MBP Heat Problem: From a thermal physics perspective

Discussion in 'Mac Pro' started by Rend It, May 9, 2006.

  1. Rend It macrumors 6502

    Joined:
    Oct 27, 2003
    Location:
    United States
    #1
    I am impressed at the bravery of some to take apart their MacBook Pros and handle the heat issue themselves. Even to go so far as to use an infrared thermometer to document the temperatures. However, there are some things that don't seem to jive with what's known regarding thermal transfer between materials. I'm starting this thread in hopes that we might try to come to a better understanding of what's really causing the high temps in the MBPs (besides the fact that there's a Core Duo in a confined space). [Disclaimer: I do not own a MBP, so I can't comment on whether I think they are hot. I'm just working on the assumption that enough people have complained to make it a legitimate problem.]

    First a question: Among those that have taken apart their MBPs, how many can document the temp change before and after? That is, have actually measured the temps at the processor or on the heat pipes (i.e., feeling warmer or cooler with your hands or lap doesn't count). The more documented info we have, the better.

    From what I've read in forums here and elsewhere, the success of that procedure seems to be 50-50. Some have improvements, others don't. This brings me to my most important point: Contrary to what everyone seems to say about applying thermal compound between an electronic device and a heatsink, it really doesn't matter if you use a large amount of compound.:eek: Flame me if you must, but the only reason to be sparing is because you would like the application to look nice, because the compound may interfere with nearby devices, or simply because using more than necessary is wasteful. i.e., using more should not reduce heat transfer. On the other hand, the pressure applied between heatsink and device is crucially important, as is the thermal conductivity of the compound. So, could it be possible that those who have replaced the thermal compound on their MBPs, and noted an improvement, are either inadvertently applying more pressure or are using better-quality compound?

    OK, now for why this is so (thermal physics of materials). If you take two reasonably flat surfaces, say even optically polished flat, and you lay one on top of the other, they will at most make contact at 3 distinct points. This follows from a consideration of the geometry involved, and the fact that only extremely rarely are materials atomically flat. For definiteness, suppose one of the objects is copper, and the other is a ceramic, like that used in processor packaging. The hot ceramic is only transferring its thermal energy at those 3 very small points. This is why the surface area of the two objects does not significantly affect the heat transfer characteristics. Note that at temps ~ 100 C, thermal transfer due to radiation is negligible compared to thermal conduction.

    Now, suppose we clamp the two objects together with some pressure. Copper is pretty soft, and so will more or less deform to give you additional points of contact. Also, as the pressure is increased, the points that are already in contact can transfer additional heat because those connections are more intimate. However, even with an absurd, unrealistic pressure, we can never bring the two objects to make contact along their entire surface. Here is where thermal interface materials enter. By applying a thermally-conductive, paste-like material, we can fill in all the microscopic gaps, and allow heat transfer across the whole surface. So, even if you put a glob of thermal coumpound on, as long as you mash the two pieces together (say, with the force of a single finger), the excess should be squeezed out, and the interface should be sufficient for good thermal transfer. My guess is that Apple may have messed up by not designing the heat transfer blocks to be parallel and in intimate contact with the processors, or the thermal compound is of poor quality.

    Sorry for the long diatribe. Any thoughts?

    -RendIt
     
  2. Kingsly macrumors 68040

    Kingsly

    #2
    I would say you are correct on all counts. Copy your post to an e-mail and send it to Apple, see what they think.
     
  3. Rend It thread starter macrumors 6502

    Joined:
    Oct 27, 2003
    Location:
    United States
    #3
    Thanks. I was beginning to think I said something egregiously wrong. Is there somewhere in particular I should send the comment, or just to Apple's general feedback website?

    -RendIt
     
  4. ZzBoG macrumors member

    Joined:
    May 7, 2006
    #4
    Rend It
    Looks like you know, what you're talking about. Are you involved in physical science of some kind?

    Can you give any scientifically approved advices for this cursed procedure (applying that paste, I mean)... Press it harder? Find the thermal characteristics of the paste?

    Resp.
     
  5. FireArse macrumors 6502a

    Joined:
    Oct 29, 2004
    #5
    ...

    isn't that all obvious?

    Didn't apple have postings for 'Thermal Engineers' about a year or so ago? Whatever they learned in their process to squeeze a G5 in the laptop shell - they'll have used to put these Core Duo's.

    They already know so much more. Something may have gone wrong in Manufacturing.

