MacBook Misconceptions about Heat, throttling, and cooling

Discussion in 'MacBook Pro' started by romesk90, Mar 28, 2014.

  1. romesk90 macrumors member

    Joined:
    Mar 28, 2014
    #1
    Hey guys,


    I see plenty of threads but none really pointing towards what I am looking for. I have also reviewed some spec documents but cannot find what I am looking for. By reading below, you will see this is not your typical "my mac is overheating! post." This is more of a "I don't know what, or why, my system is doing what it is doing" post.

    Bit of a long post, hope it makes sense. My writing is terrible, my apologies.

    In general I am wondering, what exactly are the terms for a macbook to "overheat", what is "overheating", what defines "turbo boost", what defines "fans being on". This all creates a bunch of unanswered questions for me and below is why. I understand there are multiple factors but it seems they are hidden. I have contact intel and was told because Apple can change certain specs, intel has no idea how Apple uses the CPU.

    I have gone to apple for answers but they are a bit clueless (the people I spoke to). I inquired about heat and was told "the macbook will shutdown before it gets too hot" when I asked how hot this was, they could not answer. When I said how hot my macbook got, they said it needs to go in for repairs right away. When I got it back they said it was normal. Aka, they have no idea what the norm is.


    Anyways, I have a 2012 MBP running latest mavericks. I have 8gb ram and an I7 2.9ghz turbo boost to 3.6ghz.


    What I test:

    I run Intel Power Gadget, Istat, Bresink temp monitor, to find out what is going on with my computer.

    While playing something intensive, such as Diablo 3, I notice the turbo boost is constantly running. That is, my CPU Frequency is above 2.9ghz. Now, my average temperature is about 100 degrees celcius and I hit 104. The DTS can only record up to the Tjmax which is 105 celcius. Intel recommends FULL cooling power before the Tjmax, and for an example they say 10. However, it seems apple is inconsistent with full cooling as I constantly get fans not running at full even when at 104 celcius. Sometimes they run at full tilt at 100 celcius. There is no correlation.

    Is turbo boost being activated improperly? After several reboots and running the game and monitoring, only sometimes will turbo boost not kick in, most of the time it is always on.

    A difference in processing? I have played with game settings, all on min, all at high. In both cases and everything in-between, I always get average 100 celcius, and turbo boost kicking in.

    What if I turn turbo boost off? With turbo boost off, I get a steady temperature that is 90-100 celcius. I have no lag or experience any set back from not having turbo boost in game. I can still run at full settings AND run cooler than with turbo boost on with the minimal settings in game.

    What is plimit? I keep getting plimit notifications in my console while testing my temperature. My understanding this is the power threshold when the CPU maxes out. It is just a notification and the OS decides what to do. I am unsure what this means exactly but thought I would include it.


    So, from the above, you can see I cannot make sense of what is going on.
    Why is my machine running just as good with turbo boost off, but yet runs cooler?

    Why are the fans not kicking in all the time when hot, or not running full throttle when borderline on the Tjmax.

    What exactly does apple do about the Tjmax? Some say the computer shuts down, but from what I can tell this is not the case. Since the DTS can only detect up to the Tjmax, I have no idea how hot my CPU actually gets. the CPU PECI sensor also seems to only report 104. My understanding is this can go higher but I am not sure how it is done (or even if my computer is hotter than that).


    Now, Apples warranty dictates no preventative maintenance needs to be done on the MacBooks. Also, when I inquired about thermal paste, apple told me it last forever and never needs to be changed. I have never heard of this. This leads to another question, how does one know if thermal paste needs to be replaced? Is there documentation that states apple knows it may need to be replaced? According to what I have been told, they only change the paste if the CPU has been taken off for some reason such as replacement. How does one know if there system is even cooling properly, if apple cannot tell you what cooling properly means.


    In summary my questions are:

    1) What truly happens as MacBooks CPU approaches Tjmax

    2) How can one know Tjmax was hit

    3) Can a macbook run near Tjmax 24/7 and not degrade (apple says it won't degrade)

    4) Is there any reliable way of knowing how or why turbo boost is kicking in

    5) Is there any reliable way of knowing how or why the fans are at the speed they are at

    6) can thermal paste last forever

    7) how would one know when to replace thermal paste

    8) how could one prove thermal paste needs to be replaced under warranty, if apple will not even look at that as an option

    9) How is it, turbo boost being off can run cooler and at the same quality, than with turbo boost on?

