product testing today – who is being served ?
 

As a number of recent threads have revealed, there is little consensus as to how products, wbs in this instance, should be tested. In all instances the ‘equipment required’ exceeds that needed to operate the device and observe it such as the end user would with their own computer. Is it not appropriate, given finite resources, to consider the cost/benefit ratio of various testing methodologies as applied to their intended audience(s) ?

Do we need a Testarossa to buy groceries ? I submit not (but max bling eh ?).

Procooling is understood to cater to extremists and in no area may this be seen more clearly than watercooling. And my contribution has been ever increasing levels of sophistication of test equipment and procedures (which is still on-going), leading to the distinction of ever smaller increments; the specification wars. These very small increments are of interest to three groups; product designers, those purchasing based on specs, and gearhead enthusiasts doing some armchair engineering.

Admittedly at some point the differences fall below what a user could detect even if all else in their system remained the same (which is never the case even if it seems so). Additionally most users are unfamiliar with the jargon and equipment of thermal management and are not seeking technical enlightenment; they want to be told that which is ‘the best product’. Many users prefer simple information even at the risk of it being incorrect, they have no choice if the other explanations are beyond their understanding (users, remember).

Can the procoolers here not define, and validate, a CPU based thermal test system that is not so complex or expensive that review sites, and enthusiasts, would be able to use and reference a common platform ? Review sites are not going to spend thousands for a bench, but some hundreds for several key pieces may be more palatable. Yes it is understood that if “C” and “W” are not both known with precision the value is, ahem, uncertain; but perhaps if the uncertainties were similar then the results would be at least comparable in the ‘real world’ sense.

We are good at creating confusion, can we bring some clarity also ?

An Addendum
Lets leave the high-resolution totally out of this discussion; system based testing, not bench.
I am aware of only 6 high-end testers, let them figure out what is practical as they actually have to make it work. Other threads to instruct the high-end testers please.

some suggestions:
what measurements are essential ? (sink performance relative to that platform)
what resolution ? (related to users)
what measurement method(s) ? (define procedure, calibration, base-line definition, etc.)
what 'special' equip ? (list sources and equivalents, nominal prices)

can procooling draft a test procedure for review sites ?
(can we solicit their input after abusing them so ?, lol)

Re: product testing today – who is being served ?
 

A question about drafting a test procedure for review sites:

How big of a parameter is "doing a thorough job" in the final numbers? You can't stop kids from pestering companies for free stuff and then doing a halfassed job testing it. Companies will usually give people coolers to start with (relatively cheap yes?). My gut instinct is that half-assedry (lol) probably makes their results far more worthless than an extra digit of resolution ever could help. Or even a standard testbed.

I don't think you could do even the most rudimentary water cooling system testing without:

1) Measuring CPU temp reliably (0.125C res seemed ok to me but you have to be able to tap into the diode with external readers). IHS case temp would be good enough for general reviewers? Better than MBM, but then whether the TCs stay in spec is a problem. Onboard temp readers from mobo are worthless and I ignore any testing that relies solely upon them.

2a) Water temp inlet side. The other part of the dT measurement if you're testing a waterblock. I guess 0.125C is again ok, but if you want to estimate Watts you MUST have 0.01C resolution and then collect water in and out of the block.

2b) Air into the radiator. The other part of the dT measurement if you're testing a system. I guess 0.125C is ok. This is a very very hard measurement to do well.

What kind of testing could you do with that though? Pretty show and tell stuff:

"in this standard system with xyz components the waterblock performed as so. This can be compared to other waterblocks in the same system as follows"

Can't really make buying recommendations for people unless they too use the same rad, pump, tubing, loop setup. Is that ok? Maybe so for complete systems?

Re: product testing today – who is being served ?
 

OK so here's a draft test system that would be a little bit more sophisticated. It would be capable of doing what my old test bench did more or less but at a slightly lower resolution:

Crystalfontz 633 DOW reader/logger/controller ($100US)

-DOW probe modified for water inlet
-DOW probe modified for water outlet
-Array of 4 DOW probes in a cirble for air inlet

Use the fan headers for monitoring fan rpms/PWMing the radiator fans to reach a set water inlet temp (as close as possible anyway).

Use the fan header to also record flow rates of the system using a Swissflow flowmeter ($100US)

Groove the IHS and use a small tc to read CPU temp? $200ish?

Calibrate all the probes in the same water bath from 0-50C or so and then go to town.

