Optimal pump output characteristics.

Eventually i want to understand the importance of output characteristics of the pump to overall water cooling performance. I tried to search for such a thread or article, but did not find one that addresses all the chars. If i missed it' please give me a link, so i won't try to re-envent the wheel.

Pumps characteristics:
Pumps are normally rated in gallons per hour at a certain height. Feet pressure is the number of feet a pump can push straight up in a 1" column. A pump that delivers 500 gallons per hour at 5 feet will deliver 500 gallons vertically through 5 feet of 1" pipe. A pump that has a 'max head' number or 'shutoff' number refers to the number of feet a pump can deliver before it completely shuts down.
(from http://www.aquadirect.com/catalog/pumps/moreinfo/pumps.htm)
-----------------------
Some people say high flow rate of the pump is the important, another that the pressure on pumps output is the most imoprtant, and i've never seen considerations about shutoff of the pump.

From what pump will WC system benefit the most? (considering the output characteristics)
Please describe in details.

First, I suggest you do some reading here: http://www.thermal-management-testing.com . YOu should also wander over HERE as the real H20 'tech heads' play there a lot.

Now in plain english GPH of a pump matters very little. What we are after is pressure/head. Why? Because our water cooled machines are VERY restrictive, every fitting and turn and component puts more strain on the pump. We need the pressure, the pressure is what will give us the GPH we need if we select the proper pump. For example, someone with little experience would go after the pump with the highest GPH they could find, and probably end up with a low pressure, high gph pump that would be terribly inefficient in our systems. Iwaki makes some of the finest pumps available, they are expensive and there is a reason. Have a look at the Iwaki P/Q chart for most of their centrifugal pumps. Now keep in mind that anything more than the 20 series is a waste as it will dump too much heat into the system. With that in mind, which pump would YOU choose and why?

ahhh, the graphs are blinding me....


dude, just get the biggest pump you can afford/fit/tolerate... it should to fine....

so, just remember, bigger IS better....






want better temps and no pressure loss? get these, say like a pair of em' or something... and then run em' in serial so you get higher pressure....



edit: 4500 gal/min 280ft of head,... hmmm, yeah, should be enough to make a maze two perform like a whitewater with a eheim,... you know what i'm saying?

edit: 4500 gal/min 280ft of head,... hmmm, yeah, should be enough to make a maze two perform like a whitewater with a eheim,... you know what i'm saying?

Click to expand...

LOL

nikhsub1
Flow rate is the one that does the job, so if i could i would choose a constant flow rate pump. Didn't really see anyone suggesting to get such a pump. Why is that? Is it too costy, noisy, other?

If P/Q graph supplied it would be much easier to choose the pump. But most manufacturers i saw do not present P/Q. Most of them give water column height, ID OD and maximum flow rate. How can i only from these characteristics choose the pump?
edited: completely forgot. Many thanks for the links!


fafnir:
Great idea ... i'll try to find a PC case to put those babies inside ;-)

naw man, them blue things are submersible, you all you gotta do is to dig a big big hole and pour in some concrete, or use your swimming pool if you have one and cover that up,... then dump like say like 80 gallons of antifreeze in there,.. then cement the whole thing up,...

should be fine afterwards... might want something a little bigger or better than a dtek core w/shroud though.... it just might not be adequate you know,...


anyways,... orka,... its not that we don't want better performance, its just that for what we're doing here with trying to watercool a bunch of computers we don't really need to analyse that much data and all... i mean really, "optimizing" flow and getting a drop of another degree celcius is not really work that much effort... really... i mean, we all have budget limits too right? any sanity maybe? time? assuming if we don't then maybe we'll all jump prometia/cyro/ln2 recycling/liquid helium/open a mine to get more helium/ln2 submerged cray red storm with 10000 opterons?

that doesn't make much sense does it? "ooo, look, i can get 1000 fps with my cray and a gf2 mx 32mb sdr with ramsinks..." "thats 400 more than ur barton with a r9800"

um,... no, it really shouldn't matter that much once you hit the average of the bell curve,... cause going either way a lot more is going to be really really hard, time consuming, and won't yield too much,....

i mean, sure, if you found a cheap AND effective AND realiable way to do something new, sure, share the love, but if you're trying to show off ur new pair of "hidrostal industrial flood pumps" so you can impress some chicks to get laid... its not really going to work very well... if it does, you need to find a new chick... and runlike hell from the one that was impressed by the industral pumps... really....

its for ur own good and your electricity bill will thank you later....





edit: btw, i think them pumps are at 8000 watts each... i forgot, might be less... dunno....

Pump characteristics are confusing to many of us.... I Know they have had me off the wall a few times

But here is the thing: we need to maintain the flow rate to transfer heat- moving water does this. Flow Rate.

