Let's talk about waterblock reviews...

[msv]

There seems to be people asking for WB testing at flow rates up to 3 GPM.
OK, this is ProCooling, so I really shouldn´t be surprised by anything anymore, but, serioulsy guys:
3 GPM real life water flow, is that really common, or is a WB test at such water flows of a mostly theoretical interest?
regards
Mikael S.

[pHaestus]

There are probably a few people here pushing 2-3GPM through their systems. And then Since87's rationale for the higher points is improved log scale spacing. We get all kinds here I am really not sure I can get 3GPM with current pumps and GPI flowmeter (it adds a substantial flow resistance). Perhaps 2.5 will be doable though.

[BillA]

2gpm is attainable by careful component selection
2.5 not really so simple at all, 3 is pie-in-the-sky

pHaestus
you will find that many of the newer high flow resistance wbs will preclude you from attaining even 2gpm
- that was all I could get through the Innovatek XX (it also uses 1/4" ID tubing/fittings), and my pump is rather biggish

[BillA]

since87 has made a rather interesting post here

[pHaestus]

If I have to concentrate to extract the useful info from a graph then it probably is a bit too complicated to be of general use. Good thinking though

Bill would you propose that (C/W)/$ should be included? Makes the testing of international wbs complex...

[N8]

Damn, I wrote up a nice story here, and then it just disappeared into the fourth dimension or something, I'll try again.

Here at work, we need to cool a lot of scientific equipment such as lasers, diffusion pumps and X-ray tubeheads. We designed and built a whole new building to meet our needs (physical space, electrical, cooling, data transfer, etc.) in the last couple years and moved in early 2003. The entire building has a closed loop, processed chilled water system for hooking up to our equipment that needs it. There are maybe a hundred inlets/outlets, and quite a few don't have any equipment hooked up to them. One of the inlet/outlet hookups is right at my lab workbench (hmmm, wonder why....)

The chilled water system is 15C or about 59F, a good temperature to chill but avoid condensation. Depending on the size of fitting and hose, the flow rate is up to 10 gpm. If I use 3/4" ID hose I can get about 9+ gpm. I can't remember exactly what the flow is with 1/2" ID hose, but I think it was about 6 gpm or so. Somebody can probably do the math and see if that is a good number compared to the 3/4" specs.

The pressure in the system is about 90 PSI. I have flowmeters and pressure gauges. I also built a testing unit for variable flow rates to check pressure differences on each side (inlet vs outlet. It uses an adjustable flowmeter that I can adjust the flow from 0-10 gpm, with a pressure gauge near the inlet and outlet.

This setup allows me to set a flowrate and see what the pressure differences are between each end of a piece of equipment.

I have been thinking about using this for waterblock testing since we moved in, but it is just one of those things that I can't seem to find time for. When you have a family and kids, you don't have much free time outside of work anymore.

[N8]

I guess this would also be good for testing heatercore flowrates as well.

[BillA]

not at 90psi
and a pressure - but not flow - regulator is a bit tricky I believe

[N8]

Heh, yea I guess 90 PSI might be a bit much for a heater core.

[BillA]

Originally posted by pHaestus
If I have to concentrate to extract the useful info from a graph then it probably is a bit too complicated to be of general use. Good thinking though

Bill would you propose that (C/W)/$ should be included? Makes the testing of international wbs complex...

yea, that graph is a toughie, not intuitively clear at all

re (C/W)ƒ$
this relationship, perhaps factoring in the flow resistance as well, is really the only one of relevance to a mfgr
- and one would presume, were the info available, to the user as well

never been to sure how to set it up

[pHaestus]

What leaps to mind is that you can't just have a single point on this number either. Since if you want say the performance of a block at 2.5GPM, you'll have to spend more on an industrial pump and then more on a way to cool water effectively as well. Then you get into making some pretty specific pump recommendations too though.

I like x-y-z graphs but reviewers of my papers seem to HATE them (they always get changed to 2 graphs upon revision). Always made things seem much clearer to me though.

[Les]

Originally posted by pHaestus

Bill would you propose that (C/W)/$ should be included?..........

Not even sure you should include (C/W).
You will have W for one path of a multi path cooling system.
The Diode temp which is part of C is is in all the cooling paths.
Possibly needs a subscript(or something) to differeniate from Bill's "C/W"

EDIT: All above is crap.

[BillA]

I got TechPlot just for this, though I have yet to learn how to use it

3D topos are fun, but many dislike them
for what we are talking about quite necessary as the 'sweet' (cost effective) spot will be readily apparent

wb performance in a system cannot be considered in isolation from the pump
there are a number of wbs being sold which have a VERY pronounced 'knee' in their C/W curve; if put into a system whose pump can't push it, they will do poorly

Les
expand a bit on your comments

[pHaestus]

i think Les is saying that the CPU diode temp is the result of both secondary cooling and primary cooling of waterblock. So when I estimate W from x*dT*Q, I am measuring the heat dissipated by the primary cooling, but when I measure Tdiode then I am measuring the cooling of all paths.

Is this any different from a C/W that comes from JoeC's testbed though? He has a relatively large copper top on his die sim in the shape of a socket for heat to be lost...