What about incoherent's old thermistor-based setup? It measured dP I think instead of flow, but in principle could be used as well?

Re: product testing today – who is being served ?
 

yes, this would have to be simple to and past the point of losing some data

lets consider kit and component testing apart

kit testing:
1) some reality in "C" is essential, I would opt for grooving the IHS
not hard to qualify TC thermal connection
- cons: not cheap, groovemust be 'on spec'

2a) Watt determination is not possible w/o GOOD flow rates and high res/accuracy temps
-> by virtue of the needed equipment, I have no difficulty in saying that the calculation of the heat load is outside of the normal review site's capability
- need a dyno to test engines, eh ?

2b) quite agreed, if most reviewers understood the (unobserved) variations in their air temps I suspect they would not publish
-> but we could define a procedure, no ?

keeping kits apart from components, I see no technical difficulties in kit testing with a 'qualified mobo/CPU/TC TIM joint'

component testing requires a separate determination of essential parameters
where, for example, a procedure calls for the determination of the flow rate, differential pressure, or small temperature differences then I (again, sorry) would say clearly that such testing is outside the province of those not having suitable equipment

my thought was to define an acceptable means of testing with a CPU,
not at all to 'qualify' CPUs as a source (alone) that would then enable anyone to test anything, right tools for the job

Re: product testing today – who is being served ?
 

OK to simplify things for reviewers a bit...

At this point in time I actually see a justification for lots of reviewers doing "best they can" with different CPUs: SocketA still in usage, P4s, AMD64s, Opterons, and then dual cores. IHS or running naked, the list goes on and on. We can actually use the info to try and see where things stand better than any single die simulator testing could do.

But the dP vs. Q curves that we all desire for component testing are not system dependent, are they? Why shouldn't we just have a repository for this information in a format like the old Pro/testing site that all reviewers could link to and use? No sense in everyone being limited by that piece of equipment

PROVIDED their own flow measurement tools are in clibration I guess?

Re: product testing today – who is being served ?
 

here is the problem as I see it, the dP vs. Q curves cannot be (well) made w/o a bunch of quality equip
one could postulate a lesser 'grade' of equip, and guarantee shit results (been there, done that Thanks)
this type of testing is not, to me, the reasonable province of typical review sites

no reason not to collect and 'bank' the data

good reason to question data, look at that cal thread - and I am STILL in the middle of all that

Re: product testing today – who is being served ?
 

Yup just trying to remove that barrier of entry from testing.

Saves me some work too (unless I am the one stuck doing it)

Re: product testing today – who is being served ?
 

I am a bit late to the party, but I really like the direction this is going. It sounds like this could move review testing to an area where its attainable by more for less. Resulting in faster turn around on reviews, and more goodie for our readers.

Re: product testing today – who is being served ?
 

answering in no particular order:

any sensor type yielding an indicated 0.1°C resolution can be used IF it can be caled
but we should not recommend sensors with known durability problems, coolant sensors need to be in a moisture impenetrable sheath (ck Omega, not a deal breaker)

I still have reservations about a Swissflo
-> anyone want to loan me one and I'll ck it against a 0.2" dia mag meter (with head loss) ?

I think grooving the IHS is the challange
the groove depth and width should barely exceed the dimensions of 40ga TFA coated TC wire laid flat, a dwg is needed
note that the plane of the IHS surface from the edge to the center must be established to avoid cutting too deeply into the crowned central area
- the real dificulty may be to find a machine shop willing to do such as one-offs

I'm probably at the limit of being able to speak as I have no mobo testing exp at all
got a link to Incoherent's thermistor post/thread ? (I'm not clever enough to search here)

Re: product testing today – who is being served ?
 

The advantage of the DOW sensors is that the entire digital thermometer is contained in the transistor packaging. So the wiring solder quality doesn't matter one bit. This is really the big advantage of them. Well that and the whole logging can be done in one setup.

My Swissflow seemed to report numbers within spec of my GPI flowmeter over the 0.5-3 range. I had some electrical issues with it though and it is about 4-5 years old from Joe's old test system. I can see if I can dig it up and I'll send it to you but I am not sure it's a good measure of general swissflow performance.

They are not a problem to submerge in water, but the wiring IS. No factory sheathing for them; gotta rig something up.

Ya the grooving is gonna be a pita for everyone who tries it. Could we switch to a copper shunt instead like your old TIM testing? Would that be a dealbreaker? Not sure.

