just laugh their backsides off because it does not. If one would claim that this issue puts 2009 MacPros at the very bottom rung of least environment friendly computing cr*p ever produced then maybe Apple suits will listen.

 

[DoFoT9]

I agree absolutely. But the arguments should be chosen carefully. If someone claims that this reduces the longevity of the MacPro hardware Apple engineers will just laugh their backsides off because it does not. If one would claim that this issue puts 2009 MacPros at the very bottom rung of least environment friendly computing cr*p ever produced then maybe Apple suits will listen.

good point. the power increases/performance hits might also make them listen. i doubt it though.

 

[nanofrog]

Thanks, I am an electronics engineer and I know how to interface to thermocouples and how they are used in the lab environment. I just asked the source of your information that Apple uses thermocouple in production MacPro.

My point was Apple's not using the CPU diodes as other board makers do for Tcase.

As per being an engineer, I didn't know, but you certainly had the grasp of what was going on from a technical POV, and it's interesting to discuss to me.

(BTW, Nanofrog, you say TjMax is 105-108... can you provide a link to some research? That's not what the folks at XS have found)

Intel's data is unpublished as you know, but that information is from temp monitor software pages (Real Temp), as well as what I've seen from random parts tesing (Milspec HOTL that ends up destroying parts). Some of that may be annectdotal (i.e. Real Temp), and perhaps a bit outdated (the materials testing is from '06 parts). But the interconnects keep shrinking, and the materials aren't changing (copper interconnects). Poor design (i.e. missed/over-ruled the design rules) in the EDA suite, or if the suite is actually borked. As I understand it, that's happened too.

...Real Temp for Windows...

I use it.

Alternatively, you can go through this 4000 post thread on XtremeSystems where the author of Real Temp provides some great insights into the challenges of cpu temperature monitoring.

I've read through it some time ago, but I'll try to give it a go again. As you say, it's rather long. //blech...gag....

At any rate, iStat simply reports Tcase which does not map linearly to Tjunction, and I would not rely on this.

Tcase is what we really need though, as that's what the Thermal Profiles are, not Tcore. As it seems reliable (via comparitive confirmation), all the better.

 

[VirtualRain]

Tcase is what we really need though, as that's what the Thermal Profiles are, not Tcore. As it seems reliable (via comparitive confirmation), all the better.

The sad thing is that the thermal profiles are not up to the task. On a 4 core stress test I ran (not 8 threads, but 4), the fans did not ramp up until Tcase hit 80-deg C and by that time Tjunction (core temps) were around 95-deg... alarmingly close to thermal limiting. Even more bizarre, is that the fans only ramped up by about 100RPM. The thermal profiles are extremely inadequate. In fact smacman was seeing instability on his test at these temps. Ugh!

My recommendation to anyone running CPU intensive tasks is to get SMCfancontrol or something similar and manually ramp up the fans.

 

[smacman]

I used both: temperature reads on iStat Pro and Temperature Monitor are almost the same
Just a question for you: why do you think that iStat Pro and Temperature Monitor read from different sources? Do you have any link to documents reporting this?
absolutely unnecessary I agree, but not dangerous in most cases


I read all the thread very carefully


I know but I don't understand why he wrote "the same on my MBP", so I asked for a clarification

This thread is starting to go the way of the one on Apple boards and that is unfortunate. People are jumping into it, without taking the time to go back and read it from the beginning. People are commenting on their temps, without having a clear understanding of the different temperature sensors in their machines, and without even bothering to read or understand Intel's thermal specifications. Others are arguing whether this is even an issue or not since these temps probably won't damage the CPU, and totally missing the fact that these temps are merely a symptom of a wider problem.

What is really troubling however, is a conversation I had yesterday while speaking to a supervisor at Applecare. After calling to report all of this for the 4th time in a couple of months, I asked him if he had followed any of these threads. He told me that he had briefly read them, and was very quick to note that many people posting were also not concerned by this behavior (I suspect he has been reading the Apple Support Discussions which has been taken over by Apple Defenders lately). His advice was to uninstall any temperature monitoring tools, and accept this as normal. He also confirmed there is no fix in the works from Apple.

