Common Pump Flow vs Pressure vs Heat Comparison

Posted this at OCAU.




Some notes:

*) This post was designed for an Australian audience. Ignore the prices listed unless you want a good laugh about how much us Aussies get ripped off on pump pricing. 1 US dollar = 1.28 Australian dollars.

*) The Danner Mag 3 line is my personal interpretation of a best fit line given the data presented at their website. Their data produces a line that is not even remotely smooth, so I look the liberty to smooth it out somewhat. If anything, I may have overestimated slightly. The in-line heat figure is another approximation drawn from reading about people's experiences. I believe it to be somewhat close.

*) The Swiftech MCP600 in-line heat is another approximation, based on motive pumping power which must be ultimately converted as frictional heat into the water, and some heat from the motor itself.

*) The Johnson pump may be obtained for the price stated from www.depcopump.com. The in-line heat value is an approximation based on my experiences with the Davies-Craig EBP, with both pumps sharings a similar design.

*) The Davies-Craig EBP price quoted is an over-the-counter cash price direct from Davies-Craig. The on-line order price is $201 AUD. (http://www.daviescraig.com.au/newproduct_ebp.asp)

*) The Eheim pumps voltages/frequency is not given as Eheim release correctly spec'ed models for whatever country they are targetted for. This means that so long as you buy the Eheim pump that is correct for your country, you should see the performance stated. The Eheim PQ curves on the literature did not have uniformally placed graph lines. The PQ curves given are based on an interpretation of the graphs using the global min-max scale of the graphs presented, and then following the points on the curve. I believe that this actually eliminates some of the discrepencies with the visual interpretation of the Eheim PQ curves that people have occasionally reported on.

*) The block flow/pressure resistance curves presented are for the blocks themselves only. They do not include additional resistances that may be introduced by a radiator or other heat-transfer device.

Personally I feel that the big winner here is the Danner Mag-drive Model 3 for mains-powered operation. It has it all over anything else on that chart for price/performance.

For 12V pumps the winners are the Swiftech MCP600 for moderate flow, and the Johnson CM30P7-1 for higher flow-pressure focussed setups.

Regarding the flow vs. pressure curves for the Cascade and Whitewater, where did this data come from? Very cool - I wish more data were available for other more recent blocks.

Of course, one can wish in one hand and **** in the other and see which one fills up first. I'm under no delusion regarding the amount of time and effort required to get this data. To have any of this stuff available is a tribute to the time, effort, and generosity of many willing to share.

Very nice! What about adding the MD-15R for us? There are some (including me) that have picked these pumps up for $60 - $70 if you are patient and look around. Anyway, the chart is a very nice reference tool. I would think the MD-15R would be right between the Johnson and the Danner.

I dont understand these things.

Anyone want to give an example or analogy of what the numbers mean?

Does the intersection of the waterblock and the pump mean anything?

Huckleberry said:

Regarding the flow vs. pressure curves for the Cascade and Whitewater, where did this data come from?

Click to expand...

WhiteWater data comes from BillA's testing.

Cascade data is from my own testing, cross-verified with another's independent data. The Cascade can be a little strange as depending on how (mis)aligned the middle and base-plates are, the pressure-drop can very slightly.

nikhsub1 said:

Very nice! What about adding the MD-15R for us? There are some (including me) that have picked these pumps up for $60 - $70 if you are patient and look around. Anyway, the chart is a very nice reference tool. I would think the MD-15R would be right between the Johnson and the Danner.

Click to expand...

Updated graph with the USA spec of the Iwaki MD15-R. Press shift+F5 to refresh it.

BTW - My apologies to those of you who may have color blindness. Someone at OCAU brought that up.

The top two boxes on the right are colored for the two block curves that curve up from the bottom left.

The rest of the boxes start with the highest pressures on the left and go down from there. That way you can match the box/pump position to the curve starting from the left hand side.

BigSmokey said:

I dont understand these things.

