Mac Mini performance/noise/heat?

[Snowcake]

Originally Posted by MJL

The simple fact that the fan has been modified indicates that the 2011 models are running too hot. This is also confirmed by the higher number of logic board failures in the 2011 model (especially the mid model).

I just wonder if the 2012 fan has the same mounting layout and can be retrofitted to the 2010 and 2011 models and in that way improve the cooling performance of those models.

The Intel NUC was rumored to be released in October but have not seen it yet - I wonder if they have heat issues and what temperatures they'll be measuring. Habey has announced a fanless Ivy bridge computer.

http://www.xbitlabs.com/news/other/d..._Computer.html

How do you know that the Mac Mini has a different fan than the previous versons? Source?

Originally Posted by motrek

This thread is confusing me.

A lot of people talk about "throttling"--can you explain what you mean here? Do you mean the turbo boost decreases due to temperature? This seems like it would be almost imperceptible since when all four cores are loaded the amount the CPU can turbo boost is usually very small.

Throttling just means that the CPU speed wil lower with a big amount.
For example. A intel Pentium G860 (Sandy Bridge) Would downclock from 3ghz to 1.6ghz to save itself from overheating. So you have less performance.

Originally Posted by motrek

I suppose it's possible for the OS to change its process scheduling algorithm to eliminate the load on one or more cores depending on CPU temperature but I think it's extremely unlikely.

From your previous posts I infer that your "test" is running Handbrake, which is not designed to load the CPU to 100%, it's designed to encode video. At times it may use most (or all) of the CPU but at other times it doesn't. Certainly Handbrake doesn't use 100% of the CPU all the time on my own computer, which is definitely not thermally constrained.

If you really want to stress test your CPU I would suggest using software that's well understood to max out CPU usage, e.g., Prime95:

http://www.mersenne.org/freesoft/

Originally Posted by motrek

The more I think about this, the less likely I think it is that the Mini (or any Apple product) relies on the OS X process scheduler to keep it from overheating. You can run other operating systems on these computers (Windows, anyway) and certainly Windows is not designed to keep an Apple Mac Mini from overheating.

You might have seen Prime95 use less CPU after the temps went up but there must be another explanation. Are you looking at total CPU usage (i.e., green and red boxes) or just the Prime95 percentage? Because maybe the CPU was being shared with the OS or some other software.

Originally Posted by dasx

OK, ran it for 15 minutes. 8 threads. No problems at all.
Temp stable at around 98-100ºC (208-212ºF). No CPU throttle, all time at 100%. Fans at 5500rpm all the time. Incredibly loud, sooo noisy.

Ran Torture Test just for CPU. The way I used to when overclocking windows based machines.

How could i be so stupid for not seeing that!

It was never throttling. 80% just means it is using 80% of the cpu time/power. If you was using prime/cputest the cpu would be 100% because that is what is was designed to do. If for example your CPU would indeed throttle, the speed would lower for example to 1.4ghz, but prime/cputest would still use 100% of the cpu time/power.

If the cpu is at 2,3ghz, 2,6ghz or for example the throttlingspeed of 1.4ghz, cputest/prime would still use 100% cpu time/power. 100% just means to use all the processor processing speed. That can be at every speed. Doesn't matter if it is at 2.3 ghz or for example the throttling speed of 1.4ghz.

It was never throttling just handbrake acting "weird"

Is there no program to real-time monitor the mhz/ghz on your cpu on osx? that way you could be sure if it is throttling or not. Edit i found it. MSR tools. Link> http://answers.yahoo.com/question/in...6003600AAN7vW7

[motrek]

Originally Posted by Snowcake

It was never throttling. 80% just means it is using 80% of the cpu time/power. If you was using prime/cputest the cpu would be 100% because that is what is was designed to do. If for example your CPU would indeed throttle, the speed would lower for example to 1.4ghz, but prime/cputest would still use 100% of the cpu time/power.
...
It was never throttling just handbrake acting "weird"

Right, exactly. That percentage basically just tells you if the cores are idle, not how much actual computing they are doing or how fast they are running.

Handbrake isn't necessarily acting "weird," it just isn't using all the cores, for whatever reason.

Offhand, if I had to guess, Handbrake might split the video up into 8 segments to encode with 8 threads, and it's possible that some threads finish before others, resulting in idle threads part of the time.

It is common for software to not make use of all the threads possible. Actually the easiest way to program software is to do all the work with one thread and it's often pretty difficult to make it use more threads. You should be happy when you run across any program that runs several threads with high utilization in a useful way.

[dasx]

Originally Posted by Snowcake

How could i be so stupid for not seeing that!

