I don't need a pump. I need a PUMP. Looking for outside the box suggestions.

[robble]

that pump is showing 22gpm / 8 = 2.75gpm. That could supply all your servers with enough water. I'm not sure you can get away with not having booster pumps if they are stacked vertically though it may work if they were 2 high and 4 wide. You are going to be pushing water through not just blocks but a large heat exchanger and a pretty long loop.

I don't want to derail the thread either but "gamers sell your cards", "has been profitable but competition rising", and "water miner" sounds a lot like you got a mega BC farm Have fun. Electricity costs in Hawaii are much too high for me to play that game.

 

[Diggrr]

I personally would go with the 3/8" tubing, QDC or not. When you figure the cross-sectional area of a 3/8" tube .vs the 1" feed, I do believe that the 1/2" would drop your flow too much...you are dividing a 1" feed by 4 computers, and 3/8" would keep closer to an even division of what the feed size delivers.
There's not much of an appreciable temp difference between 3/8" and 1/2" tubing, in practice >1°C at most (single block/pump/rad testing).
You can keep it at 1/2" between the cards to salvage some pressure drop.

**BTW, a motor controlling fan switch powering your pump outlet will let you throttle it. NOT a normal (cheap) fan dial, but one that says motor controller on it.

 

[robble]

note in my diagram when i mentioned the booster pumps providing enough flow even in parallel - i meant the servers in parallel - not the video cards themselves. I'd put those in series.

 

[WaterMiner]

note in my diagram when i mentioned the booster pumps providing enough flow even in parallel - i meant the servers in parallel - not the video cards themselves. I'd put those in series.

I figured as much but thanks for the clarification. I really like the parallel GPU setup but given flow rate is going to be limited no matter what it really isn't an option.

So I think that aspect is settle. Cards is series, servers in parallel.

 

[WaterMiner]

that pump is showing 22gpm / 8 = 2.75gpm. That could supply all your servers with enough water.

Yeah but 22gpm is too much for the mains. For pipe longevity water velocity should be <7 feet per second. That's 12 gpm on 1" PEX. Now that is to ensure pipe lasts 20-30 years. So I likely can go a little over 12, maybe 14-16 gpm but 22 would be pushing it. Also remember as water velocity rises so does resistance. At 15gpm in 100ft of 1" pipe we are looking at 12psi (~5ft head) pressure loss. My loop is going to be smaller but 100ft is a good upper limit. So 22 gpm needs either a bigger pipe or even more lift.

I'm not sure you can get away with not having booster pumps if they are stacked vertically though it may work if they were 2 high and 4 wide. You are going to be pushing water through not just blocks but a large heat exchanger and a pretty long loop.

I think you might be right. They are going to be mounted in a server rack. I think I already know the answer but putting the manifold horizontal and at the top of the rack isn't going to help is it? Server #1 is 6" below manifold, server #2 is 12" below manifold ... server 6 is 36" below manifold?

 

[WaterMiner]

I personally would go with the 3/8" tubing, QDC or not. When you figure the cross-sectional area of a 3/8" tube .vs the 1" feed, I do believe that the 1/2" would drop your flow too much...you are dividing a 1" feed by 4 computers, and 3/8" would keep closer to an even division of what the feed size delivers.
There's not much of an appreciable temp difference between 3/8" and 1/2" tubing, in practice >1°C at most (single block/pump/rad testing).
You can keep it at 1/2" between the cards to salvage some pressure drop.

Sounds good. One less thing to worry about.

On edit: I misread I thought you were saying using 1/2" even if QDC was only 3/8". I am not seeing why 3/8" would be better. Maybe the same but certainly not better right?

**BTW, a motor controlling fan switch powering your pump outlet will let you throttle it. NOT a normal (cheap) fan dial, but one that says motor controller on it.

I didn't know that. That is good news. Makes me feel better about not having that monster WMD-R55 blowing the loop apart. I really need a WMD-R48.5

 

[Diggrr]

Something like this should do just fine. $50 is a bargain too.
Rated for 4 amps or less, but there's another model at this site for more money. I've bought from them before.

 

[robble]

Now that is to ensure pipe lasts 20-30 years.

do you *really* think you'll be running this thing in 20 years? or even 10?

Flow may not even be that fast - it's just capable of it. You'll still be limited by resistance in the loop. You wouldn't want to booster pumps to starve.

