My ported D5 Pump P/Q results

[Martinm210]

Did the tedious testing today after buying a ball valve to control flow and setting up a plumb water column stand, I was off and running doing my first P/Q chart. Collected 16 data points just as a check to see if any one was way out of wack.

Here is my ported P/Q graph against several others that Alex graciously provided me, gotta love Petra's for being so involved in the community

Anyhow, here is the graph. My first thought was...oh man, what did I screw up, that curve has a big wow in it...it doesn't compare well to the swiftech chart at all either.


I was darn near ready to scrap the whole test and start over, but I was being really carefull with a nice plub stand, multiple measurements, etc. So before I did that I figured I would go to laings site and look at their graphs again. Particularly their D5 brochure:
http://www.lainginc.com/pdf/LTI%20D5...-%206pages.pdf

I then sniped out the 1/2" barb chart and pasted it in. Ok that looks more realistic, I opened up the inlet and outlet, so actually lost strength when less than 1.2gpm, but gained really well after that. More importantly the Laing graphs also have a similar dip around 1gpm, my data is looking more plausible now.


Now on to the big barb comparison. I mean after all porting should probably be similar to a large barbed outlet right...YES it is...now I'm feeling much better about my results


So a question might be would this mod benefit real world loops.

So far I've tested out my GPU loop with MCW60,MCR320, 5/8" T-line, and 80" of 7/16" tubing and got 2.76 GPM, so it definately helps low restriction loops.

Now I'm curious about inlet/outlet nozzles. Is there a way to bulg the curve upward in the 1 to 3 GPM range, or is that an impeller thing?

FYI,
Here are those modifcation pics again:
Before:

After:


My question is...why is swiftechs chart so far off from Laing, mine looks much more like the Laing charts and I had to go to a 4th order polynomial equation just to match the roller coaster that it is...hmm

I'd love to try a D5 impeller with a couple more blades, I bet that would be just the ticket to match water cooling needs.

[Ad1tya]

Looking good!

Have you tried temp. comparisons?

[Quazi]

Subscribed

[bobo5195]

I would of expected that your low flow results when the pump is running well below its prime to be affected by porting.

You could increase flow by modding the impeller but then a different top would make a lot of difference, then you would probably need a better motor ending up in a completely different pump so its not really worth the effort but would be extreme.

The ideal number of impeller blades is supposedly 6-8 on larger industrial models but it is by no means a hard and fast rule.

[sick_g4m3r]

Excellent work!!! This is great stuff, I commend the time you put into this. This is a great D5 database right here.

I would expect the way to increase the system pressure at the beginning would actually be to make the inside of the outlet hole smaller versus wider as this would shoot the water up faster with more pressure.

It would be totally xtreme if you added a couple other fins on the impeller. You should definitely try that, and if that helped it would be amazing.

The only thing I am worried about however is that this would make the motor spin faster and become less reliable in the long run, which I really hope doesnt become an issue. Keep us posted.

[santiagodraco]

Nice work Martin.

In response to your doubts about the charts, have you ran the same exact tests on a non-ported version? I think that might make your comparison more valid than just by testing your ported pump alone. I always questions results that are based on distinctly different tests. Not that your tests are invalid but just that without the same condition results may vary.

[Jimmer411]

I really like where this is heading. Makes me feel even more comfortable with my choice of D5 over the DDC a few weeks back. Ive been thinking about chopping off my barbs and tapping them so I can put on some EK barbs, but if we could get someone to make a modded top instead Id be set

[1badhabit]

This is a great thread. Im a D5 owner myself, Ive been looking to mod it; just haven't really had the time, or really had any idea on what I wanted to do. But this is great!! Keep the ideas coming.

[MrToad]

Hmmm... There's something niggling me about this modification.

How does it affect (if at all) the structural integrity of the impeller assembly?

[MaxxxRacer]

this is one thing that has always been perplexing to me. Most high performance pumps have a small outlet, and when you enlarge it the head pressure drops off.

I dont know much about hydrodymaics, but the only thing that makes any sense is that the large outlet doesnt provide enough resistance for the impeller to build up pressure inside the impeller chamber.

[sick_g4m3r]

this is one thing that has always been perplexing to me. Most high performance pumps have a small outlet, and when you enlarge it the head pressure drops off.

I dont know much about hydrodymaics, but the only thing that makes any sense is that the large outlet doesnt provide enough resistance for the impeller to build up pressure inside the impeller chamber.

thats exactly right. i wonder how it would do if we made the outlet a smaller diameter

[Martinm210]

Thanks guys!

To answer a few questions.

