Cool silicon, literally

[christopher_wolf]

http://www.reghardware.co.uk/2006/10/26 ... ling_tech/

Not that new of an idea as it is primarily increasing the surface area drastically and giving the thermal compound more contact much in the same way smaller vessels and capillaries bifurcate in the body into organized beds. Who said Big Blue didn't need Bioengineers?

This might even make its way into Thinkpads.

[dsigma6]

I have no large words to add to the thread, but I do find it very intriguing.

I particularly like the self-contained water cooled version, even though it's not a new idea.

[JNavas]

Better to make the chip run cooler, than to cooler the chip!

[BillMorrow]

it seems to me that the more surface area the better to conduct the heat from the chip to the cooling system..
the thermal paste makes contact with the sides of the channels, thus more surface area..

now why are they cutting the channels in a specific "tree" or "human vascular" pattern..?
does such a "format" work better than the angular channels..?

interesting article..

[christopher_wolf]

They essentially want to bifurcate into tiny channels, giving them a very high surface-to-volume ratio, quickly from larger channels. Much like the body has arteries and veins turn into capillary beds, then back into larger vessels. Reason? It is efficient to get a fluid into such areas from larger channels to the beds in such a manner because, suppose you want to fill them, you have to get every little area covered for maximum efficiency. Simply cutting them into tiny blocks would, at a point, choke the flow of it into the areas where it needs to do; just imagine trying to coat every part of that with the thermal compound and the most efficient way to do so without alot of work "pumping" it. This is also the approach they are probably going to take with the forced jet cooling approach.

On a little bioengineering aside; strictly speaking, blood isn't a newtonian fluid when it comes down to such tiny capillaries as encountered in the body due to the discrete nature of erythrocytes and other cells that must pass single file through some, but it is a very good approximation for larger vessels and pretty spot on for most viscous flows.

[vlyne]

My limited understanding is that one of the main problems with thermal paste is getting the air out (reduced conductivity) - which is one reason why a thin layer is recommended (the other being to reduce the conductive path). I can imagine that the hierarchy of small to big channels would squish the air (and paste) more efficiently ( Pressure = Force*Area effect) plus give better conductivity.

[christopher_wolf]

Pretty much; it also increases the efficiency of actually getting a maximum of the viscous thermal compound into all the little spaces to contact as much of the surface as possible. It would actually be more difficult to do that if it was a simple grid setup with uniformly small channels while maintaining surface contact at a maximum.

[vlyne]

Probably getting off the topic a bit but....

Been thinking about this while trying to swallow some spaghetti down my throat So, it's an optimatis(z)ation problem of viscous flow versus heat transfer area. Is the viscous flow in grids to do with flowing around corners? For a given channel width it shouldn't really matter and in any case there shouldn't be any significant loss from flowing around corners?? In nature, we don't tend to see much grid-like biological structures presumably because distances are optimised better in a "drainage" or branching structure but this (distance) shouldn't be a problem in this case. I suspect it's more to do with optimising the linkage between the hierarchical channels so there's less chance of blockage in the smallest channels. OK, spaghetti is down...I should try it while bent at 90 degrees

[christopher_wolf]

Spot on with the optimization of the channels between each other; think of how you would get the thermal compound in there first off. It would be more difficult, given the flow parameters, to contact all the surface with the thermal compound whilst "pumping" it through. You can't really optimize that well with a grid, in 2D, much less 3D.

There are also other reasons that nature doesn't really make grid-like structures as well; stresses on the internal elements versus scalability of the structure (given that it is being constructed with cells along with fibrils in the ECM which do have significant values of persistent length.

[vlyne]

OK, I see the light now! I forgot for a moment that the losses in viscous flow are more to do with shear stresses so narrower/longer channels -> greater stresses and pressure loss. Yes 3D optimisation would be more challenging. For a start the heat transfer would be better optimised with hemispheres (but then why do we get goose bumps when cold - to straighten out the insulating hairs?)

[rkawakami]

My understanding is that "goose bumps" is supposed to raise the hair so that it traps the air close to the skin, thus warming it and keeps it from escaping from the body.

Mind you, that obviously works better for some people than others...

[JHEM]

My understanding is that "goose bumps" is supposed to raise the hair so that it traps the air close to the skin, thus warming it and keeps it from escaping from the body.

A holdover from our more hirsute ape ancestors.

James

[christopher_wolf]

Indeed they do, vestigial leftovers as the erector pili muscles per each hair follicle.

By the way, it is refreshing (and quite impressive in terms of scientific accumen, no less!) to have a intelligent discussion on biofluid dynamics on the forums.

If anybody is interested, I would recommend the *great* 2 books "Biomechanics: Circulation" and "Biomechanics: Motion, Flow, Stress, and Growth" by the venerable Y.C. Fung. Excellent and accessible reading for anybody that would like to know more about fluid dynamics and the role it plays with mechanical/FSI interactions in the human body.

[vlyne]

By the way, it is refreshing (and quite impressive in terms of scientific accumen, no less!) to have a intelligent discussion on biofluid dynamics on the forums.

It is impressive that someone understands enough to explain it to those who are closer to our ancestors

The answer is probably in the books you mention but, the closest example I can think of, in terms of optimal structures involving heat transfer, are those "cells" that form when you add oil to a hot pan. Those cells change in shape/size as you crank up the heat. I suspect that type of structure is some sort of optimal 3D structural mechanism for heat transfer? So, instead of trying to figure out what is an optimal structure for pasting surfaces, one option could be to use an oil cooling sandwich? (altho' there is still a problem of contacting surfaces but it could be a thin malleable surface). I must admit I don't fancy having oil in my laptop (then again I don't really like that paste as well). I guess the advantage of the oil cooling is that it is a more "efficient" mechanism of heat transfer compared to conduction.