It was never throttling. 80% just means it is using 80% of the cpu time/power. If you was using prime/cputest the cpu would be 100% because that is what is was designed to do. If for example your CPU would indeed throttle, the speed would lower for example to 1.4ghz, but prime/cputest would still use 100% of the cpu time/power.

If the cpu is at 2,3ghz, 2,6ghz or for example the throttlingspeed of 1.4ghz, cputest/prime would still use 100% cpu time/power. 100% just means to use all the processor processing speed. That can be at every speed. Doesn't matter if it is at 2.3 ghz or for example the throttling speed of 1.4ghz.

It was never throttling just handbrake acting "weird"

Is there no program to real-time monitor the mhz/ghz on your cpu on osx? that way you could be sure if it is throttling or not. Edit i found it. MSR tools. Link> http://answers.yahoo.com/question/in...6003600AAN7vW7

AAHHHHH!!! OK!!! Gotta try that then.

Will post results as soon as I try it out.

----------

Originally Posted by motrek

Right, exactly. That percentage basically just tells you if the cores are idle, not how much actual computing they are doing or how fast they are running.

Handbrake isn't necessarily acting "weird," it just isn't using all the cores, for whatever reason.

Offhand, if I had to guess, Handbrake might split the video up into 8 segments to encode with 8 threads, and it's possible that some threads finish before others, resulting in idle threads part of the time.

It is common for software to not make use of all the threads possible. Actually the easiest way to program software is to do all the work with one thread and it's often pretty difficult to make it use more threads. You should be happy when you run across any program that runs several threads with high utilization in a useful way.

I do usually make programs to use several threads, and the main problem I find is making sure every thread stays into his work and doesn't interfere with others'.

I don't know if you're into programming, a basic example of what I mean is the concurrent programming in Java.
If you create a class with an integer called 'count', and you then create 8 threads to increase that integer 10.000.000 each, it might end up not being 'count'=80.000.000 (8 threads) but something lower, due to some increments not being applied.

So yes, It's so nice to see a program that uses them all just right.

[motrek]

Originally Posted by dasx

...

I don't know if you're into programming, a basic example of what I mean is the concurrent programming in Java.
If you create a class with an integer called 'count', and you then create 8 threads to increase that integer 10.000.000 each, it might end up not being 'count'=80.000.000 (8 threads) but something lower, due to some increments not being applied.

So yes, It's so nice to see a program that uses them all just right.

Yes, I'm a professional developer, and multithreading is a pain. Actually I like designing software to run with multiple threads but it can be hard, especially if you need a lot of locks and synchronization. (Of course the idea is to design software in such a way that it requires the least number of locks... in the case of your count example, the way to do that would be to have an integer for each thread that gets incremented, and then add them all together at the end.

I'm probably going to be buying an i7 Mini soon. Looking forward to it. It will be my first 8 threaded machine. It's nice to see that Intel has made some major improvements to hyperthreading such that running 2 threads on one core often makes things 10-20% faster, or sometimes even faster than that. I remember the first Pentium 4s with hyperthreading and they would often run *slower* when running multiple threads, haha.

[dasx]

Originally Posted by motrek

Yes, I'm a professional developer, and multithreading is a pain. Actually I like designing software to run with multiple threads but it can be hard, especially if you need a lot of locks and synchronization. (Of course the idea is to design software in such a way that it requires the least number of locks... in the case of your count example, the way to do that would be to have an integer for each thread that gets incremented, and then add them all together at the end.

You solution is valid for my example, but imagine a counter that you need to be global and used by all threads.
Just go for semaphores or conditional wait. Monitors are the way to go, lol.

Originally Posted by motrek

I'm probably going to be buying an i7 Mini soon. Looking forward to it. It will be my first 8 threaded machine. It's nice to see that Intel has made some major improvements to hyperthreading such that running 2 threads on one core often makes things 10-20% faster, or sometimes even faster than that. I remember the first Pentium 4s with hyperthreading and they would often run *slower* when running multiple threads, haha.

When trying performance, what impressed me more was when trying some soft I do for work. It doubles the performance of a 2012 basic MBA. WOW.
(Here)

[Maxx Power]

There is an official Intel processor stability and thermal testing suite - Linpack.

There is a windows software called Intel Burn Test:

http://www.softpedia.com/get/System/...BurnTest.shtml

that one can use to test for stability. Usually runs a few degrees hotter than Prime95 and CPUtest (both of which are prime number algorithms) stress testing tools. Those of you with 2012 Minis should use this tool to test for absolute stability and throttling.

I think Linpack as a library is available for all platforms, but the Intel Burn Test GUI-driven front end software package that allows automated runs of Linpack is only available for Windows. Bootcamp it.

[dasx]

Originally Posted by Maxx Power

Windows. Bootcamp it.

BLASPHEMY!

