Reverse Engineering the G5
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.
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.
Comments
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 G5 is hot, and there is no debating that. But that heatsink is without a doubt just a little overkill.
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 ?
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...
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!
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.
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.
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
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.
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
(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.
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
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.
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.)
Dave
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.