Reverse Engineering the G5

isegwayisegway
Posted:
in Current Mac Hardware edited January 2014
I was thinking it would be nice to have a thread with any pics of the G5 internals or components we can find...



Supposedly this is the G5's daughterboard with what we assume is a heatpipe coming off of it into some big time heatsinks, found by Appleinsider member Dfiler:





Please post anything you've found here.
«1

Comments

  • Reply 1 of 23
    leonisleonis Posts: 3,427member
    Holy Sh!t?
  • Reply 2 of 23
    isegwayisegway Posts: 133member
    Another board pic..
  • Reply 3 of 23
    fawkesfawkes Posts: 80member
    Hmmm... Interesting. That heat pipe is connected to the transistors of what is likely the core power supply for the processor. Just a guess, but IR (International Rectifier) makes power transistors and those square things next to them are inductors. It is likely that there are more transistors on the back side to which that copper plate and heat pipe are attached (there isn't another BGA chip hidden there because we'd see vias on the back side--assuming they didn't use blind/buried vias as that'd be nasty expensive). So that would mean that the majority of the heat sink is dedicated to the processor and that little "side car" with the heat pipe is the sink for the switching power supply.



    Interestingly, it is possible then that the processor is directly connected to the heat sink, while the power supply requires the fluid heat pipe! The transistors on the primary side can get direct connection to the heat sink, while the ones on the secondary side require a heat pipe... very clever, indeed!
  • Reply 4 of 23
    ipeonipeon Posts: 1,122member
    Holy cow! Man, that's a serious heatsink. That will never fit inside a PowerBook. Nor the current iMac casing for that matter
  • Reply 5 of 23
    rageousrageous Posts: 2,170member
    The thing we need to remember about that heatsink is that it doesn't really need to be that big. Having the bigger heatsink means the overall piece is cooler, thus meaning the fans to cool it don't have to turn as fast to cool it. In turn you've got noise reduction.



    The G5 is hot, and there is no debating that. But that heatsink is without a doubt just a little overkill.
  • Reply 6 of 23
    powerdocpowerdoc Posts: 8,123member
    Quote:

    Originally posted by rageous

    The thing we need to remember about that heatsink is that it doesn't really need to be that big. Having the bigger heatsink means the overall piece is cooler, thus meaning the fans to cool it don't have to turn as fast to cool it. In turn you've got noise reduction.



    The G5 is hot, and there is no debating that. But that heatsink is without a doubt just a little overkill.



    Overkill for a 2 ghz, perhaps, but overkill for a 2,5 ghz on 130 nm process ?
  • Reply 7 of 23
    eugeneeugene Posts: 8,254member
    Quote:

    Originally posted by Powerdoc

    Overkill for a 2 ghz, perhaps, but overkill for a 2,5 ghz on 130 nm process ?



    If Apple would follow the same design philosophy as most x86 HSF designers, the heatsinks wouldn't have to be that big. With PC heatsinks, the fins or pins come out of a thick block right above the processor core. The fan blows down on that block while also blowing onto the fins. I guess Apple thinks it's better just to blow air past the fins and out the back, which keeps the case better ventilated, but the heatsink hotter...
  • Reply 8 of 23
    fawkesfawkes Posts: 80member
    Quote:

    Originally posted by Eugene

    If Apple would follow the same design philosophy as most x86 HSF designers, the heatsinks wouldn't have to be that big. With PC heatsinks, the fins or pins come out of a thick block right above the processor core. The fan blows down on that block while also blowing onto the fins. I guess Apple thinks it's better just to blow air past the fins and out the back, which keeps the case better ventilated, but the heatsink hotter...



    The PC method works quite well when you don't know what kind of ventilation the case will provide--a good guess, particularly in the home-built market! In these varied cases, you don't know where the airflow and hot-spots will be, so you need something to quickly pump the heat away from the processor and into the inside of the case. Then, you must rely on the case fans to keep up with the processor heat generation--usually a good bet in modern cases. The G5 heat sink would likely melt in a PC case since it can't be guaranteed to be sitting in a wind tunnel.



    Apple engineered the air-flow channels in the G5 to give the design legs for much faster and hotter processors. The channels are designed to move lots of low-velocity (read: quiet) air precisely to all of the parts that need it. Also, because the case is sectioned, you don't have processor heat interfering with the efficiency of, say, hard drive or chipset cooling--each section always gets fresh cool air.



    As designs get hotter, engineered cooling will become more important. The last card I designed (Gigabit ethernet to ATM port card) fits into a ATM switch rack. The rack is cooled with a tray of upward-blowing fans. If one of the cards is removed, the velocity of air at the remaining adjacent card is reduced, and it will shut down if the ambient temperature is high enough. We provide empty card blanks for customers who don't completely fill their racks.
  • Reply 9 of 23
    powerdocpowerdoc Posts: 8,123member
    Quote:

    Originally posted by Eugene

    If Apple would follow the same design philosophy as most x86 HSF designers, the heatsinks wouldn't have to be that big. With PC heatsinks, the fins or pins come out of a thick block right above the processor core. The fan blows down on that block while also blowing onto the fins. I guess Apple thinks it's better just to blow air past the fins and out the back, which keeps the case better ventilated, but the heatsink hotter...