It is possible to have a pump with a very high flow rating that drops to almost nothing when there is virtually ANY restriction on it- look at any common water cooling systems pump
Pumping at a constant pressure would solve this, but these pumps tend to be more costly.....Pressure.

Pump Head gives an idea of how well a pump can deal with restriction: a pump with a 6' head will be impacted less than a pump with 2' head.

For those who can really make sense of the pq curves and xyz axis and equations like tpr2=6*r(pi+t) (or whatever they are ....) my explanation is bound to be completely insufficient

But it seems to break it down into terms that I can understand...

Well, I started out the post with the best intentions, got complicated, scrapped half the math, rewrote it to be simpler, didn't like that either, and then ended up with the following which I still don't like, but I hope it gets the point across and adds something to this thread...

With pumps, there is a good balance of power, flow and pressure, and the power part in particular must be balanced off against your radiator's efficiency.

First of all we want to establish a few basic crude points.

For waterblocks, you can basically say that when you double the flow rate, you gain about 10% extra cooling performance with the block. This varies and is specific to each block, but for now let's run with it as a rough guideline.

For flow rate, to push double the flow rate through some restriction requires at least a quadrupling of the pumping power.

Let's assume that a 120mm radiator has a C/W of around 0.06 with a moderate speed fan on board.

Let's also assume that half a pump's rated power makes it into the water as heat that the radiator has to deal with, making the water warmer.

Okay, so let's say you have a good waterblock that gives you around 0.20 C/W at 4.0lpm for a fairly hot overclocked 80W CPU. You're using an Eheim 1048 to power this setup, which is a 10W pump.

If we then replaced the Eheim 1048 with an Eheim 1250 pump into the loop which subsequently pushed 6lpm through the setup.

The waterblock will gain about 5% in performance. The C/W of the waterblock will drop to around 0.19C/W, for about a 0.8C drop in CPU temperature. The 1250 is now pushing an extra (28-10)/2=9W of heat into the water over the 1048. The water warms up by 9 x 0.6 = 0.54C due to the radiator's performance.

So the net gain in CPU temperature is 0.25C cooler.

So for our hypothetical system, we gain very little.

If we instead had a super-large radiator with some good fans on it, and the radiator had a C/W of 0.015, the CPU would be ~0.65C cooler with the Eheim 1250.

So here we can see that any pump size gains is highly dependent upon the radiator efficiency.

Without going through more math, I've basically worked out that:

*) for something like an Black Ice Xtreme you don't want a pump that draws more than 30W.
*) for something like a DTek Pro radiator 40W is the max
*) For a dual fan radiator around 50W is the max
*) For a triple fan radiator around 70W if the max

i.e. the optimal pump's power draw hinges on the radiator's performance.

Now that you've worked out how much power your pump should have at best, you need to find a pump that maximises the flow rate for that power draw.

This characteristic is highly dependent upon your system's flow resistance. A moderate to high resistance setup will benefit more from higher pressure pumps (Iwaki, Danner) while low resistance setups will benefit more from moderate pressure pumps (Eheim, Hydor, Via, etc).

So there's no real one optimal pump for all setups. Further, the gains seen by larger pumps can be quite minimal once pump heat is factored into the equation. Having said that, the best I've been able to determine is that if you live in 50Hz power countries (Australia/UK), the Iwaki MD-20R is one of the nicest pumps you'll come across. If you live in 60Hz power countries (USA) the Danner Model 3 or the Iwaki MD-15R are the two pumps who's characteristics best match up with what I'd choose. Just make sure that you use any of those pumps with at least a DTek Pro sized radiator and you'll be getting so close to about the best you'll do that you can stop worrying. Going with bigger pumps will likely send you backwards unless you have the radiator's to match, and further, you'll need to consider what the ongoing electricity costs are of running larger pumps.

but cathar, isn't there some kind of a minimum flow rate requirement for blocks like your cascade that are somewhat more or less flow dependant? (e.g. water jets or something?)

fafnir said:

but cathar, isn't there some kind of a minimum flow rate requirement for blocks like your cascade that are somewhat more or less flow dependant? (e.g. water jets or something?)

Click to expand...

Well I guess that there would be, but it is so low as to be a bit of a joke to hook up a pump that weak to your system.

I ran the Cascade with a 2.5W, 0.53m head, 200lph baby pump that measures 45x35x35mm in size (i.e. pumps don't really come much smaller/weaker than this - period) and it was basically matching what a White Water would do with an Eheim 1250 in my testing.

The jets are tiny, they truly are, and they only need to travel a very short distance. A jet velocity above around 5cm/second is all that's required to give some sort of suitable impingement action, and any velocity above that just increases the action's effect.

