post #1 of 1
Thread Starter 
The following info. should be useful for those who want to know about HSF base finish vs. thermal conduction. The link at the bottom provides a microscopic view of the metal surfaces and why smooth is good but more TIM than fills the MICRO PORES is bad.

If you want to discuss any of this info. like a mature adult, we can. Insults and hating won't change anything so please refrain from those types of posts. People are free to believe whatever makes them happy. This info. is for your edification. What you do with it is totally up to you.

We can debate surface finish forever but the reality is if it were possible - which it is NOT, then 100% metal-to-metal contact at a microscopic level would be the best means to transfer heat from the heat spreader to the HSF. By being smooth and FLAT the two surfaces have the least thermal resistance. Technically the smoother the surface while remaining flat, the higher the percentage of metal-to-metal contact. Shiny does not necessarily mean the surfaces are flat nor smooth.

Lapping is intended to make a surface flat. It can actually make the surface smoother or rougher depending on how the work is done and with what grit abrasive. Polishing is typically done for looks, not for better thermal conduction. It's very difficult to polish a surface the size of a CPU heat spreader and keep it flat. Polishing of specimens for microscopic inspection is not concerned with the sample being flat across the entire surface as is required for good heat transfer in a HSF/CPU. Micro grain polishing is primarily intended to show the grain structure of the metal. This microscopic metallurgical inspection would show you why a smoother surface transfers heat better than a rough surface where the actual metal-to-metal contact is much lower due essentially to micro high spots. (See link at bottom).

The theory that more surface area via a rougher surface or dimples improves thermal transfer is unlikely for a HSF/CPU application because air has poor thermal conduction meaning TIM must be used to FILL the voids from the rougher surface or dimples to improve thermal conduction. Unfortunately TIM is an insulator compared to proper metal-to-metal contact. So while the surface area may increase with roughness or dimples, the thermal conduction may decrease due to LESS metal-to-metal contact and the need for TIM - which is an insulator - compared to metal-to-metal contact.

As far as convex HSF bases are concerned, they are not or should not be convex by design but in fact they are often the result of poor machining - regardless of what a HSF mfg. tells you. They is no need nor desire to have an excessive pressure in the center of the HSF. Beyond a moderate pressure, more pressure does NOT = more thermal conduction. Any quality HSF design should have more than enough pressure to provide proper thermal conduction without a convex base. A convex base can reduce proper pressure and surface contact on the rest of the HSF and heat spreader surface, reducing thermal conduction. You will often see poor results from HSFs with a convex base compared to those with a flat base.

The link below shows a side vew of a smooth surface under magnification. As you can see if you had two "smooth" surfaces - a HSF and a CPU heat spreader for instance, with one on top of the other, then only the high spots would actually touch each other. TIM is used to fill the MICRO PORES or voids ONLY. It's preferred to have 100% metal-to-metal contact but this is impossible. You only want enough TIM to fill the voids, not so much as to prevent the true metal-to-metal contact of the HSF base and CPU heat spreader. This is why more TIM is bad for thermal conduction to the HSF.,r:15,s:245

NOTE: The pic is showing a machined surface vs. an electro polished smoothened surface but it doesn't matter how you smooth the surfaces as long as it improves metal-to-metal contact.

There is also diminishing returns as the surfaces get smoother and smoother because the TIM required becomes less and less - so don't become OCD over surface finish. wink.gif
Edited by AMD4ME - 3/3/12 at 7:59pm