Testing philosphy

[Jabo]

It is a simple question.
What is tha aim of all the PC cooling equipment testing:

1. To establish properties of hardware elemnts?
2. To establish performance (absolute or relative)?

with all the consequences of the above choice

Any thoughts anybody?

[bigben2k]

Yes.

To "establish performance", we have comparative and analytical test benches, which produce relative or absolute (respectively) performance data. Of course "absolute" is a moving target, with varying processor specifications.

If you want to discuss actual reviews, there's also another element of "usability" or functionality, which would also include installation and maintenance.

[pHaestus]

with all the different processors and sockets and heatspreaders and mounting forces required I don't know if "absolute" performance is reasonable to attain.

The guiding philosophy for my work is something like this:

1) I want to be sure I publish performance results that represent the best I could achieve with the wb while testing it (I have a strategy for discarding bad mounts and repeat testing when I get that odd "good" mount until that performance is the norm).

2)I want to make sure my results are reproducible enough that if I were to test the block again in several months that they would be statistically the same. This involves regular testing and calibration of the diode reader, going back and rerunning a standard wb when things look funky, and generally just being careful. With a PC being still used for testing this is HARD to accomplish.

The Procooling target audience is mostly serious wc enthusiasts and DIY wb builders. Accordingly, I think the purpose of Procooling testing should be consumer-oriented but with extensive discussion on how design of wbs affect performance. The consumer oriented part is where I am headed now. With my new pressure transmitter setup I hope to come up with a database of delta T, Flow, pressure drop, and price for all wbs tested and then for radiators and estimate for resistance of tubing. Add in pump P-Q curves and costs for everything and you should be able to make a really good online wc calculator for things like "would block x outperform block y in my system?" "What is the best cooling system I can build for $xx?" etc etc. This is a huge amount of work but something I am thinking is a nice ultimate goal.

[BillA]

the use for such testing results will depend on the 'user'
and it follows logically that different 'users' of the testing results will have possibly different needs in terms of precision and accuracy

a simple hierarchy could be envisioned:
Dell, Sony, Compaq user - no data reg, a box is purchased
Fry's user - next to no data, informed choices are not made by looking at the colored boxes
Enthusiast user - data required, but of what quality ?; not all enthusiasts are technically sophisticated - even if interested
Mfgr's representations - accurate and legally defensible (this is why soooo few mfgrs post specs)
Design engrs - best accuracy AND repeatability, any source of variation must be identified and controlled

Jabo
you getting the drift here ?
this site is for wannabe DESIGNERS
these guys really do want the absolute best in data quality, which they will use to their own ends

I am not commenting on the ability to use such data, though a better understanding is slowly permeating through most of the WCing community (with a bit of experience)

so given the existence of 'high quality' data, greater differentiation becomes possible on the basis of 'specs' (how meaningful these very small increments are to a general user is questionable)

end message ?
different strokes for different folks
here at procooling its about accuracy

[Jabo]

Wow, thanks for very exhaustive answer!
The reason I asked this questions was quite mundane and simple.
You basically answered it all indirectly hence I want to reiterate here (I got to stop trying to use this puffed up language )

The way I seee it is as follows:

1. First general group of ppl as described by unregistered as 'wannabe DESIGNERS' (that explains why you are here unregistered - j/k). These peaople need 'absolute' type of testing providing them with the most accurate data available using well described and documented scientific testing methodology (sealed chamber with full enviro control-doable at low cost at home, fully controled heat sources of different heat transfer size surface, top quality heat dissipation assembly to remove this type of possible limitation).
Experimental data obtained in this way could be of great use to wannabe designers.

2. Second group of peaople (VAST majority) are enthusiasts wanting to install a water cooling assy in their machines and looking for advice.
In my opinion this requires totally different approach. To describe it as simple as possible such testing would require 'filed conditions emulation' or to rephrase 'how an average enthusiast would put together a kit and use it' type of approach. Use of top quality measuring equipment and clearly stated methodology is also a must here.
I think such an approach would show potential users how different kits, pieces of hardware would perform comparatively if put together according to mfgrs reccomendations and manuals.
unregistered, are you catching my drift here? (LOL)

What do you think, is it feasible approach or rather not very practical?

Jabo

[pHaestus]

It's a huge expenditure of time as your "system testing" audience immediately demands a huge variety of combinations of pumps, radiators, fans, fittings, other blocks in the loop, ad nauseum. You cannot easily measure water temperature or flow rate without introducing at least some restriction (even if it's just a pvc cross)... it would be incorrect to reference Heisenberg but there is I suppose a correlary.

In my opinion this isn't the sort of testing approach that anyone with a technical background is going to find very rewarding or interesting for long, and someone without technical skill is going to do it so poorly as to make it useless.

Now having said that, what of kits constructed by a mfgr after extensive R&D? We as a DIY community could I think learn something from examining how the individual parts come together to make a whole cooling loop. The compromises between size, noise, total cost, and performance that mfgrs make are often somewhat different from my personal choices; how much different are the outcomes? THIS is a question worth throwing some time into looking at (and I have a German and Swiftech "kit" that will be tested soon as well as my personal cooling loop).

