WTF? Did you expect to see 2 1250's in series on that graph for some reason?

I have no idea what, certain loss, in the "serial" application you think I accounted for. The cooling loop curves are the same I graphed earlier, and are based on the same 'idealized' setup.

I suppose that two pumps aren't relevant at this point... I must be loosing track of this thread! The first graph you posted shows Sim20R and Sim20RZ, I thought they were about the same pump, for some reason. I understood that putting pumps in series does not result in a full doubling of the pressure: am I off my rocker?:p Nah, just OT, sorry.

From both graphs you've posted, I'm inclined to believe that blocks in series work best!:shrug:

Failure recovery.
If your flow slows to lets say half more than likely your temps would rise X amount and level off at some point. If your pump were to fail then the time would be dependent on the thermal mass of the WB and the water inside it. So a small cooler would die faster a large one would give you more time. Without doing any math I'd guess you'd have about 10-20 seconds. I read a article about a guy who forgot to turn on his pump(same as a failure basically) the water boiled and the steam blew out the block and the rest is obvious.

Parallel vs Series
I'd go with series 90% of the time but if your nearing your pumps max head then going parallel will get you more flow through each block.

Couldn't you just make a bypass valve around the rad to keep flow going. Kinda what aspirit did with his rack mount deal. Kill two birds with one stone. More flow rate through the blocks without giving up serial routing. Wonder if the loss of cooling would defeat the purpose. Of course the pump may not be working as hard either and add less heat to the water......

Bah! Probably a higly ignorant post.... :D

Well, I'm designing my CPU blocks to be non-restrictive actually. While I'm not specifically trying to make the NB and hard drive blocks restrictive, I don't see how they couldn't be to a fairly large degree - I'm going to be making them with 1/4" ID tube, (CPU's get 1/2") with fairly long passages that will have as many bends as I can figure out how to fit on the drive plate (I'm going for simplicity - bend tubing into a maze, solder the maze onto a 1/8" x hard drive sized plate. No milling, no seams, should be no leaks, efficient enough to cool the drive.) I am guessing each plate will have between 1 and 2 feet of tubing, and there will be at least 3 plates, and possibly an NB block (which would be relatively restrictive internally, but still use 1/4" plumbing)

Thanks for the input on failure recovery, it is somewhat reassuring. Big is relative, but I think I'll be on the heavy side of that set of numbers. I might do a limited simulation test once I get things built to get more definite information.

A minor planning concern I just had - I haven't dealt with one of the modern ATX type power supplies from user space yet. Do you know how it behaves when the AC input line is interrupted and then restored? Obviously it goes off when the power does, but when the power comes back does it turn back on and try to bring the PC back up?

I'm planning on implementing a 'last chance kill' relay that would interrupt power to the PC and pump if it triggered. If the supply stays off, then I only need a momentary relay. IF it trys to come back up, I would need a latching relay that would keep the power disconnected until I reset it.

Depends on what your purpose is :rolleyes: If it's to get lots of flow, it would probably work. If it's to cool the system, it seems rather pointless to bypass the cooling loop. :( I haven't seen Aspirit's setup, but if its a rackmount (presumably w/ multiple processing boards) I'd suspect the function of his bypass valve is to isolate a part so that it can be serviced without having to shut down the entire rack. At least if *I* were doing something like that (and I used to work on high reliability rackmount setups for telcos) it is the kind of thing I'd do. :cool:

Gooserider

I'm resurrecting this old thread because I figured out today that the two graphs and accompanying information above is wrong.

My calculations for systems with paralleled blocks were wrong. The following are corrected graphs.

http://uffish-thought.net/wc-gifs/p-vs-s-rev.gif

http://uffish-thought.net/wc-gifs/p-vs-s-eheim-rev.gif

For the Sim20-R, per block flowrate is:

WWS 7.5
WWP 7.8

MCWS 12.3
MCWP 9.6

For the Sim20-RZ, per block flowrate is:

WWS 8.5
WWP 5.8

MCWS 10.8
MCWP 6.2

So, for restrictive blocks combined with a pump with a 'general purpose' PQ curve, per block flowrate may be better with a parallel setup. (The resistance of the rad and tubing have an impact on whether parallel or series is best.) With a pump designed for 'high' pressure, series is still best though.

Of course, all this assumes I've got my maths right this time.