Anyone want to give an example or analogy of what the numbers mean?

Does the intersection of the waterblock and the pump mean anything?

Click to expand...

The lines curving downwards from left to right are the pump PQ (pressure/flow) curves. They plot what flow rate you can expect from the pump against a certain back-pressure. Pressure, here, is presented as meters of a water column (mH2O). If you're unfamiliar with mH2O, then 1mH2O = 1.422PSI. Also 1 US.GPM = 3.7854 LPM, or 1 LPM = 15.85 GPH (note - per hour)

The intersection of the waterblock lines and the pump lines will tell you what flow rate you will get with that block attached to that pump (but no radiator).

A further aspect to the data to consider is the nominal in-line heat that each pump adds to the cooling loop when running.

This is fairly important. Just because one pump gives higher flow rates doesn't necessarily make it better.

For example, let's compare the Eheim 1048 to the Danner Mag 3.

The Danner Mag 3 is, on average, going to add 16W more heat to the cooling loop.

Let's say we were using a DTek Pro core with a decent fan running at a quiet 7V, and this corresponding radiator setup has an inherent C/W of 0.06, meaning for every watt of heat that enters the water, the water will heat up by 0.06C.

Now with the Mag Drive 3, the water will be 16 x 0.06 = 0.96C warmer than with the Eheim 1048.

With the Mag 3 & the WW, we would expect about an 8.3LPM flow rate (not accounting the radiator). With the Eheim 1048, about a 5.6LPM flow rate.

Now let's look at the White Water flow vs performance chart here:



At 5.6LPM, we see a C/W of around 0.183, and at 8.3LPM, a C/W of around 0.173.

i.e. a 0.01C/W difference. Now if the CPU were emitting a real 100W of power, then the extra flow rate of the Danner Mag 3 would mean that it would be 0.01 x 100 = 1C cooler. However, the water is 0.96C warmer due to the extra heat of the Mag 3, therefore all that has really been gain is 0.04C, or basically nothing.

If the CPU was dumping an unrealistic 150W, then it would be about 0.5C cooler with the Mag3.

If the CPU were dumping 100W, but we had a better radiator setup giving a C/W of 0.02, then the CPU would be 0.7C cooler with the Mag 3 over the 1048.

This is why it is quite important to match the radiator setup with a pump.

Is this how you spend your vacation? Why don't you go find a nice curvy section of road to throw you motor bike through at high speeds?

Other than that... I was wondering when somebody was going to put something like this together. Nice job. Even if you do discriminate against color blindness.

So with your dual "big arse" rads and 30RZ, how much cooler do you suppose you're running over a MAG 3?

UberBlue said:

Is this how you spend your vacation? Why don't you go find a nice curvy section of road to throw you motor bike through at high speeds?

Click to expand...

If I still had my bike, that's what I would be doing, believe me. Responsible fatherhood has forced me to give up the two wheels for a little while.

UberBlue said:

Other than that... I was wondering when somebody was going to put something like this together. Nice job. Even if you do discriminate against color blindness.

Click to expand...

On a note about the above, and factoring in a radiator into the equation.

I use Toyota Camry '87-'94 model Nippon-Denso brand heater-cores as radiators. With 1/2" copper tube fittings the core has a pressure drop of around 0.7mH2O at 10LPM.

So the trick here is to plot a curve of Pressure = 0.7 x (Flow / 10)² for the BA radiator.

You then add that curve to the curve for the block and the resulting line is what you'd use.

We can guess at it though. For example, with the Cascade, a Camry-core, and an Eheim 1048, the Cascade intersects the 1048 line at 4.7LPM/1.15mH2O. The pressure drop for the Camry-core at 4.7LPM is 0.7 x (4.7/10)² = 0.15mH2O

Adding the two together we get 1.3mH2O at 4.7LPM

Let's predict 4.5LPM => 1.3 x (4.5 / 4.7)² = 1.2mH2O

We look at the graph and see that the Eheim 1048 does indeed supply pretty close to 4.5LPM at 1.2mH2O, so that would be the predicted flow rate, and indeed this is borne out in reality.