Re: product testing today – who is being served ?
 

You mean this? http://forums.procooling.com/vbb/showthread.php?t=12247

Re: product testing today – who is being served ?
 

What I found with the turbine style flow meeter is that if you dont filter the water, they seem to act as the filter. That was my second Swissflow, the first one got some hair in it or fibers from something and got gummed up.

Re: product testing today – who is being served ?
 

turbine or paddlewheel ? (the latter I think)
I and pH used a (larger - GPI) paddlewheel with some success, fouling is a problem though

but understand that a flow meter is useful only for WCing component tests which cannot be done w/o also measuring the pressure drop (witness pH's difficulties)
-> now a new difficulty; while 0.1°C resolution is sufficient for kit testing, is it sufficient for component testing ?
no
if I truncate my input data it yields 'noise' at low flow/power levels

while I see a need/place for sink and WCing kit testing using a CPU as a heat source, I am most unsure that a CPU is also suitable to test components
- I am not saying that components could not be tested by substitution into a system being tested as a kit (noting the difference etc.),
just that component parametric performance curves will necessitate equipment having more resolution (high-end, elsewhere)

In summation, I am suggesting that review sites not having a high-end calibrated test bench avoid parametric testing as they have not the tools
test cooling solutions, components only by substitution

Re: product testing today – who is being served ?
 

by components you mean Radiators, and Pumps correct?

I for one still want to see a good way of testing radiators... I mean I tried it before, you tried it before (better than me), but still there has to be a better way.

Re: product testing today – who is being served ?
 

link to the DOW sensor ?
ever cal one ?

that TIM test was terrible (by my stds of today),
no - really do need to groove the IHS if that temp is to be used

EDIT
yes, wbs, rads, and pumps
nothing about rad testing lends itself to casual activity, look at this review
http://www.phoronix.com/scan.php?pag...item=328&num=1
please give us your comments

Re: product testing today – who is being served ?
 

DOW Sensors

They are not bad for casual usage; I replaced all thermistors in my normal PCs with these for temp monitoring.

Re: product testing today – who is being served ?
 

Here are some of the DOW sensors:
Econo:
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/2795

"High Precision":
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/2812

Both will work with a CF633, CF635 and CF631

Re: product testing today – who is being served ?
 

The swissflow is a turbine style, and I had fibers or hair get wrapped around the central hub causing it to bind slightly.

Re: product testing today – who is being served ?
 

Hmm... I feel a little better about myself now... I mean theres about 2 dozen things I could point out that are wrong with that reivew.

I mean mine was even better ;) Atleast I showed who the competitors were...

Re: product testing today – who is being served ?
 

it would seem a TO-92 package could be potted into a 1/4"OD ss sheath,
lead extension is no biggie in that dia

anyone have some in a TO-92 package they want to test ?

how did you use these pH ?
dip them ?

Re: product testing today – who is being served ?
 

I have lots in TO-92 packaging Bill. Maybe 20 or 30?

You're gonna make fun of me when I tell you this but for water readings I just gooped them up as well as I could and then stuck them into a T. Not recommended for long term usage.

I found these most useful for logging air temps around the case really but if the 1/4" SS sheath isn't expensive then I'd add some in for water for sure.

Re: product testing today – who is being served ?
 

http://www.procooling.com/index.php?...icles&disp=103

thats the review where pH breaks down the TO-92 hackage ;)

Re: product testing today – who is being served ?
 

ok, send me a dozen
I'll get the sheaths and send you some
should we standardize on a potting compound ? (McMaster or Omega ?)
epoxy seal the end, heat shrink as the strain relief
wire ga ?

Re: product testing today – who is being served ?
 

Bill:
Matrix Orbital and CrystalFontz both sell these DOW readers to the water cooling/overclocking crowd. If you make a few and can explain the process then we can probably get them into their regular product catalog.

How short can the sheath be though? Those 6-12" lengths I get on RTDs and thermistors are a bit excessive for plumbing into most people's PCs...

//edit: The nice thing about the probes is that since the entire thermometer is in the IC that wire gauge doesn't matter. I think both mfgrs that sell them to PC users use wire similar to fans, but that's overkill. Whatever works and will not affect the reading in anyway. I've used wrapping wire without problem.

Re: product testing today – who is being served ?
 

I had some 3M potting compound I bought back in the DH2 days... it was supposed to be someof the best around. I bought it from Graybar.