Roscal does kits, hi and lo flow too

[BillA]

I think I agree, let me dig a bit

[Incoherent]

Originally Posted by Les

Maybe "Thermal Effectiveness" should be used to describe a radiators performance.
Is suggested inDecember's CoolingZone, by Zahed Sheikh, as "A Useful Measure of Cold Plate Performance".

Les, this article is bugging the shit out of me. I am probably misinterpreting it, ...perhaps I need to drop the assumption that "more effective = better performance"... thats annoying.
Will think a bit.

[BillA]

a mix-up in terms I believe
the article describes (as I recall it) systems effectiveness, and then tries to draw conclusions wrt components
(and what Les is suggesting - I'm on the fence aka rebuilding a part of the WCing test circuit)
- the pump heat/head loss thing recast slightly

[Les]

Link posted to indicate that authoritative(?) thermal engineers were noting the importance of the coolant's exit temperatures.

Think, I still prefer a device to be characterized by (C/W)inlets or (C/W)wo-ai at given flow-rates(Q)
Where
C= dTwi -ai or C= dTwo-ai (Not LMTD or part LMTD)
W is the change in SENSIBLE Heat of the coolant= Bill's "Heat Dissipated" = My "Ww" = dT(wi-wo)*Cw*Q
Think "Sensible Heat"of coolant maybe prefereable to Energy of coolant.

[BillA]

indeed, sensible heat is what is being measured

my problems were caused by temp compression by excessive thermal capitance in the loop, specifically at the crosses
- replaced all with some (fabricated) plastic bits, and all is well again
(undocumented was the fact that I'd used plastic crosses when doing these tests, which I still have; bleh 5 days wasted in hunting the problem)

yes Les, "C" is now taken inlet to inlet (by me anyway)
BUT I have found that a single temp at the fan inlet is rather inadaquate, the error can easily exceed ±0.5°C
- of course this depends on the desired precision

[Incoherent]

Originally Posted by Les

Link posted to indicate that authoritative(?) thermal engineers were noting the importance of the coolant's exit temperatures.

Think, I still prefer a device to be characterized by (C/W)inlets or (C/W)wo-ai at given flow-rates(Q)

I totally agree. The Win-Ain is intuitive, the Wout-Ain gives an instant feel for effectiveness.

The article is indeed about System level design decisions I guess. However I am having trouble figuring out which he is referring to with "Effectiveness", system or "cold plate".

When the effectiveness is large, the coolant entering the radiator will be hotter. The difference between Tout and Tamb will likewise be larger, meaning that more heat can be dissipated from the radiator to ambient. The system consequence is that the radiator be made smaller and cheaper, with lower pressure drop and lower fan noise.

I have trouble reading this and and not reacting... "! how many times does it have to be said, the temperature rise across the waterblock has nothing to do with waterblock effectiveness, efficiency, thermal resistance whatever, but ALL to do with the power being dissipated, the flowrate and the coolant fluid type...

Also, as the effectiveness increases, the temperature rise of the coolant increases. According to equation (2), a smaller flow rate is then needed for the given amount of heat absorption, Qactual. This will reduce the pressure drop in the entire loop.
With lower mass flow rates and pressure drop, a smaller, cheaper and more reliable pump may be used.

Edit: Apologies. My 2 year old posted this, which was an initial layout of my thoughts, not intended to be posted. Will leave it for now, is not too insulting I hope.

[lolito_fr]

IMO, the author makes a bad start when applying Equation (1) to a waterblock:
Eps=Qactual/Qmax
...
the die is not a constant temperature source!!
Perhaps someone should email this guy!?

[Les]

Equation (1) is OK.
"Efficiency = Qactual/Qmax" is a definition.
Subsequent logic, as far Equation(6), seems correct.
However that is as far as I go.

[Incoherent]

Originally Posted by lolito_fr

the die is not a constant temperature source!!

It could be considered a constant power source.

I am still very puzzled. There is nothing wrong with the logic, the problem is the point.

I think he means that the components of the system should be selected to not exceed a given die temperature, but be spec'ed as cheap and reliable as possible whilst conforming to this criterion. A high performance low flow block helps achieve this at higher water temperatures - meaning a smaller rad, lower flow pump, quieter fan.

I don't like the way it is presented though.

When the effectiveness is large, the coolant entering the radiator will be hotter...

and

Also, as the effectiveness increases, the temperature rise of the coolant increases.

leaves one with the suspicion that either he doesn't know what he is talking about or the application doesn't apply to us. The "Effectiveness" is increasing because Q actual is increasing within the framework of constants (i.e. constant die temp) he has selected. Reality is not like that unless the waterblock was trying to cool a large mass.

It is possible that the reader's an idiot.

[lolito_fr]

It could be considered a constant power source

Assuming that, isn't Qactual=Qmax? (neglecting secondary losses, as usual)
In which case, based on (1), (2) and (3) we end up with Tout=Tdie ???

(must admit to not having spent much time on the article, so perhaps this reader is also an idiot)

[BillA]

I spoke with the author several years ago trying to get him to accept one of my (puerile) articles
he is not a WCing guy at all, but like most others in Thermal Mgmt have the assumption that they can cool anything
- and the results of this assumption can be seen in old Lytron products, the Hydrocool by Delphi, and others
but they can get it right also, Delphi in the G5 for Apple - the biggest WCing app to date

the only thing I extracted was that the author was arguing for downsizing the pump with adequate system performance,
not too useful for Swiftech in pursuit of performance per se (for the DIY market)

[Incoherent]

Originally Posted by lolito_fr

Assuming that, isn't Qactual=Qmax? (neglecting secondary losses, as usual)
In which case, based on (1), (2) and (3) we end up with Tout=Tdie ???