UberBlue said:

So with your dual "big arse" rads and 30RZ, how much cooler do you suppose you're running over a MAG 3?

Click to expand...

I see about 9.8LPM with the Iwaki MD30-RZ, the Cascade and the two Camry-cores ("big arse") in series. If I calculate it, this is what I would expect to see as well, so that's another example of the above data coming into play by matching reality with prediction.

Replacing the Iwaki with a Mag 3, and I'd imagine I'd be seeing down around 6.4LPM through the system. That probably translates to around a 0.01C/W difference at the block. The Iwaki dumps 35W more heat into the system. The radiators, I predict, have a C/W of around 0.02, so for 35W the water is warmer by 0.7C than it would be with the Mag3, for a total CPU gain of just 0.3C.

However, and this is something that can't be predicted with temperatures alone. The Cascade does seem to manage CPU hot-spots better with higher flow rates. By this I mean I've had the 1048 attached to the Cascade, and even though the CPU temperatures were the same at about 2C lower ambient than with the Iwaki, the Iwaki still allowed for a higher overclock, which means that while the overall CPU temperature may remain about the same, the really small-scale temperature hot-spots on the CPU die are being better managed with the brute-force Iwaki. This is a story that raw temperatures alone does not tell.

Cathar i cant see the first pic link is broke

Edit: 1000th post!!!

Edit: Back now

slater3333uk said:

Cathar i cant see the first pic link is broke

Edit: 1000th post!!!

Click to expand...

My web host is temporarily down.

The picture is there again

This sure is informative - thanks for putting the effort into this Cathar

I am still very interested in the theory of "Pump Heat" output - I still don't understand how a 55W Iwaki pump is putting an actual 55W of heat into the water - shouldn't the number be less seeing as how that is the amount the pump *consumes*, and doesn't neccesarily *pollute* (put into the water) ?

posted by Cathar
.....though the CPU temperatures were the same at about 2C lower ambient than with the Iwaki, the Iwaki still allowed for a higher overclock, which means that while the overall CPU temperature may remain about the same, the really small-scale temperature hot-spots on the CPU die are being better managed with the brute-force Iwaki. This is a story that raw temperatures alone does not tell.....

Click to expand...

The temperature comparison you made between the Iwaki MD 30 RZ, and the MAG3 is particularly interesting - I find it strange that the Iwaki temperature isn't lower, yet a considerably higher overclock is achievable using it. Although your theory that the Iwaki's brute force helps "smooth out" temperature over the CPU die covers the occurance, it leads me more towards wondering about the value of having a high power consumption/high flow pump VS. a Low power consumption/medium flow pump (the Iwaki MD 30 RZ, vs the Swiftech 600 for example).

Also, how can one (fairly accurately) calculate a Radiator's C/W with any given fan, knowing the specs of both fan and radiator?

felinusz said:

I am still very interested in the theory of "Pump Heat" output - I still don't understand how a 55W Iwaki pump is putting an actual 55W of heat into the water - shouldn't the number be less seeing as how that is the amount the pump *consumes*, and doesn't neccesarily *pollute* (put into the water) ?

Click to expand...

The MD30-RZ's rated power draw is 80W. The 55W value is derived from observing it heat up a volume of water. The actual value may be a little lower, but not higher. I like to overestimate this value slightly to be conservative about the gains seen from larger pumps. It is well above 40W of heat in any event.

felinusz said:

The temperature comparison you made between the Iwaki MD 30 RZ, and the MAG3 is particularly interesting - I find it strange that the Iwaki temperature isn't lower, yet a considerably higher overclock is achievable using it. Although your theory that the Iwaki's brute force helps "smooth out" temperature over the CPU die covers the occurance, it leads me more towards wondering about the value of having a high power consumption/high flow pump VS. a Low power consumption/medium flow pump (the Iwaki MD 30 RZ, vs the Swiftech 600 for example).