I could call it C/"W"; or C/W* though for sure as there are lots of assumptions that go into it. I think they are reasonably well backed up with testing results though.

[Les]

Originally posted by unregistered

Les
expand a bit on your comments

Unless we are talking about JoeC's C/W
You have defined "C/W" in the manner "When considering a wb's "C/W" (the CPU-to-coolant temperature differential divided by the power), there are two constituent elements; the TIM joint, and the wb itself".from here
Then for the actual C and W you have used values from within the path from Source to wb.
It is this practiceand not the words that I have taken as the definition of "C/W"
OK in your Die losses from other paths are <2% and the difference is only small.
The losses/gains will be greater in system testing and I feel the definition should be tightened.

EDIT : Ignore all above, it is clap-trap.

DP
Basically yes .

Edit Still not mastered this link thing.Thik got it on 1st edit but....

[Since87]

Originally posted by pHaestus
If I have to concentrate to extract the useful info from a graph then it probably is a bit too complicated to be of general use. Good thinking though

I added a bit more (largely nontechnical) commentary in the oc-forums thread.

I don't expect the general watercooling public to understand that graph. It isn't even necessary for the tester to understand why it is a 'good' testing technique.

However, if someone wants to write "Waterblock Rankings for Dummies", and only provide a single C/W value, I think the method suggested is much better than just measuring the C/W at 1 gpm.

Most of the people looking for such rankings aren't even going to try to understand the underlying testing. I would think the tester would want to give his readers the most generally relevant data points he could, precisely because they aren't going to understand the rationale behind the testing.

Such a tester would also be much less open to criticism if the rationale behind his testing was available for everyone to see.

[Les]

Originally posted by Since87


However, if someone wants to write "Waterblock Rankings for Dummies", and only provide a single C/W value, I think the method suggested is much better than just measuring the C/W at 1 gpm.

Do agree, but only when the tester's techniques and undestanding are honed.
However, until then would like to see as much the raw data as possible.

[pHaestus]

On a related note Sean:

I am tempted to, along with the C/W vs. flow rate testing (and asap delta P vs. flow rate testing) also include results from a "typical" system. Something like Chevette core, 5' 1/2" tubing, Hydor L30. Yes the flow rate will vary from block to block and yes it is a step back from bench testing. But it may give "average Joe" readers the info that they really care about. Haven't decided for sure yet but it seems like a worthwhile time expenditure on the surface.

[Cathar]

My concern with "C/W per $" is that the cost of a block varies depending on country/currency.

For example, if you live in the UK, the cost of acquiring a Cascade costs barely more than buying a Maze 4 from a local UK store. If you live in the USA, it's almost double

In Australia there are some waterblocks sold (and manufactured in Australia) for which their local pricing is more than double the equivalent price in the USA after exchange rates are factored in, and this situation exists purely for discount reasons to better compete in the USA market.

The USA has seen its dollar value take a dive to the tune of a 40% loss in the last 12 months. Makes USA made blocks sold in the USA very attractive, but presents a fairly skewed view of the world as outside of the USA the ground becomes a lot more even.

What I'm saying is that depending on where you live in the world, the $$ value can range over about a factor of 2. To give "C/W per $" is really only useful to the people affected by the currency being quoted in, and assuming we're talking USA $, it definitely colors the results with a strong USA-centric bias, further colored by the currently low value of the US dollar.

Just my 2c.

[pHaestus]

A very valid point Cathar.

[Since87]

Les,

I too want to see as much data as possible. Possibly a bad idea to suggest that doing otherwise is 'acceptable'.

However, I've got mixed feeling in the case where a tester is going to do 'bad testing' anyway, and a simple change in technique could make it much less bad.

pH,

I think it might be a good idea to add the 'system' testing along with more comprehensive testing. Discussion of the meanings of the data might educate people who wouldn't even look at the data without the system test.

I dunno...

[pHaestus]

I do my best to provide data to those of you who can benefit from it. I learn much more from seeing how Les handles data than he learns from my test results I think.

[BillA]

Cathar
re the cost effectiveness
we in the US, and that northern state, can only gage cost effectiveness with the tools we have
I would say rather than USA-centric, "USA costs were considered"

I think the operative phrase is "your mileage may vary"
what is to preclude someone in Europe or OZ from recalculating based on their costs ?
- if they care enough they will do so

[Cathar]

Bill, in general I agree with you, but:

Originally posted by unregistered
if they care enough they will do so

I think that just about sums it up pretty well. As we both see over at OC.com, some people can't even care to plot a vertical line up a graph, even though they understand that it's probably important to do so.

Or witness the "Tell me what to buy" threads, and variants thereof.

Further, is the "C/W per $" factored into the block alone, or over the cost of an entire setup which is required to support that block to yield the C/W in question? What is the cost of that setup? Who is the target audience?

Just food for thought on the issue.

[Groth]

Re: C/W per $, not only is it localize in space, as Cathar pointed out, but it's localized in time. A year from now P vs Q and C/W vs Q will still be valid, anything with money won't.

There's only so much work will can squeeze out of pH, let's use him wisely.