15044 Views 1 2 3 4 5 Previous Next 63 Replies Latest reply: Feb 12, 2010 3:46 AM by Mike Hume Go to original post
Yes.. 2009 - 4,1 MacPro only.
CPU watt usage JUMPS significantly with video and audio streams. This was not observed with earlier MacPro revisions. Something could be broken in power management is osx other than just hardware Have not tested it with bootcamp. This is a noisy box compared to my silent 2006.
Hello to you all,
Sorry I've neglected to monitor this thread since posting, after my first reply I began to accept that the behaviour, although odd and not what I would expect, was "normal" for the 2009 machines.
Having returned to notice the reactions from many other people who feel this behaviour is a little out of the ordinary I'm again motivated to finding out if Apple themselves consider this to be a hardware or software problem.
In an attempt to progress things I visited my local premium reseller in Leeds, UK and demonstrated the problem using their own in store quad core 2009 machine. Upon viewing evidence in iStat which clearly showed their machine suffered the same temperature increases when playing audio they were initially confused. But then responded by saying that the temperatures were within normal limits for the CPU and that I shouldn't worry too much.
As a result Vidkidd I'm hoping you hear back soon from your local business rep regarding the reactions of Apple's own engineers to this odd behaviour?
Thanks again for everyones responses, I can take some comfort from the knowledge that I'm not the only one suffering these issues.
Granted heat is within operating limits, but seeing the CPU hit 65watts of use for simple audio playback contradict sales and marketing for the MacPro.
*Where it is called "the greenest MacPro ever.*
With this current behavior, the 2009 MacPro uses multiples more power to complete basic tasks than any other mac pro to date...
If this is a bug with the OS... Something should be patched.
As I found out and described in another thread, my Quad Nehalem with a RME HDSPe MADI audio card gets not hot if I select this card as audio out in Sys Prefs. But gets hot when selecting internal audio or RME Fireface out.
Not only might this be a problem for green aspects but also for "how long will my mac survive the runout of guarantee" since on my mac, the fans are not reacting on any temperature change or cpu stress at all.
Although I am quite "green" too, I prefer to have a machine that is working reliable. And not cycling down or even get kernel panics, when it's getting hot.
This week I've just come across a App called Geekbench that can be used to measure/rate the performance of your Mac. I'm not sure if there are rules against me linking to third party websites from these forums but it can be found here: http://www.primatelabs.ca/geekbench/
I downloaded and used this App to double check the behaviour of my Mac pro 2006 2.66Ghz Quad and its cpu temperature, under heavy cpu load instead of when playing audio. What I have observed is rather strange. On several occasions when running Geekbench I have noticed that the fans kick in really early and ramp up in reaction to the stress of the benchmark tasks. I know that this is what the fans are designed to do, but what I wasn't expecting was when I executed the benchmark routines in quick succession (even just twice) my fans just jump to full speed and won't slow down. I had to place the computer in sleep to stop them!
This behaviour is strange because its in direct contrast to what is observed when playing audio, where the temperatures increase with little or no reaction from the fans.
Firstly I would like to know what other peoples thoughts on this are?
...and secondly is there anyone reading this that has a 2009 Mac pro and are willing to run Geekbench to try see if their Mac Pro behaves like mine? I can assure people that once placed in sleep and woken up again my mac pro's fans return to there normally speeds again.
Just re-ran, PC hardly broke a sweat and cpu0 went from 33*C to 50*. The problem and difference is that a Bloomfield Core i7 running Windows 7, the temps fall off just as fast as they kicked in. As soon as cpu load drops off.
Nehalem processors will normally respond instantly to any action, and trigger a spike to the cores, to temperature, in response.
Perhaps Apple has yet to use the latest compiler tools and optimizations from Intel for their software. The behavior you and see didn't occur with the 2008 Penryn processor (or early Mac Pro).
Other tests, like CineBench10, Handbrake 64-bit, and you can get some ideas from Barefeats, there are some for Windows that can also stress test and benchmark a system.
"Geekbench 2 only measures processor and memory performance"
-MacPro 4,1 shows top scores in this test.
I just ran the 32 bit test and the 64 bit. Fan speeds hardly rose at all, just a few rpm. They stabilized plus or minus afterward.
Since I've used Hardware Monitor for many years, I wrote Dr. Marcel Bresink asking what his opinion of the heat risings in Nehalems, especially while playing songs in iTunes, might mean. It was the first he had heard about it, but with the info I sent him, he said he would look into it.
I wonder how aware Apple is of this situation?
ThHigh fan speeds have been linked to a corrupt or bad SmcFanControl plist...
Running Windows on Mac Pro always resulted in lack of fan speed control, and I don't know if that has changed or not with Boot Camp 3.0 now. Otherwise it is sometimes helpful or interesting to run tests and see what the results are, but not if cooling system isn't going to work or respond.
I asked someone I respect in the industry, whose apps I have used for a long time, including Hardware Monitor and others apps, what his opinion on this issue is. He's been my go-to man in the past for the final say or at least some really good insight. He is Dr. Marcel Bresink. http://www.bresink.com/info.html
"I would guess what you are experiencing is basically the combination of
Snow Leopard's advanced power saving techniques and Intel's Turbo Boost feature
which is present in all CPUs based on Core i7 technology.
If the total load on the system is low, Snow Leopard will first
- stop to schedule processes for the virtual Hyperthreading cores (cores 2, 4, 6, 8, 10, 12, 14, 16),
- stop to schedule processes for the secondary processor package (cores 9, 11, 13, 15),
- stop to schedule processes for the first processor package (cores 1, 3, 5, 7),
in this order.
When a processor core has nothing to do, the core, or the whole processor package,
respectively, will be powered down. A core without power will have a much lower
temperature than normal.
iTunes is an application which performs continuous real-time signal processing in
one thread of execution. If a system with Intel Core i7 processor detects the
situation "one core can be operated continuously, the other cores are idling",
it will activate Turbo Boost, which means it will overclock one core and
shut down the others. Overclocking will be done up to a degree where the
overall temperature of the processor package is still within acceptable limits.
The lower-than-average temperature readings during idle times, and the
higher-than-average readings during Turbo Boost result in the temperature
behavior you have observed when running iTunes on a system with low load.
The temperature readings themselves are uncritical. You bascially cannot overheat
an Intel procesor core, because it will automatically throttle its speed before
this could happen."
So I'm guessing that prior MPs to 09' were not Core i7, and Turbo Boost is not a feature prior to Nehalem either?
Message was edited by: Samsara
Thank you for sharing this gem!
So based on the wonderful explanation given by Dr. Bresink, this is all normal behavior for the 2009 machine, and shouldn't raise any alarms to the owners.
If anything, optimizing iTunes for the Nehalem machine is all that could be done.
What the explanation really does for me, is demonstrate how complex and robust the system actually is.
There is +a lot+ of stuff happening inside.....
To me it sounds very reasonable. But he mentions things I don't have a full understanding of. So I don't know yet if I should take it as the last word. I'd like to, especially that the whole issue seems normal and unlikely to damage a thing.
I'm going to do a little more research, but I wouldn't be surprised if he got this one right on the nosey.