That baby pump that I was using was still pushing 1.5lpm through a full setup (2 heater-cores in series, Cascade, GPU block, 4m of tubing, filter). At this low flow rate the jet velocity is still up around 1m/sec (100cm/sec), so still plenty high enough for the impingement action to be effective.

It'd take a true dribble of flow (around 1.0gallons per hour) before you could say that perhaps the Cascade wasn't receiving adequate flow for its design. Of course by then you'd be facing other issues such as the water heating up substantially because the flow rate is so low. I'd also imagine that pretty much every other block on the market would be suffering horribly at such low flow rates too.

Cathar said:

For waterblocks, you can basically say that when you double the flow rate, you gain about 10% extra cooling performance with the block. This varies and is specific to each block, but for now let's run with it as a rough guideline.

For flow rate, to push double the flow rate through some restriction requires at least a quadrupling of the pumping power.

Let's assume that a 120mm radiator has a C/W of around 0.06 with a moderate speed fan on board.

Let's also assume that half a pump's rated power makes it into the water as heat that the radiator has to deal with, making the water warmer.

Okay, so let's say you have a good waterblock that gives you around 0.20 C/W at 4.0lpm for a fairly hot overclocked 80W CPU. You're using an Eheim 1048 to power this setup, which is a 10W pump.

If we then replaced the Eheim 1048 with an Eheim 1250 pump into the loop which subsequently pushed 6lpm through the setup.

The waterblock will gain about 5% in performance. The C/W of the waterblock will drop to around 0.19C/W, for about a 0.8C drop in CPU temperature. The 1250 is now pushing an extra (28-10)/2=9W of heat into the water over the 1048. The water warms up by 9 x 0.6 = 0.54C due to the radiator's performance.

So the net gain in CPU temperature is 0.25C cooler.
...
Without going through more math, I've basically worked out that:

*) for something like an Black Ice Xtreme you don't want a pump that draws more than 30W.
*) for something like a DTek Pro radiator 40W is the max
*) For a dual fan radiator around 50W is the max
*) For a triple fan radiator around 70W if the max

Click to expand...

From Bill Adams articles http://www.overclockers.com/articles481/ its clear that water flow arte does not really change the amount of heat radiator can dissipate. So why do you say that system with longer radiators will benefit more from stronger pumps? It's realy not clear to me...

The hypothetical example did surprise me ... didn't think that the difference is so small. Seems that in general there is no need to look for stronger pump unless it will be much-much stronger. Did i understand yoy corectly?

Orka said:

From Bill Adams articles http://www.overclockers.com/articles481/ its clear that water flow arte does not really change the amount of heat radiator can dissipate. So why do you say that system with longer radiators will benefit more from stronger pumps? It's realy not clear to me...

The hypothetical example did surprise me ... didn't think that the difference is so small. Seems that in general there is no need to look for stronger pump unless it will be much-much stronger. Did i understand yoy corectly?

Click to expand...

I'm saying that when you use a larger pump that pump heat becomes an issue. Pumps dump a portion of their motor heat into the water. The radiator is what cools this water, and so the radiator must deal with the heat. The more heat that enters the water, the warmer the water will be. The larger the radiator, the less the water will warm up due to the extra heat of the larger pump.

i.e. a small radiator will allow the water to get warmer than a larger radiator when a larger pump is used. So therefore a larger radiator is capable of making better use of the extra flow that a larger pump can provide.

Yes, you understood me correctly, partially. The differences are small, but a much-much stronger pump may actually make your temperatures higher due to the pump's heat being dumped into the water, and that heat overwhelming the benefit gained by the higher flow rates through the water block.

i.e. there is an upper bound on what size pump will be effective in your system given your radiator's efficiency.

The temperature differences when you use weaker pumps drops off quite slowly, as noted in the hypothetical example. Indeed in real world examples for me I can barely tell the difference between using an Eheim 1048 and an Eheim 1250 (<0.5C cooler with the Eheim 1250). It's only once you start getting to flow rates below the performance "knee" of your waterblocks will the differences become larger as you go to weaker pumps. The reason why here, again, is due to the pump heat. The block's performance loss with the lower flow rates is offset somewhat by the lower amount of pump heat entering the system, meaning you won't actually lose that much in final temperatures.

Cathar
I understand that stronger pump will probably add more heat to the water than weaker one, but ... matching the pump to radiator seems less important than anything else. I just don't see why pump should be matched to rad. I think matching should go this way.
1) first i think WB must be choosen.
2) 2nd is radiator matched to WB. It should be chosen so it will be able to dissipate enough heat colleceted by wb.
3) now pump should be chosen and it's output should be matched to the whole circuit, so it will be able to push enough flow through wb to be at least slightly above the "knee" of WBs.
*) reservoit/T-line decision should be made prior to everything, and eventually tubing should be chosen so it will fit the components and be as thick as possible.

