Oh look. Apple has released (another firmware update) to resolve flickering screens from their precious Imac line. No love for pro owners. I'm going to go throw up now.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Audio Decoding KILLS MacPro (2009) green factor
- Thread starter handheldgames
- Start date
- Sort by reaction score
I wasn't defending Apple, hell, I even offered up the new .kext's to people to check out.
The point is:
1) This thread is way too f'cking big and to expect anyone to read through 1000+ posts to figure out what the hell is going on, myself included, is unrealistic.
2) How can we actually be sure these temperature monitoring apps, by third parties, are even accurate? From Apple's standpoint they wouldn't be, so we would need other, measurable proof, like performance hits
3) When I did try to replicate something with undeniable proof, like the supposed 25% performance drop with audio playing, I had ZERO performance drop (on the quad-3.33)
If this is a problem, and again, I have an octo-2.26 at home where according to iStat my CPU is at 60C most of the time, I would love to get it fixed, if for no other reason to lower my electric bill.
BUT, Apple is not going to trust third party app data like temperature details, and I can't replicate the performance hit. Even at home on my octo-2.26, there was only a 3% change playing audio while Handbrake was encoding in the background.
I also have the ear of a senior level Apple tech who genuinely seemed interested in looking into this (I had to call in this past weekend for a work issue and after that was fixed I casually mentioned this) and he sent me the CaptureData.app without me even pushing for it, but now I can't seem to replicate the issue. So how about you help me. How the hell can we prove this to Apple without speculative temperature data? Because to Apple, third party temp. data = they have no idea how to reliably read the temp. sensors, even if that isn't true.
Download and run Cinebench, as im not convinced the 3.33 is any different to the 2.66
Just for everyones information I am getting a brand new 2.66Ghz tomorrow, seeing as the only difference between the 2.66 and the 3.33 is the CPU we'll soon see if they have "fixed it".
goodcow, do this please as what your saying contradicts every test on multiple Mac Pros now (i've tested 4 quads, and two octos!).
1. Download Cinebench.
2. Close and disconnect everything apart from Monitor and Mouse/KB.
3. Run Geekbench.
4. Open iTunes and start playing a playslist.
5. Run geekbench again.
6. Read off the results from the Console (bottom left).
Single Core rendering and the OpenGL tests show the biggest performance loss, multithread rendering doesn't take so much of a hit.
Thanks,
CR.
Thanks for the clarification goodcow. Glad that you are investigating the issue rather than dismissing it (as it could seem before).
As to temperature readings not being Apple certified hence dismissed, I think that misses the point (as in sounding like another Apple excuse for the brush-off). Yes the extreme high temperatures may not be accurate (in either direction) but they (and the power draw) do show a significant and unwarranted side effect of merely evoking audio (whether silent or not).
This thread is highly colored (IMHO) by ambiguity. Almost no one is precise in anything without assumptions. All factors have to be detailed. It is just this ambiguity and fogginess that gives Apple an excuse not to address the issue. To make realistic comparisons and reports we need to specify everything. Obviously MP model, MacOS version, number of disks, add on boards, sound cards, firewire & USB devices connected, ... The nature of any tests needs to be rigorously specified. Ie. make it so any bug effects are exactly reproducible by anyone with equivalent kit, and that those with different systems and setups realize why their tests won't give the same results.
Perhaps we need to cook up a template for everyone to use every time test results are reported, one which details everything.
There are too many occurrences of some newcomer with say a sound card in their system, merrily trying iTunes, watching the temperature, seeing insignificant change and announcing that the bug doesn't exist.
As to this thread being too large for anyone to read. I think we all agree. Hence my attempt to make a Facebook site in which to condense relevant information. My plan was to make some repository of just pithy facts about the bug (though efforts so far have been weak). However I feel this might be better done of a web site, though my objective was to make a space that anyone could participate in, be a part of, contribute to.
http://www.facebook.com/pages/2009-Mac-Pro-audiofirewirepower-bug/295274123618
Looking forward to hearing your results from Concorde's tests (and with condensed info on your system spec).
