Heatsinks - Hitting a Wall??

Add Your Comments

Has current heatsink technology peaked? — Joe

SUMMARY: Heatsink design is hitting a wall – where do we go from here?

When “popular” overclocking started in the dark ages (three years ago), one of the areas that exploded with innovation and, sometimes, outright quackery, were heatsinks. It was obvious to all that heat was the disease and heatsinks were the cure. As more of us took up overclocking, the heatsink market went from a ho-hum to a humdinger.

We saw a LOT of Slot 1 heatsinks of all stripes and number of fans. When Slot 1 was phased out, Socket 370 heatsinks were ripe for innovation, and Alpha came along with the legendary PAL6045 to bust it wide open.

AMD emerged from the shadows with superior products – Duron and then T-Bird CPUs. In contrast to Intel’s power-miserly PIII series, AMD’s CPUs turned out to be toasters. And the heatsink challenge was ratcheted up a couple of notches – now we were looking for heatsinks to handle, 100, 110, even 120 watts. Considering that this heat is radiating from an area about the size of your fingernail, cooling AMD CPUs moved heatsink technology ahead quickly.

If you look at how heatsinks are performing today (See Heatsink Ranking), we’ve seen a LOT of heatsinks entering the top ranks. C/Ws of 0.35 or less, once an exclusive club, has now become the “entry ticket” for market success – I count 11 heatsinks capable of 0.35 or less.

Question: Where do we go from here??

In the ranking table, I put up “holy grail” C/W 0f 0.27 – water cooling. It does take a LOT of surface area and airflow to approach this mark. We are seeing very aggressive, and very noisy, 60, 70 and 80mm fans on copper heatsinks weighing well in excess of one pound.

But there are material limits to performance.

At some point, diminishing returns sets in very quickly. Doubling the size of a Swiftech 462A, for example, is not going to yield double the performance. In fact, I doubt you would see ANY performance gain. Metal A can move a given amount of heat so fast and so far; copper is a superb choice, considering its physical characteristics, cost and manufacturing ease – it’s tough to beat.

Fans hit limits also; at some point, doubling airflow does not yield appreciable performance gains. What you get, and are getting, is increased noise and higher power consumption. What might be a great heatsink turns out not to be acceptable to consumers because it’s too noisy – the cure is worse than the disease.

So back to the question:

Where do we go from here??

Possibility: Heat Pipes. The Cooler Master HHC-001 is one example where this technology seems to work quite well. But at this point, it only gets this heatsink into the “0.35 C/W or less” club – it does not leave the competition in the dust. Expect others to pursue this technology – should be interesting.

Possibility: Water. There’s still a lot of hydrophobia in the mass market, but what was once considered far-out and kooky is becoming mainstream. Following a couple of years of “kit” business, Koolance enters the scene with a turnkey product. Swiftech is entering also – expect others.

But Joe – is this a heatsink? Sure it is – broadly speaking, a heatsink moves heat and gets rid of it. Watercooling moves heat with water and uses a radiator to dispose of it. An aircooled heatsink moves heat through a base (typically copper) and disposes of heat through surface area (fins) cooled by a fan.

Air can only do so much; water can do 100 times better, hence the attractiveness of water cooling.

Possibility: Peltiers. Ah, the tiny fridge approach. LOTS of practical problems: Condensation, power, cost and very high heat loads. Not insurmountable problems, but NOT trivial. I’m skeptical on this one – too many ways to screw up, not “off the shelf” enough for the mass market. But I’ll be the first to gladly eat my words.

Possibility: CPU Throttling. Gets too hot, dump duty cycles. Intel has it built in already. Sort of defeats the advantage of Mega Gigs, though. I don’t see how you sell a 5 GHz CPU that throttles back to 1 GHz if you push it. Would you pay a premium for a car that goes 120 mph in theory, but will only go 70 mph when you push it to the max?


I’m not prescient – I have no Tarot cards or chicken entrails that tell me what comes next.

I do feel we are hitting some practical limits with current materials and designs, and that substantial performance gains will require some “out of the box” thinking.

I expect to see a lot of crowding around the 0.30 C/W mark in the near future, with noisy fans and one pound+ copper heatsinks. I think you may see more “flash” than substance in many new offerings.

Breaking 0.25 C/W with air cooling may be the “heatsink design sound barrier.” I hope we see it breached soon.

Email Joe

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>