    You will notice also that they designed their heatsink to allow the air to pass through it in a liminar way. This liminar air cooling had two advantages due to the lack of turbulences :

    - less noise

    - less dust (in surgical room this feature is essential, that's why there is a lot of liminar flow).

    The counterpart of this liminar design is the size : it has to be huge, otherwise turbulences will appear.
  • Reply 10 of 23
    matsumatsu Posts: 6,558member
    There's another pic from the side that shows two more heat pipes comming off the top of the CPU. Interesting stuff.
  • Reply 11 of 23
    mark_wilkinsmark_wilkins Posts: 154member
    Quote:

    Originally posted by Powerdoc

    liminar



    Do you mean "laminar?"



    Don't mean to be pedantic, just wondering if you meant something other than what I thought you did.



    -- Mark
  • Reply 12 of 23
    isegwayisegway Posts: 133member
    Quote:

    Originally posted by Matsu

    There's another pic from the side that shows two more heat pipes comming off the top of the CPU. Interesting stuff.



    Where is this pic? I figured there must have been some there. That huge heatsink wouldn't do much without the heat pipe pulling fluid out to the end.



    one thing though... heatpipes are most efficient in a vertical position.. I wonder how much efficiency they lose going out to the side here.
  • Reply 13 of 23
    mark_wilkinsmark_wilkins Posts: 154member
    Quote:

    Originally posted by iSegway

    heatpipes are most efficient in a vertical position.. I wonder how much efficiency they lose going out to the side here.



    Unless it's solid copper, in which case orientation doesn't matter. (I'm assuming the efficiency you're assuming is related to the convection of a fluid.)



    -- Mark
  • Reply 14 of 23
    isegwayisegway Posts: 133member
    Quote:

    (I'm assuming the efficiency you're assuming is related to the convection of a fluid.)





    Yeah... I just read that somewhere recently.



    Actually though... I don't know how it mounts... I could be wrong... it just seems like it goes out sideways for some reason.
  • Reply 15 of 23
    gregoriuspalitanaegregoriuspalitanae Posts: 7member
    Quote:

    Originally posted by Powerdoc

    You will notice also that they designed their heatsink to allow the air to pass through it in a liminar way. This liminar air cooling had two advantages due to the lack of turbulences :

    - less noise

    - less dust (in surgical room this feature is essential, that's why there is a lot of liminar flow).

    The counterpart of this liminar design is the size : it has to be huge, otherwise turbulences will appear.



    Not quite. The flow passing through the case will be very turbulent due to the fine mesh that the airstream enters through. It will remain turbulent when it encounters the fans and when it ultimately reaches the heat exchanger.



    This is not really a problem, though. For starters, at the velocities and scale we are talking about, turbulent flow over those heat exchangers is not a significant source of overall noise. (Much greater is anywhere you have flow separation, say the fans, case mesh, and even the flow tumbling off the edge of the heat exchangers. And to reduce this noise, low flow-velocities are the way forward. Hence the overall system design driver for low-velocity cooling flow.)



    And in the end, cooling designers actually prefer turbulent flow, since it has significantly greater heat exchange between a solid and the fluid. In fact, the mesh on the G5 case has most certainly been optimised to provide the best supply of turbulent air at the lowest noise and pressure drop.



    (See the redesigned xServe for more ideas on how holes - in that case circular versus polygonical - affect the airstream).



    GPX
  • Reply 16 of 23
    naplesxnaplesx Posts: 3,743member
    Very enlightening posts. We have some engineers in here, it appears.
  • Reply 17 of 23
    lundylundy Posts: 4,466member
    Quote:

    Originally posted by iSegway

    I was thinking it would be nice to have a thread with any pics of the G5 internals or components we can find...





    Please post anything you've found here.





    Supposedly, Apple has put the "Power Macintosh G5 Architectural Overview" lecture given at WWDC up on http://connect.apple.com for online Members (free). Look under the ADC TV link.



    I went to this talk in June and it is excellent. Explains the G5 architecture, the system buses, etc, and the cooling design.
  • Reply 18 of 23
    eugeneeugene Posts: 8,254member
    Quote:

    Originally posted by GregoriusPalitanae



    And in the end, cooling designers actually prefer turbulent flow, since it has significantly greater heat exchange between a solid and the fluid. In fact, the mesh on the G5 case has most certainly been optimised to provide the best supply of turbulent air at the lowest noise and pressure drop.



    Also noteworthy are the staggered pins on Alpha-Novatech and Swiftech heatsinks, I guess. (two of the best on the PC cooling side.) On the Alpha, the pins are squared instead of rounded. On Swiftech, the pins have ribs to increase the turbulence (and surface area.)
  • Reply 19 of 23
    wizard69wizard69 Posts: 13,377member
    You know guys I just had a rahter upsetting thought, we will not be able to mount these units side ways if heat pipes are involved in cooling the porcessor. Or at the very least we would have to be very carefull which way we orient the machine. I can't imagine this system working if the heat sink was upside down.



    Dave
  • Reply 20 of 23
    ast3r3xast3r3x Posts: 5,012member
    I bet it would run fine. I think the main purpose for the heat-sink is to let the heat be taken out by the fans not just dissipate it into the air inside the box. (but you know all this)



    The heat pipe should still work with it upside down but just not as well. The heat will still be transferred just not as much. I don't think its a big problem though.







    All this coming from my background and superior technical training of nothing.
Sign In or Register to comment.