BTW, although not proven to be caused by the bug, smacman reports kernel panics in high temperatures, and I had a CPU fail. So if that is at the extremes, most of us many be inadvertently shortening the lives of our machines through this bug. That is a worry (well to me anyway).
I am getting so tired of people like you, who can't take the time to read through the thread, yet expect the rest of us to rehash the entire issue to you personally.
How many times do I have to say this. IT IS NOT ABOUT THE TEMPS!! The facts:
1) Playing audio or using FW sets the CPU into an abnormal high power state (Verified with a meter)
2) A byproduct of this high power state is abnormally high temps (Verified with an IR heat sensor)
3) Another byproduct of this high power state is reduced performance in certain multi threaded applications. This loss is not apparent in all applications, but is nonetheless easily reproduced. Cinebench is a good start.
So are you trying to say that because you can't verify that the actual temperature sensor values are 100% accurate, and because your single handbrake test was inconclusive, that we are all off our rockers complaining about nothing? People like you make me want to quit posting in these kind of threads altogether...
I did some tests with Cinebench.
Without iTunes:
Rendering (Single CPU): 3597
Rendering (Multiple CPU): 14139
Multiprocessor Speedup: 3.93
Shading (OpenGL Standard): 6367
After 15 minutes of iTunes, and playing during test:
CPU A Temperature Diode: 75C.
CPU Cores: 82-84C.
Rendering (Single CPU): 2653 -26%
Rendering (Multiple CPU): 13819 -2,2%
Multiprocessor Speedup: 5.21
Shading (OpenGL Standard) : 5074 -20%
Without iTunes:
Rendering (Single CPU): 3597
Rendering (Multiple CPU): 14139
Multiprocessor Speedup: 3.93
Shading (OpenGL Standard): 6367
After 15 minutes of iTunes, and playing during test:
CPU A Temperature Diode: 75C.
CPU Cores: 82-84C.
Rendering (Single CPU): 2653 -26%
Rendering (Multiple CPU): 13819 -2,2%
Multiprocessor Speedup: 5.21
Shading (OpenGL Standard) : 5074 -20%
As I said several pages back. If this was a f*cking iMac, iPod issue it would be fixed by now. Apple just doesn't seem to care anymore.
Just make sure AppleCare is paid for, and when the machine dies prematurely from "normal" use have Apple give you a new Mac Pro.
You hope!
I had one monster run in and hassle with AppleCare over my duff 2009 Mac Pro. After more than 3 weeks of repair attempts, Apple flat refused to replace it.
And believe me, having the better part of a month being cut off, having endless frustrating calls to Apple be directed to Indian call centers, being given the run-around and lies by dozens of Apple staff. Level two and customer relations staff consistently not phoning back when they promised to. Bungled repairs, bungled shipping, ... and finally a battered machine returned.
You really don't want to go there. And afaik the only viable way is to have access to a fallback machine of your own (they won't loan a machine). And expect a month's grief from Apple.
You'll totally lose any affection for anything Apple, even your will to live.
Trust me, you don't want a premature CPU failure and AppleCare can be an evil crock.
I thought I'd seen the worst of Apple over my G5 Quad roasting, but that was nothing to the torture of the 09MP debacle.
Besides your machine will die at day 1097.
Once bitten, twice shy.
Lol, it wasn't that positive, it's just I thought the capture data app was new and you and others should know about it. No, I'm not too optimistic that any of what I said will go anywhere. That I'll only get the same responses as you did. It's totally clear to me now that however decent a tech sounds over the phone, somewhere along the line they've been trained to deflect and avoid anything the least difficult. And to outright lie at the drop of the hat.
Perhaps you just caught my tone at finding an old enemy turned friend again.
