[Skinnee Labs] The Pump thread...

[davidm71]

I doubt you getting more then 2.5-3gpm with your D5/mcp655 pump flow even on 5th/highest speed setting. Most probably flow will be 1.5-2gpm. More flow = more turbulence = a bit better cooling, but with more flow you start getting diminishing returns in temps/actual cooling. 1gpm is good, 1.5gpm a tiny bit better, 2gpm and even more - almost no gain at all (and in some cases with small/inadequate rad area even worse cooling because of increased pump heatload). For enhancing cooling/better temps, bigger rads always > then more flow. No need to concentrate on pump/coolant flow.

I was just concerned that even with my triple rad the water isn't staying in the rad long enough for the fans to cool it off as my temps could be a few degrees better. I wish I could monitor my water temps somehow...

Thanks.

[Martinm210]

I was just concerned that even with my triple rad the water isn't staying in the rad long enough for the fans to cool it off as my temps could be a few degrees better. I wish I could monitor my water temps somehow...

Thanks.

Don't worry about time in the rad, in a recirculating loop it doesn't work that way. Twice the flow rate may spend half the time per cycle, but twice the cycles so the time spent is the same. More flow is always better regarding radiator or block performance...the only negative is pump heat dump which usually doesn't matter much unless you have a double or smaller radiator.

For most typical triple radiator setups, optimal pumping powere will be one D5 or DDC at full power. You can even gain a hair more with double pumps, but its not huge gains. For noiseless extremes, you can even turn a single pump down to lower levels without too much loss.

[davidm71]

Was wondering what us the typical delta from idle to load with a triple rad and heatkiller over a 920 d0? I get like a 16 to 20 degree temp rise at stock 2.66 ghz from idle to load. I heard a ten degree delta is normal for water?


Thanks for your thoughts.

[Blueking]

Anyone knows what internal temp on pump electronics usually are at ?

[Gwiggles]

Does anyone know if the XSPC Dual BayRes Two would perform similarly to these dual DDC tops if I hooked them up in serial? It wouldn't have all the abrupt 90 degree turns, right?

[dr_dx]

Be nice to see one of these in the test - Iwaki WMD-20RZT - even at $150 it would be a cheaper/better solution than the doubles and triples.

[CedricFP]

Great work Skinnee. This sort of information is extremely useful.

[Russ750]

I had three pumps in my system for about a year now. No issues with leaks, But they are in various places thru out the system. I have done it with regular bitspower fittings. Most are compression though.

[Russ750]

Just though i would share these. This system was made possible because of Skinnee, And Martin. So special thanks to both of them for all there hard work. Anyway you can see, the three pumps. I average about 2.25 gpm with this system. (Two bucket and stopwatch method). I know it's overkill for what's in there. But i was making a future proof case.

[Russ750]

You can see i use the pumps as boosters in various parts of the system. I personally wouldn't recommend all three pumps in a row,and then running threw the loop. One could possibly have leaks that way. If they are spread threw the system, it allows the pressure to be spread out.

Pump > Radiator > Video Cards > Motherboard > Pump > Radiator > Cpu > Reservoir > Pump > Radiator. Then Back to first pump. Is how my loop is set up. I have run this setup with regular barbs with no clamps without leaks, (NOT RECOMMENDED) just for testing purposes. Hope this helps. This is just my opinion though. please refer to Skinnee or Martin as they are far more of an authority than I.

[Russ750]

In operation, double DDCs likely won't subject the system to more than about 11psi, triples maybe 17psi, RD-30 maybe 14psi.

In practice I think it's been common enough to see dual DDCs that they are relatively tested to a comfort level, but triples are pretty new and not tested a whole bunch.

We don't know about the pressure rating on most parts so it's anybody's guess on that? Perhaps the manufacturers can start posting pressure specs??

Regardless, I suspect the weak point is the fitting/barb pull out on the tubing, especially anything that's been removed/put back together, and particularly with some compression fittings that have small barbs. A 1/2" tube at 17psi is receiving about a 3.5 lb normal force trying to pull the tubing off the barb. I'd be a little uncomfortable with that sort of pressure unless I had a standard barb and worm drive clamp.

I think that with three pumps before loop your asking for leaks. But with them spread out in the loop. It significantly reduces the pressure. At least of the first few fittings/blocks. My analogy wood be put three pumps on a floor. And try to pump the water, 24' straight up a wall. The third pump in the series would have the most pressure on it..

Now if you spread the pumps out. About every 8' they would all have about equal pressure within the pump housing. At least that's how i understand it.
If i am wrong please someone correct me. I don't want to be spreading bad data.

[Russ750]

That brings me to another question. What happens to flow rate when the pumps are spread out.
Skinnee, some time while your testing.
If you could put three pumps in series, the thru the loop. Then do a test with them spread out within a loop. My initial feeling is that there would be very little effect on flow. Only the pressure would be affected. But once again i am not an authority on any of this

[Waterlogged]

That brings me to another question. What happens to flow rate when the pumps are spread out.
Skinnee, some time while your testing.
If you could put three pumps in series, the thru the loop. Then do a test with them spread out within a loop. My initial feeling is that there would be very little effect on flow. Only the pressure would be affected. But once again i am not an authority on any of this

Overall flow rate is unaffected but local pressure is.

This is a pic of a test I did a while ago for this very thing. I came to the conclusion that my gauge wasn't sensitive enough to get a solid reading and I actually needed more than 1 gauge. I have a new test planned with 2 new gauges (much more sensitive) but I've yet to get them (~$100 ea).

[skinnee]

Yup, regardless of where you put the pumps, flow rate will remain unchanged.

