2 cents.
 

Doesnt WW, Cascade and RBX blocks have a very very thin base ?
If it could be done:
Wouldnt that influence the blocks performance...?

I don’t think the base is thin through out the entire block. I think the only thin part of the RBX and the WW is over the core in which case the probe hole can be made to be stop just outside the valley of the block. This should prevent affecting the performance reading of the block (most user only care about the reading above the core). Frankly without the proper insulation on the test, this should remove more variables than it introduces.

IF something like this was to be done, it would mean the temperature probe will not be at the same position (in respect to the core) for all wbs, wich will make any datas useless for any wb to wb comparison.

joemac:

Not sure how to back calculate "W" from a baseplate temperature? The insulation is definitely a problem for the 3 barb blocks, but honestly I am not so interested in this now that I found out that weird results were operator error :)

I will, if there is interest, run tests as function of MHz and Vcore using whatever block is on my testbed this week (it's the aquajoe atm). There's a catch though. Can get someone (Les? Incoherent? Groth?) to use results to come up with better radiate-style equations for CPU power using my data? :D

1) Suggest that for the MCW5000A the relatively small increase of Wwb with increasing Flow rate is,perhaps, a consequence of only a small change of "C/W" with flow rate.
http://www.jr001b4751.pwp.blueyonder.co.uk/pHugh7.jpg http://www.jr001b4751.pwp.blueyonder.co.uk/pHugh8.jpg

or as I insist on plotting ( I share Bill's unease but ....)
http://www.jr001b4751.pwp.blueyonder.co.uk/pHugh9.jpg

Think it is a little early to judge the issue.

2) I only do beer-mat sums not computer equations.

Nice. How about power vs flow, for the last 4 blocks tested? I'd just like to see the variation in the power. I see what you mean pHaestus. Would you foresee any benefit in having 3 decimal resolution, I mean, do you think that it would be practical?


From my limited understanding of heatpipes, they are designed to work at a specific temperature, i.e. the boiling point of the liquid inside the pipe. So as a cooling solution, you have to take into account the gradient, between the liquid and the CPU, and it is going to depend on the power being supplied, as well as all the usual variables (TIM joint, notably). Bill's graph is representative indeed: the heatpipe relies heavily on the latent heat (energy from the change of phase) of the liquid being "boiled", so if the temperature is outside of that range, the performance relies strictly on the conduction/convection of the remaining metal assembly, which will be poorer. Many more variables... Testing these involves a different approach than for WBs (as Bill demonstrated). I've had the chance to think about that particular one, ever since I was asked to test one a few months back (had to turn it down: not ready to test yet!). It's a real nightmare, but one I suspect may become part of future cooling products, under one form or another.

pHaestus, if you come up with the data, I would certainly be happy to give it a shot. I am not a programmer though so it would only be a set of functions which fits your data, in Excel sheet format, not a radiate style program.
Of course this is assuming that we are able to identify and account for all the dominant parameters, at best it would be a "typical" value in a typical system, presented as "CPU power output minus secondary losses" Data at a range of water temperatures would be interesting. I would expect to see a Power/Temperature curve upon which the true CPU power output would lie somewhere.

Les, your C/W vs W chart is very disturbing, I do not like it. I can not imagine a situation where power output is dependent on anything but CPU temperature, applied voltage, speed and load level so I do not want to see anything but noise in such a chart where the above parameters are constant. Flowrate is irrelevant. Anything else, any trend, linear or curve in such data says that there's either a measurement error or in fact one or more parameters are changing. Although of course, in reality all these parameters will change within a certain range, maybe predictably.

Thinking.

Cheers

Incoherent

Some wordage:
1) The inputs are sources' heat and paths' resistances.
Varying the flow rate changes the resistance of the "die>contact area >water" path. Changing a path's resistance will effect the source's temperature.Would expect increase in heat through the "die>contact area >water" path as its resistance is lowered.
"C/W" is the "die>contact area >water" resistance.
"Heat absorbed" refers to the heat absorbed by the water.
2) IF heat only reaches the water by the "die >contact area>water" path then the plotted "C/W" is the "C/W" and "Heat absorbed" is the Heat passing through the "die>contact area >water" path.

Plotting measured "C/W" v measured "Heat absorbed" explores both 1) and 2)

Some more incoherent ramblings
If the die temperature decreases (which it does). Then the loss via secondary paths is reduced. More heat is flowing via die>contact area>water path. Agreed.

Also in Fouriers equation Q=k.A.dT/L, dT/Q or C/W, is equal to L/kA. I prefer to think of it in those terms, changing flowrate changes the effective L or A or k or all of them. These are physical parameters, if they are changing then your statement about the points in the chart being like individual waterblocks is a good one. I guess C/W is convenient.

If we adjusted the water temp to maintain a constant die temperature at the different flowrates, the secondary path gradient would remain unaltered, even though the primary "die >contact area>water" path C/W's changed with flowrate. The "Heat absorbed" is constant. If you plot this again at several temperatures I am not sure there would be any relationship except the "Heat absorbed" changing with temperature. Point being, I think the C/W vs W chart is only a few points in a chart that is filled with data points, ie every C/W can have any W so there is no trend to be gleaned.

I'll keep thinking, I am a little muddled here.

Cheers

Incoherent

"If we adjusted the water temp to maintain a constant die temperature at the different flowrates, the secondary path gradient would remain unaltered, even though the primary "die >contact area>water" path C/W's changed with flowrate. The "Heat absorbed" is constant. If you plot this again at several temperatures I am not sure there would be any relationship except the "Heat absorbed" changing with temperature."

interesting
I can do this both on an insulated heat die, and a ttv (~mobo + CPU)
will report - eventually

Data from heat die tests in any insulated condition may be revealing.
Having a leisurely re-look at your ancient data http://www.ocforums.com/showthread.p...0&pagenumber=2
Look forward to the eventuality.

jeez Les, a thread from 2 years ago ?
yea, needs a re-look

Have re-looked and possibly only seen the sensitivity to measurement.
This was very early Billa work since which equipment has been upgraded innumerable times.
Have included data in graph from Incoherent's thread
http://www.jr001b4751.pwp.blueyonder.co.uk/SecW5.jpg

The cascade data plotted here are the preliminary data reported here