Just kidding. I just won't install windows on my machine. I'd might consider installing it in an external drive but that's out of the table as Windows can't boot from an external hard disk in a Mac.

(In case you're wondering I decided to leave Windows 15 years ago and moved to Linux. Best choice ever. Even better than moving from Linux to Mac)

[Maxx Power]

Originally Posted by dasx

BLASPHEMY!

Just kidding. I just won't install windows on my machine. I'd might consider installing it in an external drive but that's out of the table as Windows can't boot from an external hard disk in a Mac.

(In case you're wondering I decided to leave Windows 15 years ago and moved to Linux. Best choice ever. Even better than moving from Linux to Mac)

Well, Windows is supposed to be a jack of all trades, at least it was before Win8. It is not very user friendly, or terribly reliable in the old days. I guess use it for what it is, and what your needs are until one day... Win8 = 8 too many for a lot of people.

I'm sure there is Linpack under linux, I could be wrong on this, but you MAY need a GUI front end to facilitate the process.

[dasx]

Originally Posted by Maxx Power

Well, Windows is supposed to be a jack of all trades, at least it was before Win8. It is not very user friendly, or terribly reliable in the old days. I guess use it for what it is, and what your needs are until one day... Win8 = 8 too many for a lot of people.

I'm sure there is Linpack under linux, I could be wrong on this, but you MAY need a GUI front end to facilitate the process.

I understand Windows is very important for most (seeing worldwide stats at least). But thank to what I do I can choose and I choose not to use it ever again.

As for the topic, I don't need that much to check my CPU. It works and has no errors under Prime. Handbrakes fine and all.

All I want is to make it a little cooler, that's why I plan to mod a Mac Mini bottom cover I bought off ebay.

[motrek]

Originally Posted by Maxx Power

There is an official Intel processor stability and thermal testing suite - Linpack.

...

Linpack is just short for linear algebra package--it's a library of functions used for linear algebra, which has all sorts of uses in scientific computing, and it runs on any platform. It's the de-facto benchmark to test supercomputer performance. When somebody says a supercomputer does X mega/tera/peta FLOPs, that's when running Linpack.

Anyway I'm sure Linpack runs on OS X but whether or not there's a nice front-end for it so it can be used as a CPU stress-test/benchmark, that's another story. Shouldn't be too hard for a developer to whip one up though.

[Maxx Power]

Originally Posted by motrek

Linpack is just short for linear algebra package--it's a library of functions used for linear algebra, which has all sorts of uses in scientific computing, and it runs on any platform. It's the de-facto benchmark to test supercomputer performance. When somebody says a supercomputer does X mega/tera/peta FLOPs, that's when running Linpack.

Anyway I'm sure Linpack runs on OS X but whether or not there's a nice front-end for it so it can be used as a CPU stress-test/benchmark, that's another story. Shouldn't be too hard for a developer to whip one up though.

I read it somewhere a long time ago that Intel specifically uses Linpack for stability testing. That's why I suggested this, since all their CPUs (if I remember correctly) should all pass Linpack (which for x86 chips there is the IntelBurnTest named after Intel with specifically tailored and compiled libraries to maximize load on x86 and x64) for many hours on end.


As for super-computing, I know all super-computers are Linpack tested, but I thought Lapack is used for newer rigs instead ?

[motrek]

Originally Posted by Maxx Power

I read it somewhere a long time ago that Intel specifically uses Linpack for stability testing. That's why I suggested this, since all their CPUs (if I remember correctly) should all pass Linpack (which for x86 chips there is the IntelBurnTest named after Intel with specifically tailored and compiled libraries to maximize load on x86 and x64) for many hours on end.


As for super-computing, I know all super-computers are Linpack tested, but I thought Lapack is used for newer rigs instead ?

Sorry, dunno much about supercomputers and supercomputing other than some trivia. I just looked up Lapack on Wikipedia and that does seem to be the case though.

[Maxx Power]

Originally Posted by motrek

Sorry, dunno much about supercomputers and supercomputing other than some trivia. I just looked up Lapack on Wikipedia and that does seem to be the case though.

Me neither, just fascinating stuff I read. I love the tech and the ideas behind the tech.

[motrek]

Just bought my new Mini (2.3ghz i7) this evening and ran some tests.

Ambient temperature where I'm testing is on the cool side, between 60-65F, so keep that in mind.

Running some of my own software that maxes CPU load, I saw a maximum CPU temperature of 96C (reached after about 8 minutes) and a maximum fan speed of 2450 RPM (slightly louder than the normal 1800 RPM but you have to concentrate to hear the difference). The software was running as fast at the end of the test as at the beginning--i.e., no throttling. The Mini got noticeably warmer towards the rear after a while, but certainly not 'hot' and just barely what I would call 'warm.'