    F
     
  6. Rend It thread starter macrumors 6502

    Joined:
    Oct 27, 2003
    Location:
    United States
    #6
    It's obvious to you and me perhaps, but not necessarily everyone experiencing the problem. I guess I should have been more specific when I mentioned design...I think the heat pipe thing works, but only when properly implemented. I haven't taken apart my AlBook G4, but I imagine there's something similar. With the new Intel books, they probably switched manufacturers and maybe the new guys aren't as careful about the placement of the processors with respect to the heat plates.

    Then again, my G4 gets kind of warm, so maybe it's an engineering issue. Maybe there needs to be some sort of spring mechanism to account for variations in height and angle of the two interfaces. What's more, they're trying to make 3 chips (Core, Northbridge, and GPU) line up all at once! It's not an easy fix, and that's why people are unhappy. It could still be the paste, but more that it's an improper application, rather than there being too much. Those pastes are generally thixotropic, and therefore need to be worked into the materials before mating the surfaces. This is born out on arctic Silver's website.
     
  7. Rend It thread starter macrumors 6502

    Joined:
    Oct 27, 2003
    Location:
    United States
    #7
    I do low-temperature (liquid helium) physics. If you decide to re-apply the paste, I would just follow the instructions on Arctic Silver's website (they are actually very good). There are generally two kinds of thermal compound: electrically conductive and non-conductive. The electrically conductive stuff usually has silver or graphite particles in it, and is therefore much more thermally conductive (usually a factor of 10). But, the elect cond stuff also makes application more difficult because of the risk of shorts or stray capacitance. Whatever you get, be sure to work it into the surface really well. I usually use a razor blade to apply, but using your fingers (with gloves!) should work, too.

    Unless you can machine your own spacers, I don't recommend messing with the the pressure applied between the processor and heat plate. You could risk cracking the ceramic packaging.

    Good Luck!
     
  8. Kingsly macrumors 68040

    Kingsly

    #8
    Steve@apple.com

    It wont actually get to him, but an executive assistant will read it and forward it to the correct party.
     
  9. mmmcheese macrumors 6502a

    mmmcheese

    Joined:
    Feb 17, 2006
    #9
    This s basically what I've been thinking all along. All these comments about "too much thermal paste" have seemed to me like they were stumbling on the solution, but pinpointing the wrong cause. Only so much paste can stay between the pads and the chip given the amount of pressure that should be holding the chip to the pipes, so other than looking ugly, it should be fine. It would be far worse to have too little thermal paste under there (which I have seen many times).

    Also, wouldn't bad heat transfer between the chips and the heatpipe make the case cooler, rather than warmer?
     
  10. mmmcheese macrumors 6502a

    mmmcheese

    Joined:
    Feb 17, 2006
    #10
    to /dev/null
     
  11. AlmostThere macrumors 6502a

    #11
    Not a flame, but I think the general theory is that too much thermal paste makes it act like a heat sink itself. The only energy transferred is where the paste touches the actual heat sink. The rest of the paste heats up and can only lose that energy to the surrounding air in the case, which heats up the machine excessively.
     
  12. Rend It thread starter macrumors 6502

    Joined:
    Oct 27, 2003
    Location:
    United States
    #12
    Just because there's a lot of thermal paste, which would act like a heat sink, that still won't prevent heat flow normal to the surface. Heat transfer to stale air is very small. Thermal conductivity of air is less than 0.0001 times that of copper or aluminum. So, while it's true that there's a much greater cross-sectional area of air through which to transfer heat, it would have to be 10,000 times greater than that of the heat pipe, and it just isn't.

    See my forthcoming post, which I copied from a discussion board at Apple.com.

    -RendIt
     
  13. Rend It thread starter macrumors 6502

    Joined:
    Oct 27, 2003
    Location:
    United States
    #13
    Some qualifications...

    NB: After discussing this same topic over at Apple's website, I realized another detail to the problem. Please read on (the following is just a copy of my post over at Apple).


    link: http://discussions.apple.com/thread.jspa?threadID=475748&tstart=0

    During the day, I had time to think more about this whole issue, and I discovered another caveat. I realized it after I posted the thermal conductivities of various materials. When all of these thermal compounds are manufactured, there is usually some mixing step involved. During this step, a lot of air is incorporated into the compound, and remains trapped. In the lab, the way we solve this problem is by degassing the paste or epoxy in a vacuum chamber, after application (or, at least after we've done all the major handling). You can actually see bubbles emerge while it's in the chamber. However, since not everyone has a vacuum chamber to throw their MBP into, the next best step is to be sure you apply only very, very thin layers of the thermal compound (say 1 or 2 thousandths of an inch).