Maximum temps? Yes, I'm willing to concede that when push comes to shove I don't know exactly what that refers to anymore, yes I was thinking cores. But I thought that xlr8 had come out against the 70C temp too. Just doesn't seem that hot. But I'll leave it to you guys to sort out. Just seems to bode really bad for me though.
Believe it or not I still think we all might be surprised with what 10.6.3 does. It would make sense in a way. Apple as we know won't admit anything to anyone, but that doesn't mean that in the background they are not working on a fix. So in Apples case they may be thinking, "Let's not say anything, let's just fix it and have it all go away". It would be the best of both worlds for them, no negative publicity they have to acknowledge, but the reasons for any would just go away quietly.
We'll see. As I said, if it's something easy for them to fix, they'll do it, no reason why not. But if it's something difficult, they won't. And I'll be using 10.6.3 to see which Apple thinks it is.
Take care, good going so far. Hope that Ars Technica thing was on the level.
Ok so i took my lunch time to do the following:
Rebooted my mac pro and left it idle for 5 minutes so the temps would stabilize. After 5 minutes idling temps were 36C for CPUA Heat Sensor Diode 32C for CPUB heat sensor diode and 27C ambient
Cinebench without iTunes
OpenGL Standard Test: 4673 CB-GFX
Single CPU Render Test: 3126 CB-CPU
Multiple CPU Render Test: 20921 CB-CPU
Multiprocessor Speedup: 6.69x
Temps remained between 36C and 41C
Right after that I fired up iTunes and played some mp3's for 15 minutes with the following results:
After 3 minutes of music there was an increase of over 10C (51C)
After 7 minutes the increase was almost 15C (55C)
Voltages for both CPUS were 1.16V while only playing music through iTunes.
Current for CPUA was 28.75A and for CPUB 28.56A on average
Power usage was for CPUA 33.38W and for CPUB 32.56W on average
After 15 minutes of music the temps sat at 62C - 65C for both CPUs
CPU Load never went above 1%
Fired up Cinebench again while still playing music with the following results.
OpenGL standard test: 3713 CB-GFX
Single CPU Render Test: 2280 CB-CPU
Multiple CPU Render Test: 18424 CB-CPU
multiprocessor speedup: 8.08x
Max temp reached during test: 68C
Total temp increase: 32C for CPUA and 36C for CPUB
Average performance loss: 20%
Come on!!! I didnt pay $3300 + tax for a machine that will take 20% performance loss over an mp3.... This means when I'm on a rendering job no music!!!.. bah.
Happens with a USB Audio device as well. Can't replicate the issue on Windows 7 Pro x64 under bootcamp.
Rebooted my mac pro and left it idle for 5 minutes so the temps would stabilize. After 5 minutes idling temps were 36C for CPUA Heat Sensor Diode 32C for CPUB heat sensor diode and 27C ambient
Cinebench without iTunes
OpenGL Standard Test: 4673 CB-GFX
Single CPU Render Test: 3126 CB-CPU
Multiple CPU Render Test: 20921 CB-CPU
Multiprocessor Speedup: 6.69x
Temps remained between 36C and 41C
Right after that I fired up iTunes and played some mp3's for 15 minutes with the following results:
After 3 minutes of music there was an increase of over 10C (51C)
After 7 minutes the increase was almost 15C (55C)
Voltages for both CPUS were 1.16V while only playing music through iTunes.
Current for CPUA was 28.75A and for CPUB 28.56A on average
Power usage was for CPUA 33.38W and for CPUB 32.56W on average
After 15 minutes of music the temps sat at 62C - 65C for both CPUs
CPU Load never went above 1%
Fired up Cinebench again while still playing music with the following results.
OpenGL standard test: 3713 CB-GFX
Single CPU Render Test: 2280 CB-CPU
Multiple CPU Render Test: 18424 CB-CPU
multiprocessor speedup: 8.08x
Max temp reached during test: 68C
Total temp increase: 32C for CPUA and 36C for CPUB
Average performance loss: 20%
Come on!!! I didnt pay $3300 + tax for a machine that will take 20% performance loss over an mp3.... This means when I'm on a rendering job no music!!!.. bah.
