Alex is going watercooled

[AvalancheDiode]

Re: Alex is going watercooled

From what I've read about water cooling all of the parts you have chosen look good.

Wouldn't running it in parallel cause there to be a massive loss of flow through each component, thus making the system less effective at removing the heat, also as the GPU produces more heat than the cpu the water would be at different temperatures which would cause expansion, different speeds of flow and then little eddies* would be formed at the Y connectors.

Actually wouldn't a Y connector produce small eddies anyway?


*An eddy is an area of liquid that is either not flowing or flowing in the opposite direction to the main flow.

Eddies can form under certain pressure/flow speed condition. Given the flow and pressure in the cooling system, eddies can not form unless you have trapped bubbles.

The reduced amount of flow won't affect cooling capability at all as long as you can pump few a liters per second into the system. Overall flow amount is decided by pump capacity and resistance through the water loop (or, in more academic term: viscosity). Serial connection is not necessarily more efficient in terms of flow amount.

Water don't expand much during heat. Your system can hardly increase temperature of coolant reservoir by 10C if it is properly dissipated by heat exchanger. There is a reason that those systems use Tygon tubes. These are soft tubes that can hold up to 40-50PSI. Given the fact that typical coolant pumps operate at 10-15PSI, the safety margin is quite large.

The advantage of parallel system is that the input to each device is independent of each other. The overall system design is a bit complex, but it allows you to add/remove individual items without touching others. It also allows you to debug water loop more easily. If one water loop fails due to obstruction/leakage, you just take that part out, fix it, and put it back from Y-connector. With a serial system, good luck with that. Y-connector (or Cross connector) will be at easy reach, while all other nozzles of water blocks will be deeper in the jungle. If I want my computer to be liquid cooled without any upgrading provision, then I'd go with submerged design.

The other advantage is that all your components under water blocks are cooled by steady flow of constant temperature. It means when CPU is on high load, GPUs will still be cooled under the same condition despite how much heat is generated by CPU.

[Warhawk]

Re: Alex is going watercooled

This is why I love watercooling, so much to think about!! So custom!!!

[AvalancheDiode]

Re: Alex is going watercooled

Avalanche, the drop in pressure from using a Y-Splitter would ruin the performance. Waterblocks need high flow to exchange maximum heat into the water, by splitting your flow you would effectively half the pressure going into each block.
Also, with radiator positioning, its commonly accepted that you put your radiator after the pump and before the CPU block because you want that water as cool as possible before it hits the CPU, pumps also create a little heat and will impart some of this into the liquid. Although, in practice the order in which you put your components doesn't affect temperatures very much, that's the widely accepted view.

Apparently people who design all-in-one computer cooling kits disagree with you about where to put radiators:

http://www.thermaltakeusa.com/Product.aspx?C=1160&ID=1776

Reservoir temperature is what you should be concerned. All your cooling power derived from how cool your reservoir is. If you put radiator in your way, then it'll run as a positive feedback on temperature of the reservoir.

All expert-designed water cooling system that I've seen are in parallel.

Like I said earlier. Resistance through serially connected water loop also kills water flow. Use bigger pump, or even more pumps, if you need that amount of flow. There no reason you shouldn't get that for a 750W system.

And, by the way, Y-connector drops flow, not pressure. According to Bernoulli principle, higher flow velocity will produce lower static pressure. Kinetic pressure comes from total flow velocity and density. Since we're talk about incompressible water, flow speed determines kinetic pressure.