    The presence of air in the paste is such that it substantially reduces the effective thermal conductivity. So, my original statement needs to be qualified. Putting a glob of thermal paste on one surface and mashing it into the other will cause the air to remain trapped for a long time, since the only way for the air to escape is laterally across the surface. Air molecules will normally have a propensity to leave the compound, it's just that the viscosity of the pastes are so large that the motion outward is strongly hindered. With a thin enough layer, and a reasonable amount of time for the molecules to leave on their own, you should have a very uniform, and therefore high-quality film of paste.

    So, if you want to cool your MBP, first make sure that in the absence of thermal paste, there is good contact between heat pads and chips. There doesn't need to be like 50 psi between them, just make sure they are touching. e.g., you should NOT be able to slide a piece of copy machine paper in between them easily. Make sure the surfaces are clean. Use isopropyl alcohol without any water (get it from an electronics shop, not Target). If possible use gloves for everything once the surfaces are clean. Apply the thermal paste of your choice (it probably doesn't matter as long as it is electrically conductive), and apply it as outlined at the Arctic Silver website. If you can spare the time, leave the finished films to outgas for a few hours before mating the two surfaces. During this time be sure to cover the area with something to prevent dust from landing in the paste (make a cover out of aluminum foil).

    This whole issue is actually pretty easy to troubleshoot. I believe there are ways to access the temp sensors in the processor and on the heat pipes from within the OS. If there is a significant difference between processor and heat pipe temperatures, then you obviously have poor thermal contact at the heat plate. If someone can measure the cross-sectional area of the heat pipes, and the lengths between various chips and the fans, I can do a quick calculation to determine what the ballpark temps should be at various places. The fans are very important, because the rate of heat flow is proportional to the temperature gradient. If the fans don't keep the other end of the heat pipe cold, then the heat won't flow that fast, and therefore the temp at the other end necessarily rises.

    -Rend It
     
  14. bigandy macrumors G3

    bigandy

    Joined:
    Apr 30, 2004
    Location:
    Murka
    #14
    some very interesting points brought up there. something worthy of taking me away from my dissertation reading, which is currently two insanely boring reads...

    neural engineering, and theoretical neuroscience.

    yawn. :rolleyes:
     
  15. NewSc2 macrumors 65816

    Joined:
    Jun 4, 2005
    Location:
    Orange County, CA
    #15
    I'm not quite sure I agree with your conclusion.

    This isn't a new problem by any means, it's well documented EVERYWHERE that you're not supposed to use more than a very thin layer of thermal paste. I've built dozens of computers and every single Arctic Silver, thermal paste, Heatsink, and CPU installation guide I've come across emphasizes this fact.

    Additionally, I have yet to read a person claim that cleaning off the thermal paste and reapplying did not produce a very noticeable drop in temperature. Almost every one I've seen (at least 15) report that their temperatures have dropped by at least 10C, in most cases 15-20C.

    I'm not a thermal physicist (chemistry major here), so I can't comment on the physics perspective of your argument (very interesting, though), but I can say that I've had extensive experience putting together and overclocking PC's.

    I can't technically explain why putting even a few millimeters-thick application of thermal paste would insulate and not transfer heat, but I've seen it in action time and time again.
     
  16. Rend It thread starter macrumors 6502

    Joined:
    Oct 27, 2003
    Location:
    United States
    #16

    My point in the last post was that I think the reason it's important to use only a thin layer is because of the trapped air. If you could remove the air, then the thickness probably matters less. In my initial post, I overlooked this point. As far as success rates, I still see some who report no measurable change. It may be because they still aren't applying it correctly, or that they're experiencing a different problem. Or maybe their measurements don't involve infrared thermometers.

    In general, the best way to make good thermal contact (short of welding or brazing) is to clamp the two interfaces together with as much pressure as the materials can withstand. Putting thermal compound in between is the final tweak to improve heat transfer. It usually improves thermal conduction by a factor of 5. However, in a situation like processor heat sinks, you probably can't clamp too hard, or you'll damage something, so the paste becomes more critical. I think what's happening in the MBP is that it's not easy or straightforward just to apply more pressure. There's probably some fixed pressure that is provided by the design, they squirt a bunch of thermal paste on the heat pads, and when they assemble it, the pressure isn't sufficient to compress the paste and get the chip close enough to the pad.