So as it stands, I am stuck with a machine, that if pushed hard while playing audio, overheats and suffers from a Kernel panic. The only solution is to run Windows, or not listen to audio, write to FW disks, or use audio apps. Applecare's response was that applications such as Handbrake, and CPU Test, are unsupported by Apple, and unless I can show instability while using only Apple software, it will not be covered under warranty. Too bad most Apple software is so poorly optimized, that it can't even push the cores anywhere near 100%. What scared me, is the tone of his response, which for the first time felt like it had come from higher up in the chain, and that he was just a messenger this time.

In posts above, it has been established that the CPU Temperature Diode sensor most likely approximates Tcase. Intel's max stated Tcase temp for the W3540 chip (Taken from their Thermal Specifications Document) is 67C. Using Temperature Monitor (which is the only application that is not ambiguous about which sensor is being monitored), my CPU Diode temperature is well into the 80C range by simply playing a song, launching an audio app such as Logic, or writing to a firewire drive. Even my CPU Heatsink measured value exceeds 67C if I play audio for long enough.

People also need to consider that the temperature symptom from this issue will vary in severity. The temperatures seen will depend on whether measured on a Quad system, or Octo system. They will also depend on the clock speed of the chip, as lower clocked chips inherently run a few degrees cooler. Ambient temp also has a large effect. Apple lists in their specs that the machine can be operated anywhere in the 10C-35C range. Given what I am seeing in a room that is 26C, I can't imagine running these things in a 35C room.

The performance issue is going to be a tough one to quantify. I have noticed myself, that it is very difficult to measure any sort of noticeable performance hit in typical applications (Virtual Machines, Adobe CS4, Lightroom, etc.) with audio playing. The only applications that do consistently show a loss are apps that stress all CPU cores simultaneously. Unfortunately these apps are few and far between (benchmarks, handbrake). In any case, something is not right. Benchmarking apps are clearly affected by audio playback, and that is not the case on any other Mac.

Lastly, the real issue of grossly high power demand is real and disturbing. No other machine consumes this much power when simply playing audio, or writing to a FW disk. Even this machine does not consume this amount of power in Windows, yet Apple claims that the power usage is "within limits".

For those of you that are actually concerned by these issues, now that it is becoming more and more clear that Apple is not going to do anything about this, what is our next step? For those of you that are totally happy with this situation, and don't see the need for a fix, please spare yourself the time or start your own thread somewhere else..

 

[nanofrog]

The sad thing is that the thermal profiles are not up to the task. On a 4 core stress test I ran (not 8 threads, but 4), the fans did not ramp up until Tcase hit 80-deg C and by that time Tjunction (core temps) were around 95-deg... alarmingly close to thermal limiting. Even more bizarre, is that the fans only ramped up by about 100RPM. The thermal profiles are extremely inadequate. In fact smacman was seeing instability on his test at these temps. Ugh!

My recommendation to anyone running CPU intensive tasks is to get SMCfancontrol or something similar and manually ramp up the fans.

I recall you bringing this up before, and it's definitely a problem. But I don't see it as a failure of Intel's Thermal Profiles (not seen this in PC's, but I've only a single system to test properly), but the code in the SMC. A re-write could be done to fix it, and sent as part of an update.

In the meantime, your suggestion of SMCFanControl is a really good one.

This thread is starting to go the way of the one on Apple boards and that is unfortunate. People are jumping into it, without taking the time to go back and read it from the beginning. People are commenting on their temps, without having a clear understanding of the different temperature sensors in their machines, and without even bothering to read or understand Intel's thermal specifications. Others are arguing whether this is even an issue or not since these temps probably won't damage the CPU, and totally missing the fact that these temps are merely a symptom of a wider problem.

