Letter to Arctic Silver: Benefits Of CPU Compound Combined with Carbon Nanotubes
By Phil Trent
Date: April 17, 2002
From: Phil Trent at Van’s Hardware
To: Nevin House at Arctic Silver
After reading the following article, I thought immediately of what effect carbon nanotubes might have on your Artic Silver cpu compound.
Here are the relevant links:
Johnson’s group determined that epoxy doped with nanotubes showed a 125 percent
increase in thermal conductivity at room temperature.
http://www.sciencedaily.com/releases/2002/04/020416073536.htm
Carbon nanotubes:
$500 a gram
http://cnanotech.com/6-0_online_store.cfm
Please let me know if you have an interest in testing
carbon nanotubes in Arctic Silver.
Regards,
Phil
===================================
From: Nevin House at Arctic Silver
To: Phil Trent at Van’s Hardware
Phil,
Thanks for the links.
We have reached the point with the conductivity of our best compounds where the layer thickness is more important than additional increases in conductivity. For example, AS3 has a thermal resistance of 0.0024C/W for a 0.001" thick layer on a 1 square inch contact area. The area of a modern AMD CPU is about 0.187 square inches so you have to divide the compound's 1 square inch resistance by 0.187 to determine the resistance for the actual area of the thermal joint.
0.0024 / 0.187 = .0128
So at a layer 0.001 inch thick, the temperature drop across the thermal joint will be 0.0128 degrees C per watt of power dissipation.
Then for a 70 watt processor...
70 x 0.0128 = 0.896C
So the thermal compound accounts for less than 1 degree C and that is the maximum improvement that could be achieved even if the compound were 100,000 times more conductive.
Factor in that you would need to use the $750 per gram variety of nanotubes to avoid contaminants and that you would need at least 30 grams to make even a small sample that was large enough to mix and you can see that you are looking at a $25,000 experiment with very little potential for anything other than bragging rights.
Still, I am sure that some high-hype thermal compound company will mix 1/10 of a gram of nanotubes into a 5 gallon bucket of standard zinc oxide thermal compound and proclaim it to be the first carbon nanotherm based thermal compound. (Of course, it will not perform any better than the compound did before the minuscule amount of nanotubes was added.)
I think you can see why the main focus of our research is in enhancing compound flow and minimizing layer thickness. This is what the addition of boron nitride did for AS3 and why it performs better than our earlier silver products.
As a side bar; the upcoming T-bred CPUs from AMD will have contact ares that are only 80 square millimeters (0.124 square inches) vs the 121 square millimeters (0.187 square inches) of the current Palominos. This will increase the importance of the interface material as the energy now must flow through a smaller contact area to the heatsink.
We can go back to our math and see what this die size reduction will mean in the real world.
If we compare 2 thermal compounds on a current Palomino that have thermal resistances of 0.01C/W per square inch and 0.005C/W per square inch (Both at 0.001 inch layer) on a 70 watt Palomino, the math works out like this:
Die size = 0.187 square inch
Compound A: 0.02C/W / 0.187 = 0.107C/W Total joint resistance at 0.001 layer
Compound B: 0.01C/W / 0.187 = 0.0535C/W Total joint resistance at 0.001 layer
70 x 0.107 = 7.49C drop across the joint for Compound A
70 x 0.0535 = 3.75C drop across the joint for Compound B
So the CPU will run 3.74C cooler with Compound B
When we substitute the new smaller die our numbers change a bit.
Die size = 0.124 square inch
Compound A: 0.02C/W / 0.124 = 0.161C/W Total joint resistance at 0.001 layer
Compound B: 0.01C/W / 0.124 = 0.081C/W Total joint resistance at 0.001 layer
70 x 0.161 = 11.27C drop across the joint for Compound A
70 x 0.081 = 5.67C drop across the joint for Compound B
So with the smaller die size, the CPU will run 5.6C cooler with Compound B. The difference between the compounds is magnified (5.6C vs. 3.74C) as the total resistance of the thermal joint increases as the area of the joint decreases.
Hope you find this useful,
Nevin House
Arctic Silver, Inc.
===================================
Pssst! We've updated our Shopping Page.
===================================