Why Are There Fins on That Chip? Getting Heat Out Is Harder Than Making It
By the BrainSnail editorial team. How these articles are written and checked, and how to tell us when one is wrong.
A finned block bolted to a component does not absorb heat so much as hand it to the air, and the whole design is about the two stubborn bottlenecks in that handover.
The two bottlenecks
Heat leaving a hot component has to cross two barriers in series and both must be attacked. The first is the path through solid material from the chip to the outer surface, which depends on how well the material conducts and how far the heat has to travel, and which metals handle well. The second is the handover from that surface into the air, which is far worse, because air conducts heat poorly and carries very little of it away per unit volume. The second barrier dominates in almost every case, which is why the design is overwhelmingly about surface area and airflow rather than about the metal.
Why fins rather than a block
The shape follows directly from where the difficulty lies:
- •Handover to air depends on how much surface touches it
- •So the same mass of metal is spread into many thin fins
- •Fins must be spaced so air can actually pass between them
- •Too close together and the air stagnates, which wastes the surface
- •Too far apart and the surface area is lost
- •A fan moves more air and shifts the whole design point
Why the paste matters
The joint between the component and the block is a bottleneck out of proportion to its thickness, because two machined metal surfaces pressed together touch only at scattered high points and the rest of the interface is air, which is an excellent insulator. A thin layer of thermally conductive paste fills those gaps, and it works because it is far better than air rather than because it is better than metal. That is why more paste is worse rather than better, since a thick layer replaces metal-to-metal contact with a mediocre conductor, and why the correct amount is barely enough to fill the gaps once the two are clamped together.
Why a bigger block is not always better
Adding metal helps only up to a point and the limit is worth understanding. Heat has to travel along a fin from its base to its tip, and the fin cools as it goes, so the far end of a long thin fin is barely warmer than the air and contributes almost nothing. Extending it further adds mass and cost without adding performance. The same applies to spreading heat sideways across a plate from a small hot chip, since the far corners of a large plate stay cool and idle. Thickness, length and conductivity together set how much of the available surface is actually doing work.
How heat is moved before it is dumped
Large modern designs separate the job of moving heat away from the job of giving it to the air. Sealed pipes containing a small quantity of fluid carry heat from the component to a remote fin stack far faster than solid metal could, by evaporating at the hot end and condensing at the cold one. Liquid cooling pumps heat to a radiator elsewhere in the case. Vapour chambers spread heat sideways across a flat area before it goes anywhere. In every case the fin stack still does the final handover to air, so the problem is relocated rather than removed, which is why even liquid-cooled systems end in a finned radiator with a fan.
The takeaway
Heat crosses two barriers, through solid material and then into air, and the second dominates because air conducts poorly and carries little away. That is why the same metal is spread into thin fins spaced so air can pass, and why a fan changes the whole design. Paste at the joint fills air gaps between high points, so more is worse. Pipes and liquid relocate the problem to a fin stack elsewhere.