Joe Citarella was kind enough to send me a number of waterblocks in return for some of my recent articles. They include:
- 2 Asetek chipset coolers
- 1 Asetek VGA cooler
- 1 Asetek “Antarctica” CPU cooler
- 1 Sleeth 1 CPU cooler
The CPU blocks are the same as the units tested by Joe. Prior to installing them, I decided to test the flatness of the base surface of each.
I began with a visual inspection of the surfaces. All had visible machining marks from the fabrication process. All appeared to have an adequate finish to allow a good thermal connection between the block and chip through thermal compound.
The Sleeth 1 appeared to have been finished on a lathe. All the Aseteks appear to have been finished on a mill.
The sleeth1 and one of the Asetek chipset coolers have the telltale marks left by ‘chatter’ during the machining process. While chatter can be indicative of sloppy machining practices, the texturing does not appear to be deep enough to have any effect on heat transfer.
Unfortunately, the surface on all the blocks was too rough to allow the use of a high accuracy interference test. Instead I choose to use a digital readout dial indicator with an accuracy of 1/100 mm.
To hold the blocks, I attached a vise to a compound slide table I normally use for optical tests. The accuracy of the slide table is well within the measuring error of the dial indicator.
To perform the test, I took three readings on each block: one over the center of where the die would go and one ½” to either side of the center.
The difficulty of such a mechanical test is that it is virtually impossible to set up with adequate precision. If the base of the block is tilted at all from the axis of motion, the reading will be skewed by the resulting slope. To account for this I averaged the readings on either side and subtracted this number from the measurement at the center. This method would leave me with a precise value for the deviation from the ideal surface at the center of the block.
Unfortunately this leaves me with the difficulty of precisely measuring the distance between the center and the outside measurement points. Fortunately my slide table has been tested to be accurate to well within 1/100mm. Any measured vertical error will not be due to horizontal positioning error.
|Chipset block 1|
|Chipset block 2|
(no measured error)
|“Antarctica” CPU cooler|
|Sleeth1 CPU cooler|
Long gone are the days of extruded aluminum heatsinks. The results of this test do not surprise me. Simply by visual inspection all of these waterblocks appear to be very well machined. Baring severe human error it appears that a well machined finish is good enough. Further lapping, finishing, and polishing is unnecessary.
It is possible to make a smoother and more precise surface, but without the proper equipment you are likely to cause more harm than good.