Does such an order of matching the components makes sence? What would you change?

factor cost and reliability in there too,....


cause a bong setupup has a scalable rate of heat dissapation that is almost not dependant on both pump speed (rate) and pressure.... it also is unrestrictive if built properly, hence a larger pump will not affect such a setup negatively assuming the other parts of the system have no restrictions....


but reliability of such a system would suffer,... so we are talking about closed systems where there is a medium between the actual water carrying the heat and the air which it has to dissapate the heat into...

and also in a passive radiator, the equations might be different again because of the lack of airflow or the constant temperature of the air flowing over the material dissapating the heat.. e.g. air would heat up aroung the radiator... so pump heat would be more of a negative thing in regards to heat dissapation, cause now we have to dissapate the pump heat too... and since the overall capability of the amount of heat dissapated is less, we might see a more significant rise in temperature with a larger pump... etc


edit: anyone ever try a condensing bong setup?

Cathar,

Another hypothetical situation:

So ideal system would be a large radiator with a good c/w paired with a pump that puts out little heat... But to get this great c/w on the radiator it naturally has to be a bit larger... and if it's a bit larger, you need a pump that is more suited to pump through it... so upgrading the pumps means bigger and eventually dumping more heat into the water?

Am I undertstanding this correctly? Hmm, so this ideal system... what would you reccommend? Radiator wise? I saw you reccommened Iwaki pumps etc, and I will look into those. Interesting information, maybe a sticky?

<ag>

so no one here has ever tried a condensing bong right?

e.g. a big water container half empty with the intake as a mist jet spraying out and the container catching the water,... fan blowing down from above....

cause so far thats the most efficent thing i've seen, though its not too reliable...

fafnir said:

so no one here has ever tried a condensing bong right?

e.g. a big water container half empty with the intake as a mist jet spraying out and the container catching the water,... fan blowing down from above....

cause so far thats the most efficent thing i've seen, though its not too reliable...

Click to expand...

Start a new thread, don't hijack this one.

Question about matching the components remains unanswered. I think it's quite important, so i'll pump the thread up a little.

Cathar
I understand that stronger pump will probably add more heat to the water than weaker one, but ... matching the pump to radiator seems less important than anything else. I just don't see why pump should be matched to rad. I think matching should go this way.
1) first i think WB must be choosen.
2) 2nd is radiator matched to WB. It should be chosen so it will be able to dissipate enough heat colleceted by wb.
3) now pump should be chosen and it's output should be matched to the whole circuit, so it will be able to push enough flow through wb to be at least slightly above the "knee" of WBs.
*) reservoit/T-line decision should be made prior to everything, and eventually tubing should be chosen so it will fit the components and be as thick as possible.

Does such an order of matching the components makes sence? What would you change?

Orka said:

Question about matching the components remains unanswered. I think it's quite important, so i'll pump the thread up a little.

Cathar
I understand that stronger pump will probably add more heat to the water than weaker one, but ... matching the pump to radiator seems less important than anything else. I just don't see why pump should be matched to rad. I think matching should go this way.
1) first i think WB must be choosen.
2) 2nd is radiator matched to WB. It should be chosen so it will be able to dissipate enough heat colleceted by wb.
3) now pump should be chosen and it's output should be matched to the whole circuit, so it will be able to push enough flow through wb to be at least slightly above the "knee" of WBs.
*) reservoit/T-line decision should be made prior to everything, and eventually tubing should be chosen so it will fit the components and be as thick as possible.

Does such an order of matching the components makes sence? What would you change?

Click to expand...

The ordering you describe is correct, although the reasons, I feel, are a little awry.

1) Choose your waterblock
2) Choose your radiator - you want the largest one that will conventiently fit inside your setup. The amount of heat that a radiator dissipates is not a static thing. It varies and is proportional to how warm the water gets. A larger radiator means cooler water. A smaller radiator means warmer water. There's no such thing as "matching" a radiator to a water-block. All you're changing is the final water temperature when you change the radiator. One other thing which radiator's may affect here is the water-flow resistance that they offer.
3) Choose your pump - it must be matched to the radiator as well as the waterblock. Even pathetically weak pumps will typically deliver adequate flow to a waterblock. The larger the pump the better, but the size of the pump becomes limited by the size of the radiator. You can buy a bigger pump, but if the extra heat that it pushes out causes the water to become warmer by a larger amount than the waterblock improves by with that extra flow, then the pump you chose to too big. The amount by which the water gets warmer due to the pump's heat is directly determined by the radiator.

oh man, i knew there was something wrong with that eheim 1048 in my koolance.....

but great, there goes all hope for that set-up, ehh,... better get a cascade or something...

how's the hydra coming cathar?