    10) Why is it, if having turbo boost on and off makes no performance difference, but yet it creates more heat, would turbo boost be activating when it does not to be?

    11) Why would turbo boost be on for extended periods of time when it does not to be? Again, i can run with it off and get the same performance but run cooler. So I could run with turbo boost on at 100-104 celcius fans blaring, or turn turbo boost off and get 90-100 celcius with moderate fans.

    12) In example 11, will either case make your laptop burn out quicker, or last longer?

    13) When will a macbook shutdown if too hot? How hot is too hot? How is this determined?

    14) Does anyone know someone whose macbook shut down because it was too hot? I have no heard or read of this yet.


    That is all, thank you.
     
  2. Cristian .b macrumors member

    Joined:
    Aug 16, 2013
    Location:
    Apple walled garden
    #2
    turbo boost starts if your mac is within the thermal limit (tjmax). macs shut down when they get too hot (over tjmax). i believe 105c is the limit when it shuts down. at least my 2007 macbook did at that temperature.and i dont think apple reapplies thermal paste at all. it wont disappear, it will harden and lose conductivity.
     
  3. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #3
    Thank you.

    However. Forgot to mention during my testing bresink would sometimes report --- as a temp. According to the developer this means either the core is asleep or its temp is higher than tjmax.

    I get --- once 104 celcius is hit. So either the computer sees that and sleeps it or its higher than 104.

    So. If the computer shuts down full at tjmax than the only way it would get there is a defect I'd have to think? But if it is defective could the computer not continue running too hot??

    I wish we had definite answers lol
     
  4. Gav Mack macrumors 68020

    Gav Mack

    Joined:
    Jun 15, 2008
    Location:
    Sagittarius A*
    #4
    Instead of focussing on turbo boost and how high it will clock and maximum temperatures a better way to think about it is the default idle temperatures of your CPU/GPU. It's pretty much the norm when I've re-pasted Macs for a 8-10C drop after proper and judicious application of AS-5 and polishing the contact plate, as Apple like many other manufacturers apply far, far too much. In fact with my 2011 17 a fortnight ago I had a 20C drop on both!

    With these lower idle temperatures you are not only helping keep your default temperatures lower, but maximising the thermal design power (TDP) of your system so you can run harder for longer without hitting that TDP barrier prematurely. I have seen first hand transcoding/render times lowered significantly on MBP, minis, iMacs and Mac Pro's to think otherwise.
     
  5. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #5
    Thank you for the info Gav Mack. I do understand this concept however, replacing the thermal paste is not an option for me as I am in warranty still and apple will not do the replacement. There AHT (apple hardware test) passes and therefore in their eyes everything is 100% okay even though it does not test for everything.

    I am also not really inquiring about the heat but yet the process. There seems to be a lot of different information going around (such as the Tjmax shutdown) that I would like to clear up once and for all for my own benefit and others if they are curious.
     
  6. Gav Mack macrumors 68020

    Gav Mack

    Joined:
    Jun 15, 2008
    Location:
    Sagittarius A*
    #6
    Sadly using solely the AHT to check if your system is running optimally doesn't take into account the shoddy application of paste and you are correct that Apple will not sanction repasting with AS-5 whilst still under warranty. I'm fortunate that I have a local repair agent who's engineers apply the standard paste with the optimal amount with good results, usually about 2/3rd less than 'factory'.

    The EFI has the temperature thresholds set and Apple optimise it for the particular system, but a large part of this is automatic and controlled partly by the intel chipset and most of all within the CPU which is a standard part. It's the same for Lenovo, HP and Dell using similar processors and GPU's so I would look at Intel's resources for more information.
     
  7. simon48 macrumors 65816

    simon48

    Joined:
    Sep 1, 2010
    #7
    Don't over think it. If it's getting hot it will throttle itself and it will shut itself down if it's way too hot.

    If you're asking out of curiosity have fun, but if this is out of concern for your computer I'd say you're over thinking it, your computer knows what it's doing.
     
  8. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #8
    Makes sense. Maybe the authorizes apple repair shop will look into it for me. I will check with them.
    I looked at the architecture of the CPU via Intels docs. I emailed Intel with further questions on things that were ambitious, they said to contact Apple who then said they have no clue :)



    It is a little column a, little column b. It is hard to trust something that does not make logical sense. In my example of being able to use my computer with no noticeable lag, without turbo boost, and run cooler, than with turbo boost on then I would question what is going on. I have no problem trusting something, as long as I either
    A) understand it
    or
    B) Have enough faith to trust it

    Both A and B are not there.
     
