SUMMARY: We talk a lot about “diminishing returns”; here’s some data which vividly illustrates this point.
One of the great things about overclocking is that we continually push the envelope seeking performance gains. As you get closer to the “limit”, you find that the toughest problem is getting that last 5%. Sometimes it’s sooo close but sooo far.
Heatsink designers continually confront this issue – how to get the most performance out of a pretty tightly defined problem. Andy Lemont over at Millennium Thermal Solutions is one such designer, and he has generated two graphs which illustrate heatsink diminishing returns.
The first graph illustrates the fan problem; The graph plots Fan RPM (= cfms) vs heatsink C/W for a 60 mm fan on one heatsink. As you increase RPMs (and NOISE), you do get better C/Ws but at a rapidly diminishing rate. Once you pass 6000 RPM, it really starts to flatten out. One option is to use two fans (you’ve seen those) or larger fans, such as 80 mm (you’ll see more of those). This is the “brute force” approach – more is better.
The second graph plots out some popular heatsinks, comparing cubic volume against C/Ws. I really like this one, sort of like buying a car by the pound. If you consider that the heatsink designer’s objective is to get as much surface area into a well defined footprint (bounded by motherboard design parameters), then a heatsink’s volume bears a close relationship to performance.
And, in fact, that’s what you see in this graph. If you want a quick and dirty indicator of how a particular heatsink will perform, figure out its volume and you’ll get a darn good idea.
However, note here that bigger is better, but also at a diminishing rate. A massive heatsink measuring 3 x 3 x 5 inches has a volume of 45 cubic inches and might get you to 0.15 C/W. It’ll also probably weigh something like two pounds.
Obviously other factors enter into the equation; for example, a heatsink of exactly the same design with a badly finished bottom will be outperformed by the same heatsink that is finely polished. Overall, though, interesting to see that buying a heatsink by the cubic inch makes some sense.
Be the first to comment