Oh dear.
My old graphs do not agrree with either versions.
http://forum.oc-forums.com/vb/showth...hreadid=183841

http://www.jr001b4751.pwp.blueyonder.co.uk/Pump1.jpg
http://www.jr001b4751.pwp.blueyonder.co.uk/Pump2.jpg
Will check*

* Edit:Will checklater.

Well Les, I can say from practical experience that an MD-30RZ (50Hz) model does indeed push about 12LPM through a single WW, and just over 7LPM with an Eheim 1250, and over 5.5LPM with an Eheim 1048. Will lose a little bit with a radiator attached, but the PD of a radiator is fairly low in comparison so long as the fittings are largish.

LOL, had me worried there for a moment.

My graphs have a radiator and tubing figured in as well. I believe that is the difference. My graphs would be consistant with yours if those terms were removed.

Also had me worried.
Like yourself I do not like posting misinformation or wrong calculations.
However all looks sweet.

I'm busy posting corrections where I can think of. If I'd paid closer attention to your graphs in that O/C thread, (Or Graystar's last post in that thread, which I seem to have missed completely.) I might have realized my mistake much sooner.

My radiator spreadsheet has the same problem, and is utter crap.

It'll take about a minute to correct the equations, and then an hour of 'trial and erroring' the fudge factors to get it 'right'. (I hope.) It's not going to happen tonight though.

Nope, you were right Les. There is another problems with my graph.

Apparently when I'd originally done it, I pasted in the 'flow resistance' for a Lytron 6320 radiator instead of White Water.

It's been a long day.

Will fix later.

Edit: My correction post has been corrected.

I had a pump fail (fail to be plugged in ;) ) and everything was fine because my rad was higher than my block and 'natural' convection took over. It leveled out at 45degC, only about 8degC higher than with the pump!...

What I'm saying is there's more to be factored in than just the mass of the block. Pump redundancy isn't needed if a little forward thinking is exercised...

Les,

Love your graph work man, very helpful to me in grasping some of the posted info.

Could I ask you one favor? Could you include the curve for the smaller Iwaki MD15 & MD 20z pumps?

All posted graphs are ,unless labelled otherwise,
for 50MHz flavours.
Although would expect different PQ curves for all pumps at 60MHz only Iwaki show two sets of curves.
Eheim being a European manufacturer I used 50Mhz data as possibly the more representative.

If can find original Excels will eyeball some 60MHz pump curves.and post in this thread.

Thanks Les,

For comparison of one pump vs others 50hz would be fine.

60hz is however a better test of what I can expect when I get a system running.

Thanks again!:)

Added Iwaki MD-15R(M)-N 60MHz and Iwaki MD-20RZ(M)-N 60MHz to second graph which now includes a Swiftech wb.
http://www.jr001b4751.pwp.blueyonder.co.uk/Pump3.jpg

I read through the rest of this thread and saw no response to this question. I am curious, too. I am new here and have been lurking, absorbing and filtering knowledge and twiddling with some design ideas for my first WC rig. I ended up here because there are fewer obviously wrong thinkers here. Y'all at least have the grace to keep your wrong thoughts more subtle than some of the other sites and forums ;) I am not the sort to jump into the pool halfway, so I plan on lots of flow restricting components as I go overboard with a fully WC setup, and over-engineered at that :D ! I like to get my ducks in a row before lots of time, money, and materials are expended, and do my best to ensure a satisfactory result. I've learned a lot lurking here, but I've got more to learn before work begins in earnest. Pardon my n00bishness :shrug: .

I've seen MMZ_Timelord in another thread talking about pushing pumps into a resevoir with a common outlet. I don't think pushing disimilar head pumps into a tank will work, there's the matter of backpressure to consider. If the pumps weren't exactly matched, wouldn't the smaller head one be fighting against the load of the difference and actually backflow if the rest of the circuit past the push tank was more restrictive than what its remaining capacity could handle? In other words, the system could not be more head restrictive than what the lower rated pump could push, and the lower head pump would be under greater load. At least that's my n00b guess.

I've had a similar thought, though, except rather than pushing multiple pumps into a tank, I envision feeding multiple pumps from a common resevoir. The resevoir will be tall, and the feeds staggered low so that gravity and some baffling will minimize pump competition on the suck end as well as promote turbulance for the inputs to the resevoir (both for air and heat exchange, see below). I was figuring on two pumps, but now I am wondering about three...

What would the effect be of having a separate radiator circuit fed from a common resevoir as other circuits? A pump running just the radiators should get a better flow rate through them than in series with other head-robbing components, and the other circuits a better flow rate without radiators (or with less restrictive radiators, did I mention I like to over-engineer?). Would this improvement be enough to cancel out the loss from the coolant-to-coolant free exchange in the resevoir?