  9. matthewadams macrumors regular

    Joined:
    Dec 6, 2012
    #9
    Well, I disabled Turbo Boost on the OSX side of things completely.. the temperature/power vs. performance gain is just not worth it... and I sometimes got the impression that the CPU ramps up Turbo Boost even for tasks which do not need it... [1]

    Another observation relates to gaming.. I conducted a few tests on Windows with "ThrottleStop". I gave Battlefield 4 a few test runs on my 15" rmbp (I dont really do games but I decided to give it a shot to see what Iris Pro can do nowadays.. coming from the 2011 dgpu disaster I opted out of the dGPU for the rmbp).. Anyway.. I throttled the CPU TDP to 45W which
    a) left more power/temperature room for the GPU unit on the chip
    b) stopped my computer from draining additional power from the battery (a common thing when running intensive tasks on windows, I believe because there is no throttling on the OS level side of things, as opposed to OSX).
    c) gave me more FPS(!) since the GPU did not throttle itself down anymore due to the thermal output of the CPU

    As for OS X, I am not sure why theres an internal throttling (you can see it kicking in by running pmset -g thermlog in terminal) going on,.. I suspect it has something to do with the fan management or maybe Apple doesnt trust Intels on-cpu power/thermal management.
    It may also be a design decision to keep the fan noise down, seeing how quickly temperatures on the CPU rise and fall and how slow the fans ramp up..


    [1] In fact I've rewritten a kernel extension to do this. There was one out there, but it did not survive sleep/wake-cycles.
    I initially rewrote these for my previous 2011 MBP.
    https://dl.dropboxusercontent.com/u/2038446/Turbo Boost Switcher.zip
    https://dl.dropboxusercontent.com/u/2038446/turboboostsrc.zip (source)
    (Original unfixed projects: https://github.com/rugarciap/Turbo-Boost-Switcher / https://github.com/nanoant/DisableTurboBoost.kext )
     
  10. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #10
    You wrote that? dude..that is what I have been using to disable and enable turbo boost. Aha, wow.

    It works great man, thanks for making it. Saved me the trouble of compiling the code that was provided for it.

    Thanks for your response.

    As for the fan, I believe it to be just as you said...they chose the silent route. You can of course, use fan control programs to adjust when the fan kicks in...but I rather "trust" the OSX fan. Disable turbo boost seems to do an amazing job on its on for make the system more stable in terms of cooling management.

    Edit: thought id just say this too, your first line is pretty much what I am seeing. Turbo boost is just not worth it. My computer runs hotter for longer, and if I lose no performance with it off...why put my computer through that? Especially if I am borderline Tjmax...I want my mac to last as long as possible. Not to replace it in a year...
     
  11. matthewadams macrumors regular

    Joined:
    Dec 6, 2012
    #11
    Nah, I only implemented the sleep/wake-handlers into the kernel extension. Since they were missing, "Turbo Boost" got re-enabled every time you put your mac to sleep (and the switcher still showed it as "Disabled").

    Thats the only stuff I fixed ;)
     
  12. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #12
    Ah okay, I need to double check mine then...I probably have the original version. I will test it out later. Thanks for the links, if I indeed have the "broken" one, you just saved me the search trouble ;)
     
  13. Gav Mack macrumors 68020

    Gav Mack

    Joined:
    Jun 15, 2008
    Location:
    Sagittarius A*
    #13
    Intel's press won't have a clue.

    Good place to start is here:

    http://www.anandtech.com/print/4083...el-core-i7-2600k-i5-2500k-core-i3-2100-tested

    http://www.anandtech.com/show/4084/intels-sandy-bridge-upheaval-in-the-mobile-landscape

    http://www.anandtech.com/show/3922/intels-sandy-bridge-architecture-exposed
     
  14. ha1o2surfer, Mar 29, 2014
    Last edited: Mar 29, 2014

    ha1o2surfer macrumors 6502

    Joined:
    Sep 24, 2013
    #14
    If you are interested in Turbo Boost and Throttling on the windows side, use Intel Extreme Tuning utility, It will have a throttling percent graph which you might find useful.