(must admit to not having spent much time on the article, so perhaps this reader is also an idiot)

I think you are right if you take that approach. Meaning that Effectiveness is always 1.00!?
Unless die temp is fixed (a thermally superconducting block of infinite mass) and thus able to deliver whatever Q is defined by R and Tin...

It's annoying me.
The conclusion I draw is that Waterblocks are more effective at lower flowrates, despite the fact that their thermal resistance is better at higher flowrates... !!?

A high effectiveness cold plate has a high liquid exit temperature and requires less flow rate for the same required thermal resistance. As a result, the liquid pumping power is reduced and the size of the radiator can be much smaller.)

Perhaps sums up his approach. I believe exit temperature is a consequence of flowrate and heat transferred, not cold plate effectiveness. But to maintain a given die temperature, the flowrate can be reduced with a lower resistance waterblock.

And one thing he is spot on with, is that the radiator can be much smaller with this approach. That might actually be the point...
<a sudden glint of understanding>
...I'll have to think some more, this might actually be a rather large point....

[BillA]

in his system the pump is the only heat source that is a variable, so downsizing it then permits downsizing the rad
this is useful info for larger sized systems (for one with little experience), but in WCing 'reasonable' pump choices do already consider the pump's heat input as a part of the selection process
- in any case it is moot for DIYs as they do not do WCing systems design, though they may bumble about at it

[Incoherent]

Originally Posted by unregistered

in his system the pump is the only heat source that is a variable, so downsizing it then permits downsizing the rad

Am thinking more in terms of the "required radiator C/W" to maintain a given water temp at high versus low flowrates. A quick calculation surprised me, need to recheck my numbers.


I am definitely bumbling.

[BillA]

"required radiator C/W"
that's gonna be shooting in the dark for sure as the rad C/W varies with both sides' flow rates, heat load, temp differential and I've no doubt missed several

this is why the 'real world' approach is to identify the rad as a limiting factor
-> then oversize the hell out of it (the DIY approach as the LOWEST CPU temp is the goal)

other systems can be designed to provide rather specific performance under defined conditions, in which case the rad and/or wb performance (size, cost) will be downsized to achieve a cost effective solution that is sufficient for the intended task

[pHaestus]

Exactly

Very few DIYers are interested in IC temperatures of 10% less than maximum mfgr specified operating temp, but if you are optimizing a system commercially then that makes perfect sense. Cooler temps really are for bragging rights, and lab/industrial equipment rarely puts much premium on that (other than brand name of course).

[lolito_fr]

Think I have clarified my mis-understanding:
Qmax is the theoretical limit, which would apply if Tout=Tdie.
For some reason I was thinking that Qmax was the actual die power disp.
doh!!

[EMC2]

*sneaks in when nobody is looking*

Hmmm... read the linked article and the discussion on "effectiveness"...

Given two choices, I'll choose that Sheikh did a "ppj" at conveying the core of his thought.

*digests the article and gets heartburn*

hmmmmm.....

*gets a wicked gleam in his eye*

Well heck, I get it now....

*builds a pump flow control system slaved to the inlet temp of the radiator and die temp of processor*

*sets control loop to adjust flow rate to maintain rad inlet coolant temp = die temp*

*makes a kit with small pump, small rad, and his new super-duper control system*

*sells it as most "effective" water cooling kit on the market*

muhahahahahaahhahahaha



*casually drops a note of serious content on the floor*


Summary of what I think his "core thought" was :

A measure of the relative thermal efficiency of a test group of heat exchangers (be they cold plates, blocks, rads, etc.), which is directly proportional to the amount of energy removed from a system under an identical set of conditions, can be measured by the deltaT across the exchanger (given the identical test conditions). Given more efficient components, a specified deltaT between the heat source (CPU, GPU, mem, etc.) and the ambient environment might be obtainable with lower coolant flow (both air and liquid), which could lead to smaller size and less noise ($ maybe, maybe not...).


Comment1: *Optimal* efficiency is quite rarely, if ever, 1.


My other 2 cents: As shown by my mad scientist device created before a moment of lucidity overcame me, thermal efficiency alone, *as defined as he did*, accomplishes nothing (except maybe a core meltdown). Ultimately the desired goal is to maintain the lowest temperature at the energy source (in this case the component die temps) in the most efficient manner (different people apply different metrics to the equation... noise, $, size, etc., etc.) within a given set of constraints (although some place no constraints other than "is it possible"). That having been said, if one examines the dominate factors in the equations for fluidic heat exchangers, there is much to be said for velocity and turbulence, which come at a cost


*kicks note into the corner under the end table and hopes nobody notices it*

*waves hi on the way out*

*skips off into the darkness of "lurk-dumb" again cackling madly*

[Jag]

Originally Posted by EMC2

...
Well heck, I get it now....

*builds a pump flow control system slaved to the inlet temp of the radiator and die temp of processor*

*sets control loop to adjust flow rate to maintain rad inlet coolant temp = die temp*

*makes a kit with small pump, small rad, and his new super-duper control system*

*sells it as most "effective" water cooling kit on the market*

Isn't that somewhat similar to what Apple did in the G5 with the monitoring software?

[snowwie]

is the advantage to controlling fan speed analagous to controlling pump speed? is peak (12v) speed THAT noisy for the DDC? is 7v on the DDC that much quieter?

edit: is there another advantage to a pump speed controller that I am unaware of?