Those results are for the block Cascade. But with some more restrictive blocks ( like those we use in Europe such 1A or nexxos ) the best choice is for lower pumps.

For a block like 1a-hv2, the best choice would be the MCP650, MCP600 and probably the DDC and CSP750 instead of MD20.

For the cascade, the DDC seems to have the same figures than the EHEIM 1250 (about 5.2 lpm and 9W) but with a more restrictive block, this lpm are less affected than for the EHEIM or IWAKI.

this topic is very usefulll to give an idea of why bigger pumps aren't the best choice, but I think you should put your figures into concept and say this results are valid for the cascade or others blocks with the same C/W and restriction, and that the best choice could evolve depending of the type of block (less or more restrictive).

It would be greet to do the same thing with many representative blocks, but I know data are missing.

/me glances up and notices that I actually recommended that the MCP600 @13.8v was the best single pump of all the above scenarios. The MCP600 @ 13.8v has about the same peak pressure head as the Laing DDC, but slightly more flow. Only if the MD-20RZ were a DC pump then it could possibly be better than the MCP600.

No, nothing different at all between what you are saying, and what I am recommending.

The slope of the pressure vs C/W curve for a particular waterblock is what also defines which pumps that a certain block might favor. One cannot base the distinction of which pump would be better simply based upon a block's restrictiveness, but rather how steep the pressure vs C/W gradient is.

For the thin-based blocks like the NexXxoS XP, and the 1A-HV2, these blocks actually have a lot more in common with the Cascade in this manner than you may realise.

I told about DDC, CSP and also EHEIM. The MCP are "polyvalent" so their good results are as good for low restrictive blocks than for very restrictive blocks.

But in your post, you put a sort of "ranking", but this ranking could evolve depending of the type of block. The MCP are great and won't move, but for some blocks we are using in Europe (seems to be 50/50 between very restrictive and low restrictive block here) the EHEIM 1048 and the DDC (which is equivalent with EHEIM 1250 for the cascade in performances) will take better rank instead of the MD20 or others IWAKI pumps.

I just want to put in concept your ranking and not saying

in all cases.

For very restricitve blocks, I think DDC and MCP650 have a better "profil" than many powerful pump like the MD20.

Edited my post above - please read it again for comments.

By "Performance", this also includes a large and highly effective radiator, such as a triple or quad fan radiator. I guess I should have made that more clear.

Are you sure about this? Have you plotted the PQ curves for these two pumps with this bloc? And then run through the math for dT based on pump heat, etc (as Cathar has done)?

[perso, j'ai un gros doute :p ]

I suspect that there are some low restriction waterblocks for which the 1046 is not as "below the knee" as you mention.
The curve for the Swiftech 6000 comes to mind - there seems to be very little additional benefit as flows increase.

C'est mon petit doigt qui me l'a dit :p

I have the PQ curve of the DDC and the EHEIM 1250. They cross themselves at about 5.2 lpm. (first I estimated the PQ curve of the cascade from the cathar's results with the EHEIM 1250,1048 and MCP650 and I used it with the PQ curve of the DDC, and I saw 5.2 lpm like for the 1250)

I used the power/flow curve from rosco to determine the heat given to water by the pump (I used the same efficiency than cathar with the MCP650 alias D4) and found about 9W

5.2 lpm, 9W, same figures for the DDC and the EHEIM 1250, so same perfomances :p

Buzz, you have the PQ curve for the DDC at 13.2v.

Somewhat different, and lower, at 12v where the block tops out at less than 6LPM, and peak pressure is ~20% lower.

Using the 13.2v PQ curve is fine, but be aware that it is a non-standard voltage in PC's and must be achieved through an externally powered means.

I always found it somewhat odd that the DDC was "originally designed for PC cooling", yet somewhat misleadingly all the PQ curves and values reported are at its maximum 13.2v level.

Still, even if the DDC is the equal of the Eheim 1250, that still leaves a fair amount performance gain to be achieved through the use of a stronger pump for blocks.

The pressure vs C/W curve for the Swiftech MCW6000 is very flat in comparison to many other blocks, and I would hazard an educated guess that it is even flatter than many of the European high-restriction low-flow blocks.

Too late now, but I'll run some figures with the MCW6000 tomorrow.

