Tom rounds up 34 HSFs

[bigben2k]

http://www6.tomshardware.com/cpu/20030113/index.html

Bill might have some competition...

[pHaestus]

Not quite. Still not a bad setup for using an actual CPU; I have similar temp monitoring and am also now using K7Burn in real time from DOS with temps taken on separate machine. I need to get the info on what pins to get current and voltage from though; would be useful to have for sure.

[bigben2k]

Using an actual CPU?!?

In the last cooler test, we ascertained the temperature data with the aid of the computer-regulated CPU simulator KT-2. With the KT-2, an aluminum block shaped like the die is heated with the aid of a power resistance. The thermal sensor of the KT-2 is located just below the surface on the inside of the block. This enables a highly accurate, reproducible temperature measurement. Despite this, taking measurements using a simulator has a few drawbacks:

-the shape and size of the heating surface do not correspond exactly to the die area of a real CPU;
-the temperature distribution of a real CPU surface cannot be simulated;
-the output does not correspond to the values of a real CPU.

During the course of this test, we used the same kind of measurement platform used by AMD in their cooler tests for the very first time.

[pHaestus]

uh yes Ben.

Read the article.

Note pics of modified A7V-333 to read current and voltage.

That board had to come right from Asus w/o their COP (or POS or whatever). They discuss the Max1617 by name as well.

[bigben2k]

re-read...

The measuring platform consists of a modified A7V333 motherboard from Asus and a temperature sensor evaluation kit developed by Maxim. The board was modified in such a way as to enable the measurement of both voltage and current on the processor during operation.

This makes it possible to take precise measurements of the power dissipation during operation. It simply results from the product of the current drawn multiplied by the drop in voltage on the processor. In addition, using the temperature sensor evaluation kit, the exact die temperature is read out and displayed on the controlling PC.

We used an additional temperature sensor at a distance of about one inch above the ventilator fan in order to determine the ambient temperature. The processor (AMD Athlon-XP-2400+) is then heated with the aid of a DOS program that guarantees a constant processor utilization of one hundred percent. Instead of thermal pads, we used Arctic Alumina thermal paste in all of our tests.

Oops...

[pHaestus]

The question then becomes whether "the best" in situ test with a CPU is of more utility than a mediocre test with a bench testing setup. Hard to say, really. I have some specific reasons for testing with a CPU at this time; otherwise I would greatly prefer a good die simulator.

[bigben2k]

Define mediocre!

The results should be relatively accurate, but is no means to calculate the resulting temp on one's own PC.

This is going to bring up the old "so what if this cooler is 1% better than that one" question.

[pHaestus]

I am defining mediocre as a die simulator that doesn't measure current and voltage simultaneously, that doesn't mnimize secondary heat losses, and that uses temperature measuring of unknown linearity or uncertainty. The Innovatek die simulator may (or may not) be any good. Without honest specs posted regarding it, though, I am dumping it in that category of mediocre. +/- 1% would be not too bad if that was in the final "C/W" result; +/- 10 or 20% isn't so good though...

[bigben2k]

Harald Thon measures the amperage, with a mod?!? He seems to rely on the onboard voltage meters.

To refine this, maybe we should be looking at what the CPU is actually doing with this power, instead of trying to re-create/simulate it. Near impossible, of course...

Otherwise I agree: there's lots of secondary losses here. How would you isolate them?

[gmat]

I see two big names missing from that round up:
- Zalman -> very popular among silence afficionados
- Alpha -> WTH are they thinking, not including the kings of air cooling ?

[pHaestus]

also thermalright

[BillA]

good ol' inq

http://www.theinquirer.net/?article=7213

catch the very last line, lol

[gmat]

There's a good point about the heatpipe system used in Shuttle mini PCs... would be neat to include it in a round-up.
Also, nothing in terms of price, for PCs at work i ended up with Taisol HSFs, which are very good at cooling (and rather silent) *and* cost about $23 here, versus Alphas or Swifties which cost between $55 and $70...
I mean, if we do a (cooling performance)/(price*noise) the picture changes totally. THG could have done that at least... bah. Who reads them nowadays.

[bigben2k]

I don't see the absence of price as an issue, given the different prices a user would encounter throughout the world.

There's also an availability issue, where some HSF might only be available in some countries. THG is based in Germany, so that might have been a factor. Maybe they just asked for samples, and tested those received (and ignored those non-freebie'ers).

If price was included, it would also have to be converted to many other currencies, and the availability would have to be listed, and that's just too much work. IMO, it's best left up to the user to make that determination.