BillA , on your example of the pressure pulses in an engine, it has very little to do with the actual velocity of the charge in the Plenum. It does involve the pressure and vacuum pulses generated by the charge passing through either an expansion or contraction phase in the plenum, but the concept is based on sound pulses traveling through the fuel air mixture and arriving at the valves/port-entrance at various stages of the reciprocating cycle. This however takes for granted that the medium is able to be compressed to a certain degree..... two stroke engine exhaust design is a perfect example of this technology. It uses both the vacuum pulse and compression pulse generated by the exhaust pulse passing through the defuser section and baffle section, receptively. by changing the lengths and angle of these cones, one is able to generate a intensely impressive powerband, not too unlike that of an Turbo-charger on a 4-stroke..

yea, it was a poor analogy for unstable/transitional fluid flow

I do however agree fully, flow resistance in this place( water-cooling) is directly proportional to the mediums(water) velocity, there is no arguing over that one from me. It is in fact this annoying hurdle that steered me into trying to optimize block design for maximum turbulence, in stead of maximum flow. That way, the plumbing to and from the block will not be the bottle-neck, but rather the block.

Of course I am of the opinion that flow-rate not being used to generate turbulence, is flow-rate utterly wasted....

Laminar flow is not really an issue with watercooling computers. Laminar flow is more prevalent in very, very low flow rates. With the rates of flow we use in our systems, it is not an issue. All these blocks that claim better performance with turbulence are just marketing hype.

For the bulk solution perhaps, but there is always a boundary layer between the surface of wb or rad and the bulk water. At the surface (with proper resolution to see) the water should be fixed, in fact. Increased water velocity improves thermal transfer because it will decrease this layer. Increased water velocity also directly affects Reynolds number (the measure of turbulence). Hard to discern what block designs are improving performance due to increased surface area vs. actual changes in flow dynamics though.

#rotor your comments are as usual in line with mine. There are many block designs that would be extremely low flow and high resistance (Lytron?) that most likely will outperform these blocks that are "high flow" simply for the sake of bulk water movement.

On a somewhat related note, how costly is it to have you make waterblocks? I have the crude beginnings of a plan...

oooohh

the wb bug has bitten another

Back to the main topic. nikhsub, I have had a very similar experience to yours. I increased tubing at one point to 1/2" and also increased pump size. The overall difference was nil.

I've said it before, I'll say it again. In watercooling computers, once you get to a flow rate beyond 20-25 gph, the benefit of any higher flow is almost nothing.

You can show me all the charts and graphs you like. I speak from first hand experience as well as the experiences of many others who have made the mistake of adding more pumping power and larger tubing to their watercooled computers.

I will say that there are some applications that may benefit slightly from more flow, specifically those setups where flow was horrible to begin with. But once the minimum flow rate is achieved, no amount of pumping power or tubing enlargement is going to make much difference.

Cost is directly proportional to the amount of hoops you are going to make me jump through :) [I'm joking]

..... if it's a design I already made. then I'm cheaper than almost anyone else out there...(doesn't sound good, I know :) ) but when it comes to custom requests, I'm sure you will agree that cost can not stay as low as with a production model.

I however pride myself in being one of only a handful that actually has the ability to produce a waterblock on the fly, with information and design schematics, sent to me via email...... NASA can't, or won't even do that.... :D

Jim's, you might have a good point there.... using the turbulence factor as an sales pitch might be a good idea, only problem, almost everyone in the world seem to think flow-rate is what makes a block work.... very much similar to the predicament AMD has with regards to CPU speed. GHZ is not what makes a PC fast, it's what you do with that cycle, that will make it fast. Same goes for flow-rate/turbulence .

Jim
I'm not contesting the validity of YOUR experience, I do accept that you saw what you saw

but your conclusion
"In watercooling computers, once you get to a flow rate beyond 20-25 gph, the benefit of any higher flow is almost nothing."
is simply wrong

you are swapping several components in a specific system, and then trying to generalize a universal conclusion
you are wrong

I have tested radiators, waterblocks, have 6 different pumps, and have assorted tubing from 1/4 to 3/4 in. ID

flow rates above your described 0.3 to 0.4gmp can provide a measurable benefit
but it is the ENTIRE system that must be evaluated to achieve higher flow rates
NOT just swapping the pump and tubing
-> as any single component can effectively limit the ACTUAL flow no matter what the pump or tubing size

you did not identify your wb, rad, or if any 90s were in the system;
but if you go from a flow rate of 0.4 to 1.4gpm, you will see a substantial difference (if your measurement capability is functional)

#Rotor
quite agree with the 'virtues' of turbulence (and am fabricating a 0.7gpm @ 40psi system right now)

but your comments are masking a point:

for a given wb, higher flow WILL result in greater turbulence, a reduced boundary layer thickness, a higher convection rate, resulting in a reduced thermal gradient across the wb bp - hence lower CPU temps

yes, yes - wbs can be designed for more turbulence at lower flow rates
but even so the above statement is still valid

(why do you promote turbulence (= drag) also on the 'top' side of your wbs ? - ain't no heat 'up' there)

True, most of the heat will already be suspended in the liquid, resulting in the top of the block, almost not seeing any heat at all, with the emphasis on ALMOST... :) and I am somewhat of an perfectionist.... so as you can imagine, I very much do not like doing things halfway, even if the gains is only fractions of a degree...

