Posts:
149
Registered:
Apr 12, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 10, 2006 11:19 PM
in response to: RendIt
|
|
|
There is one thing that I am increasing confused about which (from what I can find) nobody has addressed as of yet.
I was under the assumption that the MBP's entire chasis is supposed to act as a large heatsink. Sure the fans are supposed to shoot some air out, but for some reason I had the idea that the case itself was also supposed to transfer some heat.
It just seems to me that if the case is now magically cooler after the cleanup/regrease - doesnt that actually mean that the CPU is being cooled LESS efficiently?
MacBook Pro
Mac OS X (10.4.6)
|
|
Posts:
633
Registered:
Mar 3, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 10, 2006 11:27 PM
in response to: dmcbride
|
|
|
I was under the impression that the lower case is the heat sink - not the entire case...
MacBook Pro 2.0GHz Intel Core Duo
Mac OS X (10.4.6)
My God, its Full of Stars !
|
|
Posts:
15
Registered:
May 10, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 10, 2006 11:45 PM
in response to: syntrak
|
|
|
1) I completely agree.
2) I'm not sure yet how the torque affects the mating. From other forums I'm involved in, it seems that both the heat pads and the chip boards (the boards that support the CPU, GPU and Northbridge) are rigidly fixed in their respective places. So, there isn't a large amount of "give" built into the design. This means that the quality control of the boards and heat pipe assembly have to be very high, if the mating is to be consistent from unit to unit. There's a post above where it's shown how to determine if improper mating/alignment is the problem.
3) Assuming the mating is OK, as outlined above, then a proper application of thermal paste should lower the temps that the user experiences on the case. Ideally, thermal paste should NEVER harden. If it does, then only after a few thermal cycles you would end up with a poor thermal connection to the heat pads. When shopping for paste for this application, make sure it's thermal conductivity is on the order of 1.0 W/m-K, and make sure it can withstand regular excursions to 125 C or more. The combination of air (both in the paste and in the surroundings) and high temps (above 175 C) will eventually cause any paste to harden.
At work, I use a product from NuSil called R-2940. It's thermal conductivity is as good as any of the non-Arctic Silver pastes, and you don't have to worry about shorting or stray capacitance, as it's insulating. The best thing about it is that it's rated to 240 C! So, it would never harden, and it would provide a reliable contact for life.
-RendIt
PowerBook G4
Mac OS X (10.4.6)
1.25GHz AlBook
|
|
Posts:
15
Registered:
May 10, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 2:30 AM
in response to: Daffy_Duck
|
|
|
You are neglecting the fact that nobody would use paste for a heat pipe. In their respective applications, both the copper pipe and the thermal paste would probably transfer heat almost equally well. Even if we assume that the area of the chip is equal to the heat pipe (the chip is actually a larger area), the actual conduction in the two would be:
heat pipe: 2 W/K
cheap thermal paste: 10 W/K
The heat pipe is several cm long, and the thermal compund is only about 0.003 cm thick. To get the actual heat flow, just multiply above by the temperature gradient across each item.
PowerBook G4
Mac OS X (10.4.6)
1.25GHz AlBook
|
|
Posts:
231
Registered:
Feb 11, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 3:18 AM
in response to: RendIt
|
|
|
My point is that you can't just focus on conduction because inside the heat pipe, other heat transfer methods come into play.
Here's a quickie from Google:
http://www.cheresources.com/htpipes.shtml
"Heat pipes have an effective thermal conductivity many thousands of times that of copper."
So, due to convection and evaporation, the heat pipe has a higher conductivity than the copper itself. The paste is the weak link here.
By the way, can you tell me where the 2 W/K comes from?
MacBook Pro 2.0GHz / 7200RPM
Mac OS X (10.4.6)
Also have AMD desktop with Windows XP
|
|
Posts:
30
Registered:
Mar 1, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 6:22 AM
in response to: RendIt
|
|
|
Is anybody else who's in Italy/Europe have trouble getting hold of Artic silver thermal paste ?
Ive been told there is a 3 week wait here in Italy from essedi & bow. Maby something to do with sudden demand from owners with dodgy MBP's
Can any body recommend a good alternative ?