Will the fact that we are in a closed system negate my attempt to limit pump competition on the suck end of the feed? Or would this actually be a good thing as the combined draw of the pumps would be helping the more flow restricted inputs? Or would it make them worse by helping the less flow restricted inputs more and cause me to have to tune the inputs to satisfy my anal nature? Or am I in the same problem area that I perceive the push tank to be in because all I am doing is moving the point of push to the other side of the circuit (although I do have 3 ins and 3 outs)?

Well, I think that's enough n00b musings and questions for now. Am I thinking clearly, or am I just another clueless n00b :confused: ? Give it to me straight.

LPorc,

I was indeed trying to say that two pumps would utilize the same reservior to DRAW water from and pump it throughout the system and then return to the reservior.

I will not be doing this on my current system as space limitations would not allow it. :(

The idea of a separate pump for just the radiators and one for the CPU/GPU/HDD/NB, etc. is interesting. I would like to know if that would allow better removal of the heat since the radiator would have MUCH higher flow with that setup. :shrug:

Ah, I must have misunderstood, or perhaps confused someone's reply.

I have the advantage in that I'm starting from scratch with my over-engineering, so once I figure out what all I want in the box, I'll build a box to hold it :cool:

The idea of using different flow rates for radiator and cooling loop was discussed on OCAU. You might need some aspirine though :D

Thanks! I activated over there and read the thread. The numbers were pretty straight-forward, although for such a short thread it did meander a bit.

The bottom line I got out of it is that in theory we could get a small gain, but the gain is small enough we might not get it in practice. From my point of view, more importantly, is in theory we don't get a loss. The plumbing might be a little more complicated, but it does give us some options, and can let us use disimilar pumps effectively.

Speaking of pumps, anyone here build their own or investigated other types of continuous flow pumps?

Yes, I think the main advantage to your idea is flexibility rather than extra performance, as you gain more from increasing flow through the blocks than through the rads.

Re pumps, I have successfully "converted" a 180w draining pump (83lpm/5.5m) into a 3-stage unit capable of approx 45lpm/11m. For watercooling, but not PC:)
Its dead now, but I still have it lying around somewhere...

I just read this whole thread and I learned a hell of a lot, now I need to go eat some tylenols and take a nap. You guys are crazy.

Welcome to the club ;)

Going back to the bits about pumps failing, back in the good old days when I ran an Original (Slot-A!) 750Mhz Athlon (running at a whopping 963Mhz :drool: ;) ), I had a too had a pump failure (also unplugged lol). This is what I found when I plugged in my temp-monitoring software :

http://www.geekops.co.uk/photos/0000.../Too_Close.gif

Bear in mind that I'd left the machine on for about an hour or so, so I'd guess the temps were about stable. It was amazing how soft the tubing went when I turned the pump on, but fortunately, no leaks / breakages!!

P.s - Amazing thread btw - I've not being paying much attention to w/c'ing recently as my degree (read resits :rolleyes: ) took priority. I'm starting to get back into it now and this thread was an amazing recap / learning experience. Well done to all - now if we could just condense it into a singl page, I'd say you have the beginnings of a WC'ing bible! ;)

generally when were talking pumps max head is more important i watercooling loops and the reason is because of the restrictive design of cpu blocks, Nb blocks gpu blocks etc.

an example of this is comparing the swiftech mcp600 pump (690lph, max head 3.2m) to the eheim 1250 (1200lph, 2.0m max head) the eheim is designed for aquariums and ponds hence its more of a free flowing pump and has an open impeller, the swiftech has a closed impeller and produces higher head, this is beneficial because when we get into restrictive setups especially with blocks like the cascade and white water. the pumps with higher head will have a higher flow rate.

well thats my 2c, back to biology practical :s

Hi all

What a great thread! I live in Denmark, and the ppl. over here tend to go with the theory that flow doesn't matter so this was indeed interesting to read!

Does anyone know (or care to do) a real life test using the same system and different pumps to show what the "real world" gain of better head/flow is? since it can be a bit difficult to piece it together from the different "lab" results..

Are you thinking *cough*-deleted-*cough* ? ;) yeah im from DK too and ppl really insist that flow is secondary - and great thread - i read it awhile ago and theres alot of good info here

Wow....I suddenly got the urge to go back to school and register for thermodynamics and fluids class...:)