Click to expand...

Indeed. The Iwaki is severe overkill. It'd be good with an evaporative cooling tower setup where water had to be pump up some distance to then free-fall. The evaporative effect would happily handle the heat load of the Iwaki, and it is here that the Iwaki would dominate all other pumps, allowing for a 7' high cooling tower, yet still seeing excellent flow rates where other pumps would be struggling to push any flow at all. This is one of the reasons that I bought the Iwaki, which is its potential for good use in such a scenario.

Evaporative cooling towers and/or oversized radiator setups aside, the Swiftech MCP600 has an extremely attractive set of characteristics. For a single CPU block and a radiator, it's perhaps the single best pump on that list, but its top-end flow rate is a little low. Given some lower restriction blocks, the Johnson CM30P7-1 would be a better choice.

Basically, of all the pumps there, I'd choose the Johnson myself if I wanted a 12VDC pump first, but it is pricey (~$150US), otherwise I'd always choose the Swiftech MCP600 (~$80US).

For a mains powered pump, the Iwaki MD15-R would be the number one choice, but again it is a little pricey (~$150US), otherwise the Mag 3 (~$45US) gets the nod for near unbeatable performance and value for money.

felinusz said:

Also, how can one (fairly accurately) calculate a Radiator's C/W with any given fan, knowing the specs of both fan and radiator?

Click to expand...

Not really possible. Have to measure it. There is an additional factor I left out above though. For heater-core style radiators, higher flow rates does tend to give better cooling performance. So while the Iwaki's and Mag3's may push more heat into the system than a MCP600, the radiators will actually be working slightly more efficiently. For larger pumps, the actual water temperature increase will be slightly less than what is predicted on the basis of the pump's heat. In fact since this improved radiator efficiency applies to the pump + CPU heat it is potentially possible, if the CPU is pretty hot, for water temperatures to actually be lower even after accounting for the pump's extra heat. This is again something which would have to be measured, and may throw a completely different complexion on the benefits of using larger pumps.

"For heater-core style radiators, higher flow rates does tend to give better cooling performance." Why doesn't it make a difference with the other kinds? (What ARE the other kinds? [Wait... Are those the ones with the tube wandering through the fins, as compared to all the little holes through which the water can flow?]) So, basically, the Mag 3 is an excellent choice, unless you can't handle the watts?

Cathar said:

*) The Johnson pump may be obtained for the price stated from www.depcopump.com. The in-line heat value is an approximation based on my experiences with the Davies-Craig EBP, with both pumps sharings a similar design.

Click to expand...

I can't find this pump anywhere at their site...

Also, it would be kind of neat to see the MD-20RLZT on there.

posted by Cathar
.....The MD30-RZ's rated power draw is 80W. The 55W value is derived from observing it heat up a volume of water. The actual value may be a little lower, but not higher. I like to overestimate this value slightly to be conservative about the gains seen from larger pumps. It is well above 40W of heat in any event.....

Click to expand...

Ah my bad - I understand now

In a larger circuit, using multiple waterblocks, and several radiators, I guess a high flow (overkill) pump with a high power draw would perform better (due to multiple radiators to help kill pump heat, and a need for a high flow rate) than in a simple CPU block/Radiator setup, where pump heat pollution is more noticeable.

Given that, I am really hoping my CustomSeaLife Velocity Titanium One will prove to measure up pump heat wise, and be useable in a PC watercooling setup. If it doesn't it will be returned .

After some testing, as soon as my comparison pump arrives, I guess we will see for sure.

Very interesting data. It's nice to see it all compiled together like this. I'm glad I got a Mag 3!

You talked about flow drop through a Big Arse and a Camry core - are those both single pass? I'd kinda be interested in seeing an estimate of flow rate drop with the popular Chevette cores.