What is really troubling however, is a conversation I had yesterday while speaking to a supervisor at Applecare. After calling to report all of this for the 4th time in a couple of months, I asked him if he had followed any of these threads. He told me that he had briefly read them, and was very quick to note that many people posting were also not concerned by this behavior (I suspect he has been reading the Apple Support Discussions which has been taken over by Apple Defenders lately). His advice was to uninstall any temperature monitoring tools, and accept this as normal. He also confirmed there is no fix in the works from Apple.

So as it stands, I am stuck with a machine, that if pushed hard while playing audio, overheats and suffers from a Kernel panic. The only solution is to run Windows, or not listen to audio, write to FW disks, or use audio apps. Applecare's response was that applications such as Handbrake, and CPU Test, are unsupported by Apple, and unless I can show instability while using only Apple software, it will not be covered under warranty. Too bad most Apple software is so poorly optimized, that it can't even push the cores anywhere near 100%. What scared me, is the tone of his response, which for the first time felt like it had come from higher up in the chain, and that he was just a messenger this time.

In posts above, it has been established that the CPU Temperature Diode sensor most likely approximates Tcase. Intel's max stated Tcase temp for the W3540 chip (Taken from their Thermal Specifications Document) is 67C. Using Temperature Monitor (which is the only application that is not ambiguous about which sensor is being monitored), my CPU Diode temperature is well into the 80C range by simply playing a song, launching an audio app such as Logic, or writing to a firewire drive. Even my CPU Heatsink measured value exceeds 67C if I play audio for long enough.

People also need to consider that the temperature symptom from this issue will vary in severity. The temperatures seen will depend on whether measured on a Quad system, or Octo system. They will also depend on the clock speed of the chip, as lower clocked chips inherently run a few degrees cooler. Ambient temp also has a large effect. Apple lists in their specs that the machine can be operated anywhere in the 10C-35C range. Given what I am seeing in a room that is 26C, I can't imagine running these things in a 35C room.

The performance issue is going to be a tough one to quantify. I have noticed myself, that it is very difficult to measure any sort of noticeable performance hit in typical applications (Virtual Machines, Adobe CS4, Lightroom, etc.) with audio playing. The only applications that do consistently show a loss are apps that stress all CPU cores simultaneously. Unfortunately these apps are few and far between (benchmarks, handbrake). In any case, something is not right. Benchmarking apps are clearly affected by audio playback, and that is not the case on any other Mac.

Lastly, the real issue of grossly high power demand is real and disturbing. No other machine consumes this much power when simply playing audio, or writing to a FW disk. Even this machine does not consume this amount of power in Windows, yet Apple claims that the power usage is "within limits".

For those of you that are actually concerned by these issues, now that it is becoming more and more clear that Apple is not going to do anything about this, what is our next step? For those of you that are totally happy with this situation, and don't see the need for a fix, please spare yourself the time or start your own thread somewhere else..

Ouch.

I'm wondering how many filings with engineering have actually been made, as a flood would seemingly be hard to ignore.

 

[VirtualRain]

@smacman... +1... Well said. I would advocate cutting newbies some slack... but there's no excuse for not reading the thread before taking a stubborn position on this. I was in denial at first myself, but saw the light.

At any rate, sorry to hear this feedback. I think for those who's work is impacted by this, we can only hope that the 2010 Gulftown rigs don't suffer the same problem. Then if that's the case, sell the 09 and buy a new one.

 

[nanofrog]

+1 Well said.

Sorry to hear this feedback. I think for those who's work is impacted by this, we can only hope that the 2010 Gulftown rigs don't suffer the same problem. Then if that's the case, sell the 09 and buy a new one.

Unfortunately, as the chips are a Tick (same architecture with a pair of additional cores on a die shrink), it's going to be similar (actual temp values may change, but the issue remain).