[BillA]

"(that explains why you are here unregistered - j/k)"
ok j/k, not offended
I've been here for some years Jabo, singing quite the same song

if you had more perspective regarding this site, you might consider what procooling would be if I had never posted any data here
- like so many other 'tech' sites mumbling about 'what if' with the un-informed majority running off the qualified (OCers is the perfect example)

[Jabo]

pHaestus, I was affraid you were going to say just that.... what about fellas with resources to do so (manufacturers) cross testing the hell out of their equipment... no, stupid idea, being biased allegations flying in the air would make it difficult to breathe

I don't know... I really feel that average enthusiast leel user testing method is needed... but how to do it...

I see so many peeps posting bollocks about choice of pump/block/rad setup that it makes me cringe

I think your idea oc a calculator for performance estimation of different combinations would be great (based on pressure drop mapping and resulting coolants velocity dependant thermodynamics).
We can cook a little proggie where you input block, cpu(die/IHS)-block contact area, TIM, ideal tubing layout, and the rest (btw, tubing pressure drop is calculateable(is there a word like that at all LOL) off the bat, no need to measure it)

[Jabo]

Originally Posted by unregistered

"(that explains why you are here unregistered - j/k)"
ok j/k, not offended

In my life I found auto-irony and laughing from my good self to be priceless qualities, glad you got it too

[pHaestus]

jabo even if you have some dumb tech (that would be me) generating radiator and wb performance data and digging up pump P-Q curves, the REAL kicker then becomes airflow through radiator. So many overclockers use some sort of Enermax/clear LED/bling bling fans w/o P-Q curves. And then many more undervolt (in a variety of ways). How to determine the actual airflow (and therefore radiator) performance? How to account for whether someone puts their radiator so it gets fresh outside air as opposed to on a blowhole at the top pulling in case air? This "ambient" temp change can be BIG for final performance. And what about those guys that put 90 degree elbows on the pump intake? How the hell does one deal with craziness such as that in a calculator? Average enthusiasts don't have any temp probes other than case and CPU; these are of limited utility as input terms.

Seems to me that some really applied articles on plumbing (I have done one), part choice for quiet, cost/performance, or balls out cooling are needed WITH data to justify choices. TON of work though

[Jabo]

Yeah, as usual, real life turns it ugly head back here and bites us where it hurts
Oh well, looks like it was just a food for thoughts here, a pity...

Good God, morons with bright (LED) ideas should be somehow kept well clear form such a calculator, I was more thinking of rather 'typical' rational setups (no elbos, crappy fans, raditors sitting just above heaters etc.) and a simulator being able to give you a different performance values based on CPU and h'ware choices.
I guess you are right that this would require some truly herculean effort of data gathering and keeping it contanstly upd to date as new stuff is being released quite often these days (how to get all the equioment samples to get minor loss coefficient figures?).
Cheers for taking time to write all this

[jaydee]

Originally Posted by unregistered

this site is for wannabe DESIGNERS

Ouch, you make it sound like a bad thing.

[BillA]

no jd, we ALL learn
when I first hit procooling I knew nothing about WCing a computer
I just study REAL hard

the road is open to those who want to work (yes Dave, and have also the ability)

[pHaestus]

I find it funny that so many of our site's readers hold my tests up as a "standard". There are still gaping holes in my testing and my knowledge. It's still very interesting though so doesn't feel like work.

[talcum]

Originally Posted by pHaestus

I find it funny that so many of our site's readers hold my tests up as a "standard". There are still gaping holes in my testing and my knowledge. It's still very interesting though so doesn't feel like work.

The Holy Grail is to get to where you don't have to add the disclaimer, "NOTE: These results are NOT comparable to tests done by others.", because if others tested rigorously, then the answer would be the same.

[BillA]

Originally Posted by talcum

The Holy Grail is to get to where you don't have to add the disclaimer, "NOTE: These results are NOT comparable to tests done by others.", because if others tested rigorously, then the answer would be the same.

would that rigor alone were sufficient, but it is a bit more difficult

progress is being made
I have spent most of the week making and testing 3 flow restrictors in 3 sizes, and curves generated with 2 different types of pressure taps
- these will circulate between JoeC, pHaestus, myself, and anyone else publishing test results wishing to sing in the same key

not standardization by any means, but at least the offset will be known

[Groth]

Originally Posted by unregistered

3 flow restrictors in 3 sizes, and curves generated with 2 different types of pressure taps

More details and pictures?

[BillA]

lol
6" straight copper tubes,3 ea in 1/4, 3/8, and 1/2"CTS, smooth, moderate dent, Big dent
the problem is that with a (pseudo) venturi the pressure recovery is high even with substantial restriction

test setups are:
1) cross/CTS adapter -> connecting tube -> test tube -> connecting tube -> CTS adapter/cross
and
2) CTS tube of relevant size with a 1/8" tap in sidewall -> connecting tube -> test tube -> connecting tube -> CTS tube w/tap

flow rates from 0.3 to 4.0 gpm

have not yet plotted the data, will be interesting to see how well the 2 'pressure tap' methods compare
(god I wish this stuff was automated, 3 more weeks perhaps)

[Groth]

How easy is it to make consistent pseudo-venturis? Think their accuracy/reproducibility will beat other options like orifices?

[Groth]

BTW, I use the 'tap in the sidewall of CTS tube' method for my differential manometers. Gave me better results than crosses or tees.

[BillA]

the operative phrase is CHEAP
next is easy
then fast

my bfh works fine

reproducibility is perfect, 1 unique set doing the round robin bit
replacement, if necessary, will be very tedious

EDIT:
I found downsizing the cross (to reduce the internal volume) helpful