Happens with a USB Audio device as well. Can't replicate the issue on Windows 7 Pro x64 under bootcamp.
I agree with most of what you're saying but speaking as a software engineer, I think you're wrong when you call the Capture Data step a dead end. Yeah, it wasn't productive in your case but it's probably a standard part of the bug report process for the phone reps. If an engineer gets a bug report without this data, they may well bump it back as incomplete or take it even less seriously.
It sucks that whichever engineer received your bug report was too lazy to look into this properly but I think it's still *very* important for people to go ahead and get this data if a tech asks for it.
try my fix and report how performance is doing
http://groths.se/ABF.zip
beta test, you should have another OS to boot and remove the fix if something screwy happens
Forgive me for a somewhat simplistic post to what is obviously a very complicated issue, but I am not seeing any significant performance loss in Cinebench multi CPU render. With iTunes playing music in the background the score is 17,330, without iTunes the score is 18,069. This represents a 3.1% performance hit. A small performance hit would be expected while running another application while rendering.
There are obviously some people experiencing kernel panics due to heat issues, but as I posted some time ago I'm not seeing the high temps other people are seeing and my performance hit seems to be in line with iTunes CPU usage. I have a 3.33 GHz W3580 Xeon which I installed myself. My temps rise to around 55C and hold there as long as I'm playing music. That's nowhere close to what a few people here are reporting.
I'm seeing temps rise, but not to a point at which I'd be concerned. I'm also not seeing any noticeable performance drop except in benchmarks, and even then it's very small. I'm not at all minimizing what others are obviously experiencing, but this is a non-issue for my 2009 Mac Pro.
There are obviously some people experiencing kernel panics due to heat issues, but as I posted some time ago I'm not seeing the high temps other people are seeing and my performance hit seems to be in line with iTunes CPU usage. I have a 3.33 GHz W3580 Xeon which I installed myself. My temps rise to around 55C and hold there as long as I'm playing music. That's nowhere close to what a few people here are reporting.
I'm seeing temps rise, but not to a point at which I'd be concerned. I'm also not seeing any noticeable performance drop except in benchmarks, and even then it's very small. I'm not at all minimizing what others are obviously experiencing, but this is a non-issue for my 2009 Mac Pro.
I brought a Killowatt meter to work today to test the quad 3.33.
As soon as I play audio (or even select "line-in" in QuickTime X's screen capture mode) the wattage draw jumps from 190W to 240-250W.
I've sent a screen capture and an iPhone video showing the wattage change on the Killowatt meter to a senior level technician at Apple, however I believe he said he's usually out on Wednesdays and Thursdays (definitely Thursdays) so I probably won't hear anything back until Friday at the earliest.
Again, I had called in for a work related issue and once that was resolved casually mentioned this and he seemed genuinely interested in trying to help without me even having to prod him. I'm not replicating the performance issue though, at least not in Handbrake, but this is definitely not green.
As soon as I play audio (or even select "line-in" in QuickTime X's screen capture mode) the wattage draw jumps from 190W to 240-250W.
I've sent a screen capture and an iPhone video showing the wattage change on the Killowatt meter to a senior level technician at Apple, however I believe he said he's usually out on Wednesdays and Thursdays (definitely Thursdays) so I probably won't hear anything back until Friday at the earliest.
Again, I had called in for a work related issue and once that was resolved casually mentioned this and he seemed genuinely interested in trying to help without me even having to prod him. I'm not replicating the performance issue though, at least not in Handbrake, but this is definitely not green.
I saw approximately a 3% performance hit in Handbrake on an octo-2.26, however the power consumption definitely seems to be a huge issue, particularly on the quads.