 

[johan851]

On serial vs parallel. johan851 based on what you are saying (>2 gpm per rig is unlikely) I think I have no choice but to put the 4 cards in series right? Otherwise if rig is closer to 1 gpm then each branch of the SLI will be 1/4 gpm or less. Way too low. Right?

The parallel manifold and limits on flow rate make putting the cards is series a requirement wouldn't you say? Temps aren't really an issue. If I can keep all the cards <60C year round it is fine for what I am doing. I already have software which dynamically adjusts the clock based on core temp for each core.

Yup, that sounds right to me. Series is the most reasonable choice.

I personally would go with the 3/8" tubing, QDC or not. When you figure the cross-sectional area of a 3/8" tube .vs the 1" feed, I do believe that the 1/2" would drop your flow too much...you are dividing a 1" feed by 4 computers, and 3/8" would keep closer to an even division of what the feed size delivers.

Not sure I follow. If you have four paths with equal resistance, why should 3/8" vs 1/2" matter coming off the main? They'll all have the same resistance either way and flow will be divided evenly in both cases. Seems like overall resistance would be a bit less with 1/2".

 

[WaterMiner]

Not sure I follow. If you have four paths with equal resistance, why should 3/8" vs 1/2" matter coming off the main? They'll all have the same resistance either way and flow will be divided evenly in both cases. Seems like overall resistance would be a bit less with 1/2".

I take it you think 1/2" would be better. It sucks Koolance doesn't make a 1/2" QDC as that is likely the least restrictive point. Still I think 1/2" would still be a good idea as you are dealing with a point restriction rather than resistance along the entire tube.

Would you use 1/2" barbs and tubing even if largest QDC is only 3/8"?
Alternatively know any QDC in 1/2"?

I thought about just using a non QDC but I need to be able to cut off coolant for a single server. Using a pair of ball vales with a union would work:

[ mainifold ][ ball valve ][ union ][ ball valve ]----- line to server

they parts aren't cheap so there is no cost saving over using a QDC and it isn't exactly "quick" disconnect.

 

[Diggrr]

It's more of a preference than a fact based issue.
When you divide a given flow by a group of tubes with a greater combined cross-sectional area than the feed line, water speed is slowed (actual GPM remains the same).
I prefer to keep the water speed consistently high throughout the loop instead of having places where it speeds then slows, especially just before a resistance is met (when it hits the water blocks).

It may not produce a noticeable pressure feedback effect in our water loops, but I prefer not to leave it the chance that it may.

 

[Diggrr]

So much for memory.
4 x 1/2" tubes are a good match for a 1" line (.7788 .vs .7853), though if you add a 5th, the water would slow.

If you have no plans to expand, then 1/2" would be fine. If you do wish to expand and add another server, then 3/8" should be used.

There is still no realistic penalty for using 3/8" over the 1/2" lines. QDC's do have a flow penalty, but maybe not enough to matter.

 

[robble]

Why are people saying koolance doesn't make 1/2" QDCs?

I recently bought a pair.

http://www.frozencpu.com/products/1..._13mm_12_VL3N-F13B.html?id=nA7Sx2rY&mv_pc=269
http://www.frozencpu.com/products/1..._13mm_12_VL3N-M13B.html?id=nA7Sx2rY&mv_pc=268

 

[LoneWolf121188]

Have you considered something more...industrial? For example, a 10 second search on McMaster found this...24gpm at 20 ft for $600. Search for "water pump" on there and see what comes up.

 

[WaterMiner]

Why are people saying koolance doesn't make 1/2" QDCs?

I recently bought a pair.

http://www.frozencpu.com/products/1..._13mm_12_VL3N-F13B.html?id=nA7Sx2rY&mv_pc=269
http://www.frozencpu.com/products/1..._13mm_12_VL3N-M13B.html?id=nA7Sx2rY&mv_pc=268

Sorry. 1/2" threaded fitting.
They have 1/2" and even 5/8" ID but only 1/4" and 3/8" threaded. Sorry for the confusion.

Looking at Koolances website I realized I don't need threaded connectors. I was going to use threaded connector to attached the QDC either to manifold or to panel adapter.

Koolance has a barbed QDC w/ panel adapter built in.
http://www.koolance.com/water-cooling/product_info.php?product_id=886

SCORE