MacBook Pro
Mac OS X (10.4.6)
Win XP Pro
|
|
Posts:
231
Registered:
Feb 11, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 8:02 AM
in response to: RendIt
|
|
|
I want to illustrate the effect of thickening the layer of paste between the interface with some actual math. I know I'm probably beating a dead horse but someone might be interested in the mathematical proof.
Thermal conductance is the measure of how much heat passes through a material per unit of time. It's measured in W/K, where W = Watt and K = Kelvin
Thermal conductance = kA/L
Where:
k is the thermal conductivity (W/mK)
A is the cross sectional area (m^2)
L is the thickness (m)
I'll do 2 calculations the first assuming a thickness, A, of .003" (.0000762 m) which is roughly the thickness of a piece of paper. I will use a k value of 1 W/mK. That's the thermal conductivity of basic thermal paste. I will use a surface area of 2in^2 (.0013m^2).
Tc = kA/L = (1 * .0013) / .0000762 = 17.06 W/K
Tc = 17.06 W/K
---------------
The next one assumes the thickness of the paste is the rough equivalent of 3 sheets of paper or .009" (.0002286 m)
Tc = kA/L = (1 * .0013) / .0002286 = 5.69 W/K
Tc = 5.69 W/K
--------------
So, by tripling the thickness of the layer of paste, the AMOUNT of heat that flows through the paste per unit of time decreases from 17 W/K to 5.7 W/K. That means it will transfer about 1/3 as much heat.
MacBook Pro 2.0GHz / 7200RPM
Mac OS X (10.4.6)
Also have AMD desktop with Windows XP
|
|
Posts:
231
Registered:
Feb 11, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 8:08 AM
in response to: Daffy_Duck
|
|
|
Now watch this.
What if we spread a paper thin layer of Arctic Silver 5 which has a thermal conductivity rated at 7.5 W/mK? Remember, in the above equations I used basic paste which has a thermal conductivity of approximately 1 W/mK.
Tc = kA/L = (7.5 * .0013) / .0000762 = 128 W/K
Tc = 128 W/K !!
-------------
Quite an increase from 17.6 isn't it?
By the way, if you were to replace the paste with a 3/4" thick piece of plywood, Tc would be about .0089 W/m. In case you were wondering like I was. On second thought, it would actually be lower than that because there would be more air to deal with in the plywood to metal interface.
MacBook Pro 2.0GHz / 7200RPM
Mac OS X (10.4.6)
Also have AMD desktop with Windows XP
|
|
Posts:
43
Registered:
Feb 16, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 9:41 AM
in response to: syntrak
|
|
|
The lower case contains a heat pipe, whose purpose is to route/transfer heat from one place to another. Heat from the dies is transfered to the heat sinks on the heat pipe, which then transfers it to the heat vents on the left and right sides. The fans blow air across the heat vents, transfering the heat to the air and out of the case.
Powerbooks
|
|
Posts:
633
Registered:
Mar 3, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 10:22 AM
in response to: RendIt
|
|
|
Rendit - Your posts are great ! Please contact me by email - phil.angell@mac.com
MacBook Pro 2.0GHz Intel Core Duo
Mac OS X (10.4.6)
My God, its Full of Stars !
|
|
Posts:
782
Registered:
Jan 24, 2003
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 10:34 AM
in response to: dmcbride
|
|
|
dmcbride - We do not think it means the cpu is being cooled less efficiently. As explained by Daffy and migs - the proper application of thermal paste allows more heat to flow from the cpu to theheatsink to the heat pipe where it is moved OUTSIDE the case via the cooling fan - exactly as it was designed and supposed to work. The case is cooler because more of the heat is removed via the heat pipe and fan and their is less reliance on the case as a radiator.
At least that is the current trend in thinking.
Quad G5, FPiMac 1.25, Mac Mini 1.66, Ti PB 400, Indigo Clam 366, iPod Video
Mac OS X (10.4.6)
Canon 20D DSLR, i9900
|
|
Posts:
15
Registered:
May 10, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 11:40 AM
in response to: Daffy_Duck
|
|
|
OK, but the other heat transfer methods only come into play inside the heat pipe. So, it's still heat conduction down the pipe, albeit with a different effective thermal conductivity. And if I understand that webpage correctly, if they are using capillarity to effect the transfer of heat, then there is no convection involved. They don''t even mention the word on their webpage. Recall that there are two types of convection: forced and natural. Forced convection is what happens at the fan end of the MBP. It works pretty well. Natural convection requires large temperature gradients and an open structure through which to set up flow (think about baseboard heaters- they work best in a large empty room). It works moderately well, once you've satisfied the prerequisites.