    The net result will answer your last question/comment: The thermal conductivity of the paste is ~9 W/m-K. To get the thermal conduction of the entire interface, you need to multiply this number by the cross-sectional area, and divide by the length of the interface. So, for an area of 2.5 cm^2, and a gap of 0.001", you get a thermal conduction of about 90 W/K. Now, if the film is 2 millimeters thick (0.080"), we then have a thermal conduction of about 1 W/K. Thickness makes a big difference. You can think of the paste like a thermal resistor between the chip and pad. The thicker the paste, the bigger the resistance to heat flow. As I said, if you squeeze the two surfaces together (and leave them!), the paste should extrude out the sides, and you should be OK (again, neglecting the trapped air). But, I don't think the design of the MBP allows for that sort of pressure.

    -RendIt
     
  17. Hodapp macrumors 6502a

    Hodapp

    Joined:
    Jul 10, 2003
    Location:
    New York, NY
    #17
    What's amazing about this to me is that the other 50% of people are smart enough to be able to take apart and reassemble their MacBooks, but too retarded to properly apply thermal paste- which is a VERY basic concept.
     
  18. ddrueckhammer macrumors 65816

    ddrueckhammer

    Joined:
    Aug 8, 2004
    Location:
    America's Wang
    #18

    I am currently studing to be an Engineer, albiet Civil and not Electrical or Mechanical. Your reasoning sounds good from what I know of thermal conductivity. It sounds like Apple needs to either go to using some kind of thermal pad, as other suggested they used to use, or have a machine made that accurately applies the thermal paste. I personally wouldn't trust my companies reputation to a factory line worker on an issue like this. Perhaps they already have a machine that isn't calabrated correctly. Forgive my ignorance, but I know nothing of circuitry manufacturing. Most likely the pads would be the cheapest solution. Does anyone know how these work? I assume they have to be heated so that they make contact across the entire surface of the chip or they use some sort of conductive adhesive.
     
  19. Krevnik macrumors 68030

    Krevnik

    Joined:
    Sep 8, 2003
    #19
    Actually, with the grease I saw on my system, the layer was WAY too thick. Even if you are okay by pressing out the grease, they aren't doing that, so you have a layer many times thicker than recommended. Also, by not applying enough pressure, you don't get proper mating either, meaning that you can /reduce/ the amount of contact area.

    The heat never changes, really, the difference is what percentage goes where, and what is done about it once it gets there. Now, when this particular design is working correctly, the VAST majority of the heat should be dumped into the heat-pipe where it can be pulled away by fans if the pipe starts getting too hot itself. However, with poor mating, this doesn't happen. Instead, what I think is happening is that /not as much/ heat that is supposed to be in the heat-pipe is getting there, which keeps the pipe cooler than the fan turn-on temps. The rest of the heat is gotten rid of through other means (combination of ways), and produces the hot temp across the F-key bar and the like. Now, since the heatpipe never reaches the turn-on temp for the fans to go full blast because it isn't receiving the full percentage of the heat it is expecting from the chips... the heat isn't gotten rid of very quickly. If the heat isn't gotten rid of very quickly, then your system will remain like this, radiating/conducting heat through other case components, and the like.

    Correct the connection between the heat-pipe and the chips, a higher percentage of the heat goes through the heat-pipe, the thermal system becomes aware of the heat that needs to be removed faster, and so the fans turn on and remove it when needed. That is how you can get a cooler case by improving thermal transfer, considering the firmware doesn't turn the fans on if the keyboard or HDD is warm.
     
  20. Kreamy macrumors member

    Joined:
    Jan 2, 2005
    #20
    I'm not entirely familiar how the heatsink is actually attached to the logic board - but i would think that there'd be some form of washer to limit how much it can be tightened on so as not to crush the die?

    Taking this into play, your scenario of 'it doesn't matter how much is applied, it'll seep out of the sides' would be true - however, i think that for something as conductive as the silver or ceramic formula, the heat would be spread across the entire paste, and probably get trapped in the excess that lies on the sides of the die (effectively trapping the heat between the chip and the sink).

    Of course, this excess seeping out is inevitable so long as the process is performed by a human, but i believe that it's why we're instructed to keep the paste on the flat surface of the die and ensure that none is left on the sides, to reduce the excess.

    Ultimately, it is very important how much paste we apply even if the distance between the chip and heatsink surface is uniform. However, you're looking at the situation from a purely scientific viewpoint and not considering that the heat output difference between one correct method of applying thermal paste and another more correct method is quite negligible. But then again, I dropped science at 10th grade - so that's just my two cents.
     

Share This Page