You are making the point there, "for a given block". What I'm saying is that so many people are thinking of flow-rate, as being the mechanism at work to produce better cooling, hence the so popular spiral designs everywhere, Now I'm not saying the spirals ain't good.... They are good, very good. . You see, these blocks have been designed to make it easy for the liquid to get through.... how do you make life easy for water to get through??? you remove any turbulence generating characters in the design.... there is my problem.... what I'm saying is, don't let the block worry about flow-rate, the block needs to worry about getting the heat into the water, nothing more. if flowrate, or lack there of, is a result of the blocks design, the wrong thing to do, is to go open up the block, and take a dremel to it.... the right thing is to go buy a stronger pump...
heehe I sound like an American.... "if the car ain't fast enough, add more CI. to the motor".... :D

LOL, yes and we Europeans put in direct fuel injection, turbo charge 5valves per cilinder and twin spark, and take care of aerodynamics... :D

Hello,

This is the first time I'm posting/replying in this forum (although I'm not exactly new), so HI all! :)

About this setup, I just think the problem as mentioned be4 is the high flowrate trough the radiator.
In the radiator, the longest the water can stay in into it, the more it will be cooled down, BUT it is also true the faster the water can go though the waterblock, the more heat will be dissipated.

I think with other radiator for higher flow, less passes and more tubes, your new pump would outperform your new one, but as it was mentioned be4, you've lost your "sweet point" of your setup. The heat exchange in your waterblock is higher, but your radiator now it's less efficient at this flowrate. Just try to get another one.

Francisco

PD: Sorry for my bad english, but I'm not english :D

BillA, as my results are limited to my own system, I certainly agree that different systems may yield different results when it comes to flow rate. Obviously your testing is far more advanced than mine, and covers a wider array of blocks, radiators, etc.

My system is a chevette heater core 6" x 6" x 2", two 120 mm fans, one push, one pull, inline chillerblock, 500 GPH pump. There are two loops in the system, the above describes the hot side. The cold side is CPU wb, NB wb, inline chillerblock and reservoir, 500 GPH pump.

The main point I am trying to make is that once you get beyond a minimum flow rate, any further increase results in minimal gains. I believe even your own graphs show this to some extent.

NOTE: The testing that I did was with the chillerblocks removed and the radiator and reservoir used in a normal configuration. It consisted of the pump, CPU wb, NB wb, radiator and reservoir.

our difference lies with the definition of "a minimum flow rate"
had you said 1.5gpm I'd not have even posted
but 0.3 to 0.4 is far below what can be useful

but looking at your system description it is very clear why you see no benefit from higher flow rates:
you cannot achieve them

your 'chiller' is killing your flow (potential), and also any cooling benefit as the chiller 'can't cope'
-> chillers ONLY 'work' (and I misuse the word here) at very low flow rates
because their capacity is so limited

just saw your note

NB wb in series or parallel ?
(I would suspect this as a contributing factor)
and either way confuses the issue: in parallel diverts some of the flow, in series severely limits it

I agree 100% with that.

From all the data that BillA (and others) have shared, if I looked at a range of flow rate where the difference in the cooling ability (c/w) is less than 5%, it seems like 300 gph would be a good target flow rate, but that's still an off-hand observation. 300 gph effective flow rate is not easy (read cheap) to achieve, and that only takes into account the WB, not the rad.

It then became apparent that 300 gph is not only hard to achieve, but most blocks have a cross-sectional channel that is too large, which, although less restrictive, makes 300 gph a futile attempt.

300 gph is also quite useless to a rad. A heatercore would be too restrictive for that kind of flow rate anyways.

So I'm back to the old addage: either you use a high flow rate to achieve turbulent flow, or use a lower flow rate, but figure out the best way to induce turbulation in the water.

Then there's the fins...

Just check out the waterblock design thread I started.

300 GPH actual flow rate? With what pump? I would consider 1 gpm (60gph) to be a good minimum flow rate to design flow for. My testing is of course limited to "typical" blocks and so I am not sure what flat plate wbs need in terms of flow rates to be effective.