I admit that I find any CPU temperature over 80C disconcerting, based on decades of building my own computers, but I do trust that Apple engineers have done their due diligence and made sure that all the components operate within their specs and tolerances.

I was half expecting to be disappointed with the Mini since it looks like it has laptop cooling hardware in the teardowns, but I don't think this is the case. You can read plenty of complaints on the internet from rMBP owners complaining about fans running at near-full speed while under only moderate load, whereas my Mini's fans don't run anywhere near full speed even when under full load. So, I'm very pleased with my new computer.

PS: I have seen the "black screen" problem twice after only using the computer for an hour or two. Annoying, but I expect an OS update will fix this... otherwise I will spend the $7 to get a Thunderbolt-to-DVI adapter.

[Yebubbleman]

Originally Posted by motrek

Looking at the teardown of the 2012 Mini, the heatsink and fan look like they are laptop-sized.

Does anybody know if this is enough to cool the 2.3GHz i7 model without generating a bunch of noise and/or limiting the performance, since the Turbo Boost feature overclocks less when the processor is hotter?

I'm a little worried because I will be doing some very processor-intensive stuff on the Mini at times, and I've read about how people are annoyed by the amount of noise put out by quad-core MacBooks under load.

I had a Mac mini Server (Mid 2010) that I used as a non-server Mac mini (I wiped the drives, and installed the client version of Snow Leopard and used it as a normal Mac). Occasionally, I'd play StarCraft II and Civilization V. For the most part, these were the most system intensive apps that I used on it. Mind you, this is not a quad-core Ivy Bridge Core i7, but rather a dual-core Penryn Core 2 Duo. Similarly, I wasn't using the Intel HD 4000, but rather the NVIDA GeForce 320M (Unlike the Intel HD 3000, the Intel HD 4000 is an evolutionary degree faster than the GeForce 320M whereas, the HD 3000 is supposedly comparable if not slightly weaker). On StarCraft II, I noticed no major fan activity. With Civ 5, on the other hand, I did. The machine was very warm to the touch, but the fan noises were still fairly minimal for being substantially louder than usual. It's nothing compared to my gaming PC tower. Long story short, I wouldn't worry; it might make some noise and be a bit loud at times (but only relative to it being quiet), but it's nothing to worry about.

[dasx]

Originally Posted by motrek

Ambient temperature where I'm testing is on the cool side, between 60-65F, so keep that in mind.

Do you usually work at those temps?
I'd like to see what my "modded mini" does at those ambient temps. I just don't wanna get a cold.

[entzoe05]

Originally Posted by nsfw

I tag along on this thread.. I have the 2011 i5 with 6630M/2 drives and the fan spin up is pretty maddening. I don't play games much so I sort of regret not getting the 2.3.
Anyways, wanted to see if any 2011 owners got a 2012 as well. Is the cooling better? I suspect the intel4000 runs cooler than the 6630m.

I'm experiencing this loud fan issue with Mac Mini 2012. Got this last week. Noticed the following; 1. Copying files 20GG 2. Rendering 3mins movies from iMovie. Fan would go as high as 5500rpm.

So I went to a local BBY store to check their mac mini demo display. Asked the sales guy if I can install iStat pro so I can check the fan rpm and cpu heat reading. I ran the iMovie, and export a 15mins movie. Attached image is the screen shot of the CPU and FAN reading.

The only difference between the mac mini is;
Mine: i7 2.6 Ghz, 16GB RAM, 1TB Fusion
BBY: i7 2.3 Ghz, 4GB RAM, 1TB (stock)

What is this telling?

[philipma1957]

Originally Posted by entzoe05

I'm experiencing this loud fan issue with Mac Mini 2012. Got this last week. Noticed the following; 1. Copying files 20GG 2. Rendering 3mins movies from iMovie. Fan would go as high as 5500rpm.

So I went to a local BBY store to check their mac mini demo display. Asked the sales guy if I can install iStat pro so I can check the fan rpm and cpu heat reading. I ran the iMovie, and export a 15mins movie. Attached image is the screen shot of the CPU and FAN reading.

The only difference between the mac mini is;
Mine: i7 2.6 Ghz, 16GB RAM, 1TB Fusion
BBY: i7 2.3 Ghz, 4GB RAM, 1TB (stock)

What is this telling?

HALF the story my guess is the photo is from the best buy computer. so where is the photo from your computer BTW the cpu is 92c with only 25% of the cpu in use that is a lite cpu load. all I can say is show us a photo of your computer running the same stuff. it should be a little hotter. why your cpu is hotter and you have 2 drives not one. fusion is 2 drives more power. oh you have more ram more power used. so your machine should pull 5 watts more then the best buy machine ,5 watts is a guess but close at least 3 maybe 7 more.