TEMP COMPARISONS
No I havn't tried temp comparisons, Cathar already pointed out in the tubing sticky and I truely believe this that water cooling is fairly resilient even at low flows. I saw 10% in the CPU + GPU loop I was running, but 10% flow rate improvement isn't much for temps if I'm talking 1.5 yo 1.65 GPM. I'm still pretty weak on the thermal part of this hobby, but from what I understand this flow rate change won't mean much in temperatures because of how flat the thermal properties are on most water blocks. For example look at the Apogee GTX here:

Even if you were to improve your flow rate from 1 GPM to 2 GPM, you only gain about .005c/w, for a 150watt CPU that's only going to amount to just under 1C. What's important to see is how the curve slope really steepens as you approach .5GPM and lower, so not to take the wind out of my own sails, but flow rate doesn't amount to a whole heck of alot unless you start getting way down there under .5GPM.

This is just all about modding, tweaking, and going doing that little extra to get to a very small improvement. For me, I'm just having fun learning more about how this all works and this is how I learn it. Have an idea, try it out, and measure how it worked. That part's more fun than the actual benefit, just one of those things...

IMPELLER MOD
The only thing I'm willing to do to the impeller is maybe some hand sharpening of the blades, that thing is just spinning way too fast to be anything but perfect and adding a blaed would have to be a machine shop type fabrication which I could get access to and I do have a few years of machinist experience, but I'll leave that to someone that really wants to make money with something like this. If it was feasible to make an impeller for low cost that made a big improvement in this pump and manufacture it for a low price, that could possibly be a marketable item. My guess is that it would be overly expensive to fabricate to make profitable. I'll see if cleaning up some of the mass production edges makes any difference for me. The most obvious to me is the rounded leading edge of the impeller on the inside, I think if these were sharper you might get some overall gain.


NON PORTED TEST
No I unfortunately was too impatient as soon as I saw the first pump gain I immediately modded the second one the when it arrived, the only thing I can see is the shape of my pump curve matches well to the Laing curves, particularly the larger barbed version. I believe the Laing curves are accureate, the D5-38/810 is what I think what you get when you buy any variable D5 or MCP655 pump. I'm not sure why Swiftechs curve is so different Looks to me they too some test data and apply a second order poly to the data that forced the curve in one smooth sweep, maybe that's just the way they produce curves as practice. I'm going to recreate the Laing curve and have it for my flow prediction spreadsheet I'm working on as the base placeholder.


D5 vs DDC
The D5 and DDC3.2 are fairly close in the real world loop restriction range of around 1.0 to 2.0 GPM range, with the DDC with top still having the advantage. The port mod seems to have improved things after 1.2GPM making them even closer. Overall I think the DDC is still the better pump for most watercooling applications, the D5 doesn't show it's strength until you have a low restriction loop like the MCW60 plus rad, in that case the modded D5 is better. The DDC2 on the other had is a really strong pressure pump, it's really too bad this impeller is being replaced by the 3.2. What was clearly a substantially better pump is now only a little better with the new 3.2 in favor of reliability, I guess laing was just having to much loss in failures which is a smart move really, they still own the market, so fewer returns means more profits. They are both good pumps and more than adequate for most systems, the DDC is a little better at higher restrictions, and the D5's strengths are in low restriction. The DDC with top is still better, but rember how close they are and remember how little effect flow rate above 1 GPM has on temperture.

IMPELLER STRUCTURAL INTEGRITY
I'm not sure, but I would guess anything that's simply making the pump work more efficiently wouldn't have a negative effect on the impeller, but I could be wrong.

CURVE SHAPE
Yeah I don't understand pump design either, and have too noticed you basically have the ability to shift the curve around, but generally you give up on one end to improve the other. There is always some areas of efficiency you can improve like the sharp outlet, this is a result of Laing choosing to use the same interior mold for both the 1/2" barb and 3/4" barb, just a cost savings is all. I am interested in trying some reduction nozzle combinations out to see if you can bulge the curve more on the pressure side. A completely new top with more refined channels and tighter tolerances would probably help in the pressure side, but it's a complicated top to machine, it would almost have to be cast to make affordable. Not really sure why the smaller openings help the low flow pressue side, I guess friction acts as sort of a valve to help the impeller ratchet more pressure.

NEXT ON MY LIST
One thing that is certain to me is my non-variable pump is not as powerful as the variable D5. I want to make the same curve using the same exact top to compare motors. Not sure exactly why this is, but it is at least at the right hand of the curve. Plotting it out to do a side by side will help me understand what's going on there...maybe I just have an oddity?

Then I might try a few nozzle combinations to see if I can get any more out of a real world loop.

[sick_g4m3r]

NEXT ON MY LIST
One thing that is certain to me is my non-variable pump is not as powerful as the variable D5. I want to make the same curve using the same exact top to compare motors. Not sure exactly why this is, but it is at least at the right hand of the curve. Plotting it out to do a side by side will help me understand what's going on there...maybe I just have an oddity?

Then I might try a few nozzle combinations to see if I can get any more out of a real world loop.

Great post!! I cant wait to see you do this!