It's going to take an OS X update to fix it it seems, as it's not happening under windows on the same systems. So deductive reasoning points to the lack of optimization for Nehalem architecture in OS X as it currently exists.

 

[Kallel00]

Is the overheating issue also while playing movies (ie, HD movie + audio out) ?

Asking because I dont want the fans to start revving faster each time I play a movie

 

[Inconsequential]

What is really troubling however, is a conversation I had yesterday while speaking to a supervisor at Applecare. After calling to report all of this for the 4th time in a couple of months, I asked him if he had followed any of these threads. He told me that he had briefly read them, and was very quick to note that many people posting were also not concerned by this behavior (I suspect he has been reading the Apple Support Discussions which has been taken over by Apple Defenders lately). His advice was to uninstall any temperature monitoring tools, and accept this as normal. He also confirmed there is no fix in the works from Apple.

I've had a response back from apple care and say the temperatures and power usage *are* within limits *however* the drop in performance *shouldn't* be happening and have sent me a piece of software to run. They also asked me to run every possible test and system setup, which I will do this weekend before my Applecare rep gets back on Monday.

In posts above, it has been established that the CPU Temperature Diode sensor most likely approximates Tcase. Intel's max stated Tcase temp for the W3540 chip (Taken from their Thermal Specifications Document) is 67C. Using Temperature Monitor (which is the only application that is not ambiguous about which sensor is being monitored), my CPU Diode temperature is well into the 80C range by simply playing a song, launching an audio app such as Logic, or writing to a firewire drive. Even my CPU Heatsink measured value exceeds 67C if I play audio for long enough.

I think you may have a more dodgy system because my 2.66 has *never* (as far as I can remember) gone above 70ish Tcase (again I will confirm when im back!)

The performance issue is going to be a tough one to quantify. I have noticed myself, that it is very difficult to measure any sort of noticeable performance hit in typical applications (Virtual Machines, Adobe CS4, Lightroom, etc.) with audio playing. The only applications that do consistently show a loss are apps that stress all CPU cores simultaneously. Unfortunately these apps are few and far between (benchmarks, handbrake). In any case, something is not right. Benchmarking apps are clearly affected by audio playback, and that is not the case on any other Mac.

Well my tests listed above will show (over a decent time period) a reduction in performance, but again, I can't do that quite yet.

Lastly, the real issue of grossly high power demand is real and disturbing. No other machine consumes this much power when simply playing audio, or writing to a FW disk. Even this machine does not consume this amount of power in Windows, yet Apple claims that the power usage is "within limits".

For those of you that are actually concerned by these issues, now that it is becoming more and more clear that Apple is not going to do anything about this, what is our next step? For those of you that are totally happy with this situation, and don't see the need for a fix, please spare yourself the time or start your own thread somewhere else..

Agreed. I will be buying a small power meter to compare to iStat myself which should show apple that it is *not* what should be happening.

I will keep going with applecare until the end of January, next step is *shrug*.

However, we aren't really organised here. We should push applecare in the direction of the loss of performance, as the power usage and temperatures are almost certainly a result of that.

I will, when i've finished my testing for applecare, will post in full, my results in a new thread dedicated to gathering a collection of data without all the chit chat, which can occur in this thread (as showing this link to engineering just makes them confused and just makes us seem like we don't know what we are on about).

We need concise, accurate and continual proof of an issue with everyones machines. In the course of this thread we have found that its not just audio but USB/FW HDs etc that cause the issue too and I will be testing these things as well.

We are all intelligent people here who *are* able to group together in one cause with one "voice" so to speak. But rambling on about things isn't going to help.

If we need someone to start then next post and show my findings then I will do that and we can proceed from there.

 

[lbodnar]

Abnormally high temperature differential between diode and heatsink signals a problem with thermal interface material between the IHS and heatsink.

In any case Apple's assertion that their hardware only supports Apple software must be ridiculous even to themselves. Any software was probably compiled with Xcode in the end...