(semi)ok Buzz: I dont think the DDC puts out 9w - it only just absorbs that much. (not that it would make much difference at this level)
My little finger does agree with yours in that it will be a different scenario for a 1a-hv2 type block though ;)

So the best noise/performance choice would be two 1048 pumps in series? Or do they become a lot noisier when placed in series? Does anyone have experience with 2 1048's in series?

Probably not. Even if the flow curve is perfectly flat the transistion from a 1046 to a 1250 looses by just .3C (remember we're dealing with a 9 w pump here, which isn't a huge problem for a large-ish radiator). If theres even a slight downward curve on the C/W per flow curve, the block will pick up the couple thousands of a C/W needed to offset the 7.5w gain.

Of course, if you assume a small enough radiator, all bets are off :)

This is a very nice guide! The only thing I question is the estimation of heat dumped directly into the water by the pumps.

which is exactly the basis for this whole exercise
at 3 gpm and 150W I'm looking at a deltaT of 0.20°C, so at 10W the expected deltaT would be 0.0133°C
at a more reasonable/measurable flow rate of 1.5 gpm the deltaT is 0.38°C, so 10W would be ~ 0.0253°C
I would not be too uncomfortable at differentiating between 5W differences, but lesser increments would not be 'measurable' with the equipment I have

next time I'm into pump testing I'll gin out some measured values

So, you're saying it makes sense to go with the 1250 because it might perform about the same as a 1046 in this situation?

Cathar,

Would you still maintain your recommendation «Single 120mm radiator: First choice: Swiftech MCP600 or AquaXtreme 50Z. Silence Choice: Eheim 1048» for a radiator that followed somewhat the ideal specs you mentioned in a topic sometime ago: 160x160x32mm, or 160x160x45mm?

Cathar
Not happy with the data.
Particularly the assumed constancy of heat imputs
Are the "pump heat" to water measured or a dream?

Eheim 1048, 1250, MCP600 @ 12.0 and 13.8, MCP650, Iwaki MD-30RZ are all measured values, since I actually had those pumps in hand.

Eheim 1046 is an extrapolation based on the other Eheims.

The rest of the Iwaki's are extrapolations based on a 63% of rated input figure as determined from testing the MD-30RZ, due to the common shared separated magnetic armature design. The MCP600, which is also of similar design, as well as the March 893-09 that I have here, all have a ratio of in-line heat to pump motor input of around the 65-70%. Given the strong trend, I felt it safe enough to run with those figures.

In the full article I had already explicitly explained that which were measured values, and those which were extrapolations.

I really wanted to include the Danner Mag 3, but I did not feel comfortable with assigning a pump heat transfer value to it without either measuring it, or having a number of other pumps from the same range to extrapolate from.

With a Cascade + rad in the circuit?
Not happy

With a Cascade, no rad.

Not happy? Get over it, or do better.

Happy that info is presented.
Always been a poor experimentalist, and now being slightly physically limited present zero.
However always hope for pure data

Les, I do know where you're coming from, although I do get a little frustrated sometimes with the perennial suggestions from you that I'm making this stuff up as I go along.

I'm happy to admit what's extrapolated/estimated vs what I measured. I am also aware of the variance of pump heat output as the flow rate changes. I chose to measure with a Cascade in the loop to provide a single value that I could use. Yes, it does not strictly apply, but in what I measured for the various pumps it offers a pretty good mid-point between stupendously blocked flow (trickle) vs open-flow, with those extremes varying by around 10%, and at most 15% from the values I measured with the Cascade in the loop by itself.

I don't have the time, or really the exacting equipment, to generate detailed heat vs flow graphs, so yes, I did compromise here and settle on a catch-all figure. Even given the heat vs flow variations, we're still only talking about 0.1-0.2C variations at the absolute extremes of the flow rate ranges for the >50W pumps, from what would be calculated given the simplified "mid-point" heat transfer figure.

As a cooling god you have to be extra cautious.
For example
I am curious about the statement "C/W"=0.003 fora 3.5mx3.5x2.5m room .
With brick walls i would make it more like 0.03c/w

F**k I hate the "cooling god" appellation.

No double brick walls here in Australia (well, not many).

0.003 with the door open accounts for it. Just a really, really rough measurement.

Close the door, and double brick the room, then yeah, 0.03 would probably be more accurate.

Apologies, didnt notice the 'knee' was very similar - i was under the impression it was steeper than the cascade by quite some margin.