Whitewater Comparison

[pHaestus]

nikhsub1 from our Pro/Forums loaned me an original Little River Whitewater so that I could compare it with the currently available DTek Customs Whitewater block. This testing is really right at the limit of what is possible with my testbed btw.

Here's the Review! (http://www.procooling.com/reviews/html/little_river_whitewater_waterb.php)

Also be aware that Scott told me "take as long as you need" on the testing when he sent the blocks... last Christmas!

[Wildfrogman]

Nice review. I was always wondering how close the dtek blocks performed to the original. Now I know and its good to see they are very, very close :) .

[Cathar]

Thanks for the confirmation Phaestus.

One thing of note. A visual comparison of the thickness of the two plates gives a bit of an optical illusion. The bottom edges of the LRWW are chamfered, so the base-plate does appear a little thinner than it actually is. Using a pair of calipers the two blocks (well, the DTek WW I have here) are within 0.1mm difference on the thickness of the bp.

Also, re: the CPU internal diode. I picked up an AMD mobo that reads the on-die diodes of the AthlonXP CPU's. Across 3 CPU's I noticed a lot of variation in reported temperatures for the exact same cooling, voltage, speed and load. One CPU at 2.15v/2700MHz would report ~+7C over the water temperature. Another reported ~+10C, and yet a third reported ~+13C. While keeping a single CPU consistent means having repeatable results, it is readily apparant that the CPU internal diodes are not calibrated in any way to be consistent from CPU to CPU. This doesn't affect your tests but will be something to keep in mind should you have need to change the CPU used.

Thanks again for the great work.

[pHaestus]

If I change the CPU then I am forced to recalibrate the whole setup again. I use a water bath and a Digitec 5810 thermometer for the calibration.

[Groth]

Cathar, it's all about manufacturing tolerances - minor variations in interconnect resistance, gate capacitance, transistor leakage, etc. You can't blame the thermal diode alone for your varying readings; not only will the thermal diodes in two chips have slightly different non-ideality factors, but the the two CPUs will have different power consumptions. Low-power mobile CPUs come from the same wafers as desktop chips...

[Cathar]

Groth, the difference between +7C and +13C is close to 100%. Can't tell me that one CPU is consuming 50% of the power of another due to manufacturing tolerances.

[pHaestus]

+/- 3C is right at the tolerance for some diode/diode reader combinations. The one I use is fortunately much better than that, and when calibrated has solid 0.1C accuracy.

[Groth]

No, I'm not saying one CPU is consuming 50% of the power of the other. Just saying that blaming *all* the difference on the thermal diode isn't useful.

By the way, I've seen a 20-25% difference in power comsuption in a sample of 4 bartons. It was a dead end project due to calibration impossibilities, but it was consistant.