300gph (actual) is beyond the pale (translate: pointless)
2 - 3gpm is adaquate even for the Swifties

as #Rotor and others have observed:
its not the volume of flow that is as important as is the maximization of the turbulence potential of the available head

Occam's razor

Possible neither "volume of flow" nor turbulence should be maximised to get the best performance from "available head"
Using Kryotherm for designing a thereotical wb with a 50x50mm baseplate for an "available head" of 30KPa(10.04 ft water) the best designs I came up with were laminar flow.
For example for 90 fins this would be my choice from one Turbulent and two Laminar ::
http://www.jr001b4751.pwp.blueyonder.co.uk/LesD8.jpg

Links to the other 90fin designs:-
http://www.jr001b4751.pwp.blueyonder.co.uk/LesD7.jpg
http://www.jr001b4751.pwp.blueyonder.co.uk/LesD6.jpg


Note that the K/W( =C/W) is for cooling a 50x50mm heat source.

Edit: Deleted "However" from "However using ....."

deleted

Get a serck radiator and I think you will see much better temps :cool:

Do you know where to get one? I've got a running alert on EBay, and I haven't seen any.

Besides, according to BillA's revised numbers, the Serck isn't that far off from a heatercore.

U mean the numbers from overclockers.com:s radiator roundup? Anyhow. The serck will definately get the gph higher. And if I recall correctly it doesnt perform worse with higher flow?

Edit: I've seen what u mean now. The serck and big momma(heater core) perform on par @ low flow. So a bypass as i think u proposed should do the trick. Anyhow. I want a serck as i think u would put less stress on the pump and that gives u less vibrations. I know u can isolate the vibrations but i dont wanna put pressure on my pump as i find that more comforting.

Edit2: Come to think of it I find it hard to believe that a bypass would give u better temps but im surerly wrong :D

Edit3: Just came to think of it, the serck is made out of aluminium.... A standard heatercore is often made of copper/brass. I find it entriguing that it beats a copper/brass radiator with almost the same design (surplus from BillA:s round up). It would be fun to see the difference between a copper/brass lets say black ice extreme(if it indeed is made out of copper/brass)? And one made out of aluminum.

Edit4: just saw that the surplus radiator had about 50% less opening area..... Anyhow... still think the copper/brass vs aluminum comparison would be fun to see :)

Edit5: Since i live in sweden i think it would be kinda hard for me to find a serck. For the moment i have a chevette heatercore bought from www.dtekcustoms.com and i think i will stick with it for a while since i am running without a peltier and want my case to be portable(to a certain degree :D)

Well, you're not likely to find two heatercores, exactly the same, where one is copper, and the other, aluminium.

The aluminium seems to be best at transfering heat to air (from previous discussions) where copper is best for transfering heat to/from water. Many radiators have copper tubing, with soldered aluminium fins. That seems to be the best combination of metals.

Metals aside, the (internal) design of the rad is far more important, since the airflow that we're using is very low (relatively, in mass).

I've heard this is a myth that was created with the alpha coolers that were among the first hybrid heatsinks. The fact that they aren't entierly made out of copper is that they would weigh too much/cost too much. When i see the slk800 beat the shit out of the lets say alpha pal 8045 and swiftechs dito i begin to wonder. Sure heatsinkdesign has alot to do with it...

Copper vs aluminum is nothing more than conduction vs convection. Where conduction is paramount, copper wins by virtue of higher conductivity. Where convection to a lousy medium (read: air) is paramount, surface area tends to rule. Even though aluminum has a lower conductivity, when you consider the ratio of density to conductivity, aluminum is actually better. All this means is that you can get a lot more surface area from a given mass of aluminum. This extra surface area makes up for the lousy conductivity.

If you have two identical shapes, copper will win. If you have two identical masses, you can generate a lot more surface area with the aluminum and win a convection-to-air competition. This is the real reason for hybrid heat sinks. The fact that aluminum is cheaper and easier to work with also explains why you'll normally find it in applications convecting heat to air.

Well then its true that aluminum is not as good as copper :) So it was just a myth :cool:

more lives than the proverbial cat,
this one really does need a stake driven through it's heart
or, what's that ?

a silver bullet ?

I know... I wish it would sink in, so that I could at least remember the details of the explanation, for when I'm on the spot...

myv65, can i quote you on that one ? That'll save me a lot of time.

Did'nt the 'ALu gives up heat better than Cu' thing come from the fact that ALu WILL cool down quicker once the heat source has been removed than Cu?. size for size, shape for shape?...

ALu would cool down the same as Cu if they were of equal mass though right?,or store the same amount of heat?...