[Martinm210]

I think that with three pumps before loop your asking for leaks. But with them spread out in the loop. It significantly reduces the pressure. At least of the first few fittings/blocks. My analogy wood be put three pumps on a floor. And try to pump the water, 24' straight up a wall. The third pump in the series would have the most pressure on it..

Now if you spread the pumps out. About every 8' they would all have about equal pressure within the pump housing. At least that's how i understand it.
If i am wrong please someone correct me. I don't want to be spreading bad data.

Yes, that's a good point.

If you spread the pumps out, that will help keep the max pressure down. In a sense, it's getting used up by the blocks and added back in by the pump. You'll still have fairly high pressure because of all the pumping power, but the peak would be reduced because you're distributing it over the whole loop.

Nice setup!

[MagisD]

OK so I'm horrible a reading those charts , Can anybody sum up the average increase from single pump to 2x of the same ?

[Russ750]

OK so I'm horrible a reading those charts , Can anybody sum up the average increase from single pump to 2x of the same ?

Those charts list a fairly open system. Basicaly you get about 25 to 30 percent better flow. With two pumps and about 30 to 35 percent better flow with three. In a simple system. Much less increase with setups that have more restriction

Kinda the same as crossfire or sli. The first video card gives you, lets say about 60 fps. A second gives you 90, third gives you 105, forth give you 110.

If you look at my box above, you will see that i could gain about .2-.3 gpm, if i would loose all the 90's. But then it wouldn't fit. So thus three pumps to keep me above 2.0 gpm

With your overkill box, you biggest fight is going to be keeping your flow rates up. With six radiators your water temps will be maybe 1/2 a degree less, than the same system with only three quad radiators. Thats with anoverclocked cpu and quad sli/crossfire and motherboard cooler. And about 5 hours of Vantage to heat the system.

If your water speed drops from 2.0 gpm to 1.5 gpm because of so much restriction (90's, Six radiators). Then the actual cpu and gpu temperatures will raise about 1 degree compared with the same setup with less radiators and 90's. You very easily meet the point of diminishing returns and start to go backwards, with restriction. This is do to slowing down the water over the water blocks. At least that's my experience. Skinnee did a great job, with showing what a normal system would flow like. The only way you will know exactly where the balance is to test your system starting with one rad and working your way up. Same theory goes with the pumps. For me it's all about the balance of proper flow rate, cooling and aesthetics, Cost be damned.

[MagisD]

I know i went way beyond the point of diminishing return. But i also know i never want my fans or pumps above silent on 24/7 oc while gaming :-) and I planed that for max load of my next upgrade or 2.

I was just sorting out if i should go serial, Dual loops, triple and if I should throw in a couple more pumps.... rads are barely any restriction. All those 90' and the rads equal about 1 med restriction block. 2 in Series mcp-35x should do it if not i'll throw 655 on rad only loop.

Sent from my X10a using Tapatalk

[MagisD]

Btw anybody have a method for approximating gpm without a flow meter i never picked on of those up :-)

Sent from my X10a using Tapatalk

[Russ750]

Just by chance skinnee did you keep track of cpu temps while testing. That's if you even did these test with a waterblock on a motherboard. If you did, could you toss up another graph just to show heat temps relevant to flow. I know there on Skinnee Labs, I just though it might help on this thread.

[Russ750]

I know i went way beyond the point of diminishing return. But i also know i never want my fans or pumps above silent on 24/7 oc while gaming :-) and I planed that for max load of my next upgrade or 2.

I was just sorting out if i should go serial, Dual loops, triple and if I should throw in a couple more pumps.... rads are barely any restriction. All those 90' and the rads equal about 1 med restriction block. 2 in Series mcp-35x should do it if not i'll throw 655 on rad only loop.

Sent from my X10a using Tapatalk

In your case i would personally do a dedicated loop for the rads, and motherboard block. Since flow speed thru radiators has almost zero effect on water temps. Run that into your large reservoir. Then have a second loop, for just your cpu, videoblocks that runs into the same large reservoir. As far as pumps. The ddc's have a little better head pressure, thus a little better flow, but less reliability. The D5 flows a little less in a normal loop but has better reliability.

That being said, i have been running ddc's for almost three years 24/7 with no failures. They are always attached with thermal tape, either to a copper plate or to the aluminum case.

My biggest problem with the d5 was mounting it cleanly in a system, back when i was choosing pumps. Now there is a ton of choices. Here is my favorite new example.


Which reminds me Skinnee if you get a moment, and a chance to toss one of these in the flow rate graph. I will buy it and have it sent to you. I need to call Koolance and see if these can do series. From The design it looks like only parallel, Major bummer

[Russ750]

Btw anybody have a method for approximating gpm without a flow meter i never picked on of those up :-)

Sent from my X10a using Tapatalk

Two five gallon water jugs. Take a one gallon water jug pour it to an empty 5 gallon water jug. Then mark the side, where it fills to. Repeat four more times. Pour the water into the other 5 gallon jug. Now pump from the one jug to the one with lines. Start your watch at the same time as you start your pumps and wallah.

Koolance quick connect make this really easy to do. Because you can have your loop set up for use. Then just have an extra set of fittings. for when you want to test. I never though i would ever say anything about koolance but the have some cool stuff now.

[skinnee]

OP Updated: CPU and SLI loop charts updated, EK Dual D5 and Jingway PQ linked.

Working on a few more D5's today...

[larrys2006]

I plan on using a EK Supreme HF Full Nickel or copper unit on a i7 920 cpu with 3/8 id tube and 1 pump and 1 rad, my problem is understanding this chart. It looks like a Watercool Single would be a good pump for cpu and later gpu cooling but I am not shure, what do you think? Also how about the tank do you think they are mostly all the same?

[Church]

How about adding this separate MCP35X-H top to test? Interesting, how it might do on 3.25.