My point in that calculation is that for most purposes, a regular silicone-based thermal grease will do an OK job for the power levels involved. A proper application will reduce the high temps people are experiencing. Now, if the processor were putting out something like 100 W, then maybe the particular choice of thermal paste matters a little more. The design of the laptop is to take heat away from the chips, put it onto the pipe and take it to the fan, which then releases it to the air. If at any point, something fails in the chain, then you have a problem. The design should not rely on other forms of heat transfer.
The 2 W/K just assumes a 2.5 cm^2 cross section for the pipes, and a few centimeters length. Of course if the heat pipe isn't just copper, then it's a different number.
PowerBook G4
Mac OS X (10.4.6)
1.25GHz AlBook
|
|
Posts:
231
Registered:
Feb 11, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 12:43 PM
in response to: RendIt
|
|
|
"if they are using capillarity to effect the transfer of heat, then there is no convection involved."
When a liquid moves across a surface of a higher temperature, whether by capillary action or any other movement, there is in fact heat transfer by convection. In heat transfer, a liquid is treated much like air but with a different conductivity and density. You mentioned natural convection. That's what's happening in the heat pipe. Except it takes advantage of a phase change. It's still convecting heat from the copper surface near the chip. The reason for that is that at the surface, the liquid is essentially boiling. The molecules in the liquid are moving very rapidly around on the surface at this point. Because it's moving, it's exposing more of the atoms in the liquid to the higher energy level of the copper more quickly than if the liquid were stationary (conduction transfer). That's what convection is all about. Then, the liquid in it's gas state is carried to the other end of the heat pipe and it's heat is lost to the metal, again, largely through convection. The fan removes the heat from the metal, again largely through convection.
"The 2 W/K just assumes a 2.5 cm^2 cross section for the pipes, and a few centimeters length. Of course if the heat pipe isn't just copper, then it's a different number."
I'm still confused about where that number comes from. Perhaps an equation or link to somewhere would help.
MacBook Pro 2.0GHz / 7200RPM
Mac OS X (10.4.6)
Also have AMD desktop with Windows XP
|
|
Posts:
15
Registered:
May 10, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 2:05 PM
in response to: Daffy_Duck
|
|
|
OK, I see your point. The thing is, when I think of free convection, I think of the temperature (and therefore density) gradient in the fluid as being solely responsible for the motion of the fluid and therefore the transport of heat. See here:
http://en.wikipedia.org/wiki/Convection
If capillaries are used, then you've got a sort of fluid pump. So, in the heat pipe there is convection, in the sense that a fluid is moving and it carries heat with it, but it's not exactly free convection, either. Does that make sense to you?
Actually, I'm guessing that because it's not just conduction through copper, or just free convection (as with a hollow tube containing a fluid), but is using a sort of forced convection with the capillaries, that explains why it's so efficient at transferring heat. Wish I had thought of that.
My calculations are exactly what you have done at the bottom of this thread page. I used the thermal conductivity of copper, the 2.5cm^2 area, and a 5 cm length from chip to fan. It's sort of a moot point now since we know that the heat pipe is much more efficient than originally envisioned.
-RendIt
PowerBook G4
Mac OS X (10.4.6)
1.25GHz AlBook
|
|
Posts:
231
Registered:
Feb 11, 2006
|
|
|
|
Re: MBP Heat Problem: From a thermal physics perspective
Posted:
May 11, 2006 2:34 PM
in response to: RendIt
|
|
|
No problem man. It's been awhile since I thought about this stuff. I don't have to deal with heat transfer in my current job but it was my favorite subject in college.
The liquid basically carries the heat through the pipe. Most of the heat transfer occurs at either end (because the temperature gradient is the largest) and the primary method of transfer is convection.
Yes, I get the same result with those numbers. But I think the geometry is probably wrong. My guess is that the copper is quite thin at the interface. So most of the heat transfer takes place in the liquid. The copper is just there to hold the liquid in.
If you calculate with a thin interface...perhaps 2 mm you get a Tc of 500 W/m.
MacBook Pro 2.0GHz / 7200RPM
Mac OS X (10.4.6)
Also have AMD desktop with Windows XP
|
|
|