Do you mind posting your entire test results (with and without audio playback)? If your performance loss is less than what everyone else is seeing, then maybe Cindori's theory that it is attributed to the lack of Turbo Boost due to high temps is true. Your temps have always been the lowest (I suspect because you actually applied your thermal paste correctly unlike Apple's staff that tend to squeeze it on like ketchup on a hot dog), and perhaps Turbo Boost still works for you even with audio playing. I'd still like to see your full results though.
I must say though, I am rather surprised to hear you say that this is all a non-issue. Even if your performance loss is lower, and your temps are "within limits", you still have a system that unnecessarily draws 100% load power when simply playing audio back. Back to the old analogy of buying a new sports car that consumes 15X more fuel when the stereo is on. Out of principle, that would be unacceptable. What some Apple customers will accept as a "non-issue", does not cease to amaze me. This is supposed to be a highly engineered, quality workstation.
After seeing this thread I too downloaded a 3rd party app "Temperature Monitor v4.8) and was getting 45 deg c average on cpu cores and up to 95 deg c after running almost any app (itunes, dvd player even safari) within about 3-5 minutes.
I contacted apple tech support who basically said, "you can't trust these 3rd party apps". He also said "if there is not excessive heat coming from the machine or it's working as it should, then all is ok..."
So what next? How can I get apple interested in this? I like the idea of being able to show a huge (unacceptable) increase in power usage, but is there an easy / "apple approved" way of doing that?
(I have a 4.1 MacPro, QuadCore, 2.93 GHz, 8GB Ram - and I'm no expert)
I contacted apple tech support who basically said, "you can't trust these 3rd party apps". He also said "if there is not excessive heat coming from the machine or it's working as it should, then all is ok..."
So what next? How can I get apple interested in this? I like the idea of being able to show a huge (unacceptable) increase in power usage, but is there an easy / "apple approved" way of doing that?
(I have a 4.1 MacPro, QuadCore, 2.93 GHz, 8GB Ram - and I'm no expert
)
Unless you have high ambient temperatures, 55C for iStat CPU temps when playing audio is abnormally high. Under normal conditions, iStat CPU temps should be less than 10-deg above ambient (eg. under 30deg. C). Keep in mind your core temps will be about 15-degC above what iStat indicates.
Download Hardware Monitor and graph a test from stable cold-boot idle temps to several minutes of audio playback and share the differences with us.
BTW, I've said this before, but any attempt to compare temps is meaningless in an Air Cooled system if ambient is not specified.
Yeah, here are my results with the Killowatt... (from page 13).
I also checked the temp increases were real with an IR thermometer.
This isn't the first time they've used this excuse. Granted, the actual value may not be accurate, but the rise is, as the variances subtract out. So if it says there's say a 25 - 30C rise, it's accurate to the temp sensor (say +/- 3C if it's a diode in the chip).
A simple "litmus test" per se, is stick your hand at the exhaust vent, and get an idea of what idle feels like. Then run the test, and put your hand back there again. If the numbers are so far off, your hand won't get much hotter. However, so far, those that have done this noticed a significant change.
But if you want accurate temps, you need a copper block with a K type thermocouple to be used with a Digital Multimeter with a Temp function (drill a small hole in the edge of the block to insert the thermocople, then place the block between the CPU and cooling tower), or an InfraRed Thermometer to obtain readings. The IR is easier, and cheaper if you don't have a meter with a temp function.
One of the members has used an IR Thermometer, and the temp increases where real.
By the way... why isn't this story on our very own Macrumors website? They are certainly talking about the 27inch iMac's screen problems and every burp of the iPad or iPhone... why no mention of this clearly universal 2009 Mac Pro flaw??? Has anyone tried talking to Arn or the other higher-ups?
A few weeks ago I tried to get the spotlight on this issue but nothing came of it. The fact that every second page or so somebody posts claiming their machine is not affect or that it's a non-issue probably does not help our cause...