    What you are seeing when playing game is an Intel Throttle state lowering the clock and voltage to cool the CPU down. This also happens when the dGPU is active because there is simply not enough power to supply both GPU and CPU under full load. When disabling turbo boost you are hurting performance greatly but also keeping the power consumption and heat from fluctuating causing more fluent frame rates. Turbo boost will kick in a lot when opening a program, loading a large webpage, extracting a file, installing a program, booting up and shutting down.. and so on. It will also stay full active even if the CPU is fully pegged on all 4 cores (if the cooling is sufficient)

    This can be seen in a luxmark benchmark if you choose both GPUs, HD4000/650M and the CPU. You will get a very low scored compared to a PC with similar specs due to temp and power throttling.

    In regards to heat difference with turbo boost, the cooling system may only be able to cool it properly under non turbo boost full load and the CPU is enabling turbo boost and deactivating it very rapidly causing higher temps. If performance is the same whether or not turbo boost is enabled is a major cooling system problem.

    Here are a few examples.. This is with my G46VW, a 14 inch laptop with a GTX660m. It includes a 180 watt adapter and uses around 140 watts while both CPU and GPU are fully loaded.

    This was taken with turbo boost disabled

    [​IMG]

    This was taken with turbo boost enabled

    [​IMG]

    A 2013 rMBP with a 3740QM and a 650M/HD4000 will get ~730. (my new rMBP) Those benchmarks were taken without even using any GPU and I'm only 200ish points off. That is disappointing to be honest but it's all due to thermal/power constraints. I never dipped below turbo boost even for 1 second and max temps are hovering around 83c. With just the CPU being fully loaded my wall adapter is saying I'm pulling 65 watts. If I was charging I'd go over 85watts easily. I even idle lower than a rMBP does at 9.8 watts so the laptop is super efficient.
     
  15. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #15
    Thank you for the insightful post.

    I do not have my computer boot camped, perhaps I will to try out those programs.

    You mention if performance is the same with turbo boost on or off, that that is a cooling issue.

    I notice with my laptop, once it gets hot it never really gets back down to idle temperatures. For example, I usually am around 50 Celsius when browsing the internet. I played diablo earlier and went to supper, my laptop is still at 60-70 Celsius stable (cpu is about 90-95% idle). Unless I turn it off, the temperature does not drop once it heats up. The laptop will hover at 80 Celsius for quite some time after playing a game or doing intensive stuff such as android development (opening up the android emulator shoots heat to 100 Celsius).
     
  16. ha1o2surfer, Mar 30, 2014
    Last edited: Apr 1, 2014

    ha1o2surfer macrumors 6502

    Joined:
    Sep 24, 2013
    #16
    What you are seeing are the consequences of using small heatsinks, small fans, conservative fan curves and an aluminum body. Aluminium tends to want to hold onto heat (or any temp for that matter) a little longer than plastic.

    It could also be a problem of OSX not disabling the dGPU after you closed diablo. This would cause the temps to be higher because Apple made the design decision to have both CPU and dGPU on the same heatsink. This causes a heatsink to not be used in the way they were designed. The thermal paste Apple uses also is applied in an ample amount causing heat to be trapped in the pockets that form when the heatsink is pressed onto the chip. This can cause heat build up that can't be carried away by the heatsink. As shown below.. All manufactures are guilty of this...

    [​IMG]

    Heat should be carried away from one end of the heat pipe to the other. When there are 2 different heat sources being carried away to different sections of the pipe it can cause the entire pipe to constantly be transferring heat to the cooling running chip and not to the fan. Heat will also want to get trapped in the location shown below..

    [​IMG]

    This is an 11 inch laptop with a 45 watt quad core and a GT 650m. This is one way you can use 1 fan and combine heatinks the correct way. One side is the "hot" side and the other is the "cool" side. This means 1 fan can cool both without throttling and won't heat up the other chip if it's not in use.

    [​IMG]


    This shows how a heat pipe works. Imagine two different heat sources trying to be cooled.. it isn't ideal IMO They probably did this ( in conjunction with not giving the mac a big enough power supply) because they didn't foresee a user putting a max load on both GPU and CPU in a gaming scenario. In their defense, they don't market a Mac as a gaming machine and it's not built to be one at all.

    [​IMG]

    I don't claim to be an expert but after studying heat sinks and modding heatsinks in laptops and desktops this design is frowned upon. There are also other issues with Apple's design such as the thickness and length of the pipe which can also lead to heat issues. I rate a heat sinks' ability to cool a component based on how long the "cool down" time is. That is; the time is takes to drop a temp from a fully loaded chip back down to idle temps after letting the computer get up to idle temp. For example.. My G46VW has a fully loaded CPU temp of 83c, once the load has disappeared, it takes 23 seconds for the temp to drop to 38-40c which is my idle temp. My Macbook takes over 2 minutes and the fan is still screaming after 30 seconds from when the full load has disappeared. This is not an official test but it's my way of testing.