Also, the smaller diameter of the outlet would increase pressure like the way when you put your thumb over a garden hose.

[Martinm210]

Also, the smaller diameter of the outlet would increase pressure like the way when you put your thumb over a garden hose.

I think that's just sliding over to the left side of the curve, but I'll give some different options a try. I think it's more about what inlet and outlet pressures make the pump more efficient in the 1-3GPM range which is most critical to our needs.

I just don't understand how pump efficiency really works, guess that's what testing is for..

[Xeon th MG Pony]

Why are the charts so different? Simple: It is a safety factor they use to cover them selfs. Over clocking is NOT over clocking, what you are doing is using up the safety margin that is built into the system, they test a chip to its max stable operating speed, then they knock it back by X amount and label it as such, so you may have 500MHz to 1.5GHz head room, when you OC you are simply eating up that safety factor, just as with how allot of manufacturers will take the max a unit can perform then knock it back by X amount to give a safety factor.

When I design things I use parts that are 25% over rated, so any thing I build will have a 25% performance over what I would rate it for, so any one doing their own tests would find that their curve will be 25% over the curve I would give, that how ever, does not mean that it is a good idea to use it at 100% of its true limit as you loose efficiency and may shorten its life and such.

but thats the answer.

[Martinm210]

Interesting, my flow rate differences are now backed up by a max head test, I even used the same top on both pumps to confirm one modded top wasn't too different than the other.

Anyhow this is what I found:

MCP655 Variable (Ported Top 1): Max Head = 161.5" or 13.45'
Swiftech Sticker Model MCP655, Part No: 71310, Volts 8-24, Watts 3-33, Amps 1.9, MFR: 03/07
MCP655-B (Ported Top 1): Max Head = 126" or 10.5'
Swiftech Sticker Model MCP655-B, Part No: 72839, Volts 12, Watts 24, Amps 2.0, MFR: 04/07
I figure I lost about half a foot to one foot of max head with the top, so the swiftech chart and 11' head spec matches the -B model I have. I wonder if it's just the -B model or if Laing did the same thing with the D5 as they did with the DDC2 to DDC3.2 limiting??

FYI
I also stuck a really reduced nozzle in there (3/8" down to about 1/8") in the inlet and outlet of the -B model pump to see what if any differences there are with max pressure head.

Outlet nozzle - Did absolutely nothing still 126" of head
Inlet nozzle - Gained....up to 133" of head or a .58' or about a 6% pressure head gain.

If anyone does this mod I would leave the inlet alone, only port the outlet. I need to experiement with some inlet nozzles now and see what improves a real world loop.

So, I wouldn't recommend the 655-B model over the variable anymore, it's clearly weaker. In fact the MCP655 variable model is almost 30% stronger in max presssure head than the MCP655-B.

[Waterlogged]

In the course of my "discussions" with sick, I noticed that in Laing's PDF that the "B" (basic) version of the pump is locked @ setting 4. This is what is likely accounting for the difference.

[Martinm210]

In the course of my "discussions" with sick, I noticed that in Laing's PDF that the "B" (basic) version of the pump is locked @ setting 4. This is what is likely accounting for the difference.

Ahh...THANK YOU!!

I skimmed through that, and missed it completely, but I see it now. Now I just need to find the resistor that controls that and FIX IT

[Waterlogged]

Ahh...THANK YOU!!

I skimmed through that, and missed it completely, but I see it now. Now I just need to find the resistor that controls that and FIX IT

[sick_g4m3r]

Now I just need to find the resistor that controls that and FIX IT

Or raise it higher than the five setting. How would you know what to raise it too?

[Martinm210]

Or raise it higher than the five setting. How would you know what to raise it too?

Yeah, exactly my thoughts. I'm real weak on electronics, have done a little with resistors and LED, but I know a few basics.

The variable thing should be nothing more than a potentiometer or variable resistor to divide voltage. If the D5 is capable of running at 24V, you should be able to put a regular resistor in place that makes it higher than a setting 5 running at 12v without hurting anything.

I'm going to look into this more, I have one of each pump so with a multimeter I should be able to figure it out if I can get to the resistor, but I might need some help from some electronics gurus.

[Xeon th MG Pony]

it is a pot, rheostat have a distinct feel, and are rather dated. Since it is a brushles it is controlled via an oscillator thus it is a low current device, thus further lack of a need for a rheostat.

[Martinm210]

Thanks guys, I'll have to take some pictures up close and see if you can help me figure this electronics part out.

[Martinm210]

you should also record pump RPM on the data, for tells you how much more load was placed on the motor or lessened as per mod.

I'm only using a regular 12V PSU, but after I get it back in the computer I can record the RPM change.

[AndyM]

I'd look over Liang's documentation on the D5 before messing with that, seems these are frequency controlled. I read that the variable pump will hold a set speed irrelevent of the input voltage, to the limits of the pump operating at a certain voltage.