    If I am wrong please state your facts because I love to learn.. one other thing, there is a reason Apple advertises CPU speed of their notebooks with the non turbo clock speed more often than the turbo boost speed.
     
  17. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #17
    ha1o2surfer, thank you for such a great post.
     
  18. ha1o2surfer macrumors 6502

    Joined:
    Sep 24, 2013
    #18
    Sure thing! I find it all so interesting and fun to learn about!

    Another thing I should mention that I will add. When I said "think about trying to cool 2 different heat sources" A heat pipe only carries heat in 1 direction. (as seen in the heatpipe stages photo" This will help understand what I meant by Apple not using the heatsink in a way it was designed to be used.
     
  19. ha1o2surfer macrumors 6502

    Joined:
    Sep 24, 2013
    #19
    If you didn't have heat/power throttling issues then a Mac should easily get to this score in Luxmark.. But I have yet to find a Macbook Pro (including mine) with similar specs get anywhere near this

    [​IMG]
     
  20. Epilorossa macrumors newbie

    Joined:
    Mar 29, 2014
    Location:
    Russia
    #20
    Hello!
    Faced with the problem:
    In my MacBook pro 15 late 2013 fans not works above ~ 2000 RPM. only help the programs which controls the fan speed.
    Are there benefits to fix it?
    SMC reset does not help.
     
  21. dmccloud macrumors 6502a

    dmccloud

    Joined:
    Sep 7, 2009
    Location:
    Anchorage, AK
    #21
    I just ran the OS X version of LuxMark on my rMBP and got the following result:

    [​IMG]
     
  22. nickandre21 macrumors 6502a

    Joined:
    Jun 21, 2012
    #22
    Well aht or asd will just tell then in a jist if the fan is speeding up properly or not. For a proper test they need to use apples cooling service diagnostics. I had it done on my retina i had no complains but the engineer wanted to check everything is fine before callling for parts. It first checks the sensors and fans. Then it will set a fan to a particular speed and ramp up it will measure the time it takes if its out of factory specs it will fail. Then it moves on to loading the cpu and seeing if the fans quickly get rid of the heat. it checks with various parameters if the conductivity between heat sink and chip is fine as well as checking if anything is using more power. This test can be found in apple service triage (ast) along with other tests. Asd is a stress test which can stress test memory and check few parameters same with aht.

    One engineer had agreed with me that apple does apply a bit to much of paste, even on the tech manuals the amount given is to much they cant do much but do as apple says. On my retina which went for a logic board change i somehow found idle temperatures dropped by around 5-7c and max stayed below 97c where as prior to that max used to be below 101c
     
  23. ha1o2surfer macrumors 6502

    Joined:
    Sep 24, 2013
    #23
    The 4960HQ is a much more powerful CPU than my 3840QM. The Iris Pro is also much more powerful GPU then my Intel HD 4000 yet I score higher.
     
  24. romesk90 thread starter macrumors member

    Joined:
    Mar 28, 2014
    #24
    Just a quick update:

    I went into Apple and they ran a ASD (apple service diagnostics).

    During the test, doing some CPU intensive ones would get my fans going at max but temperature never went above 75 Celsius.

    So,

    Does ASD throttle CPU? Are the readings from iStat, intel power gadget, and temperature monitor all wrong? There is a BIG difference. Outside of ASD I get temperatures of 100 Celsius easy. My fans do not go on max until closer to 100 Celsius.

    What do you guys think?

    The ASD was ran for several hours, it never got as hot as when I run OSX to test temperature. This makes me believe the settings for cooling are different, and/or ASD has some limits on power consumption.
     
  25. Merode macrumors 6502

    Merode

    Joined:
    Nov 5, 2013
    Location:
    Warszawa, Poland
    #25
    I have a question that might be interesting for the OP.

    There's a Blizzard game called Hearthstone. When I play it on my Late 2013 13" rMBP (2,4GHz, 8GB, 256GB), after a few minutes fan kicks in and the case is getting warmer and warmer. I haven't checked temparatures or how stressed the CPU is, nothing. It might be worth looking it up.

    Now, here's the funny thing. The very same game was just released for iPads - devices that have no fans whatsoever. They can't have fans kick in because they don't have them. Why would then my rMBP go ballistic with fan when I play the game? It is running a little loud.

    Isn't it strange to put it mildly?
     

Share This Page