[Bisley_Bob]

Re: Alex is going watercooled

Avalanche, i'm sorry, but a thermaltake all in one cooling kit is no proof of where to put your components in a loop, they are absolutely rubbish. You seem to think that water coming out of a waterblock is a lot hotter than the water entering it, in a low flow system perhaps, but if you have a high flow then the water only has to change in temperature by a tiny amount in order to keep the cpu cool. Look at it this way going rad > res > pump > cpu means you cool the water in your rad, then it gets to sit around in your warm case in the reservoir before it goes to your cpu.
Please link to these expert designed loops that are in parallel, or alternatively go to some dedicated overclocking/watercooling forums and see what they do there. They will all reccomend serial loops.
Found an expert designed loop for myself here:
http://www.vadim.co.uk/product5359/C...QX+Intel-775G2 Look at the 3rd image down the page and trace the water loop, start at the pump. Pump > Rad > Rad > CPU > NB > Rad > Video cards > Res > back to pump. And vadim are good at what they do.

[=98=]

Re: Alex is going watercooled

haha, seriously, if you're going to quote an expert watercooling kit builder, don't quote Thermaltake, they are awful. Go onto any decent overclocking forum and say Thermaltake kits are good, and you'll be ripped apart.

If you want a good kit, quote Swiftech. I have talked to some experts who build custom kits and systems on overclock.net and they said Rob has the correct idea as to the layout. They said do not go parallel. Thanks for the advice anyway.

[AvalancheDiode]

Re: Alex is going watercooled

Bob:

The cooling in these rigs work because of the amount of water they use, not the cooling capacity they've integrated, which probably overkills anyway. This is not the best way, but they have to compromise with limited space within a case. There is a Koolance outboard cooler laying at my work we used to cool some electronics, and they use approach I described.

Alex:

You can ask those people on overclock.net how many of them have ACTUAL experience in designing a cooling system. Most of these people are buying aftermarket kits (which IMHO are rip off). A few month ago when I was thinking about how to build (DIY, i.e.) a water cooling system, the conclusion is to use large water body, immersed heat exchanger and submerged fish tank pump. In this case, you get best of all world: cheap, quiet, power efficient and low maintenance. There's still a catch: PSU. Aftermarket water-cooled PSUs are one of the biggest rip-off in this business. If you have access to junk yards and some machining and welding tools, it will cost you next to nothing. Now you know why I don't take advice from that forum. They use serial because they can't use parallel. Aftermarket kits usually comes with 3/4" or 1/2" ID Tygon tubes which are ridiculously large. If you use hardware supply, then you can avoid using these limitation and go with 1/4" or 3/8" ID tubes. Parallel is not only possible, but easy with small diameter tubes.

As for my experience in designing a working cooling system. Well, that's unquestionable. I did and I maintain even larger system.

[Bisley_Bob]

Re: Alex is going watercooled

I am intrigued as to what large system you maintain. I'm sorry, i may well be wrong about this, but i just cant see how someone who quotes thermaltake kits as being the way to do things could be someone who knows their stuff.

Bob:

The cooling in these rigs work because of the amount of water they use, not the cooling capacity they've integrated

Are you saying the water temperature will continue to rise? But it only stays effective because there is a large enough body of water to dissipate that heat? In that case what about the people who use a t-line instead of a resevoir?

[=98=]

Re: Alex is going watercooled

You can ask those people on overclock.net how many of them have ACTUAL experience in designing a cooling system. Most of these people are buying aftermarket kits (which IMHO are rip off).

Nope, 90% build their own custom kits. They don't like to help people who buy aftermarket watercoolers.

I actually have a separate thread on there.

[AvalancheDiode]

Re: Alex is going watercooled

I am intrigued as to what large system you maintain. I'm sorry, i may well be wrong about this, but i just cant see how someone who quotes thermaltake kits as being the way to do things could be someone who knows their stuff.

Are you saying the water temperature will continue to rise? But it only stays effective because there is a large enough body of water to dissipate that heat? In that case what about the people who use a t-line instead of a resevoir?

A laser system that sucks 60Amp at 600V that will blind or electrocute you depending on how wrong you try to use it. The other systems include a smaller laser system that we try to sell with an internal cooling loop, and detector arrays that requires chilled water to cool down the TE side.

Alex:

I think they still buy radiators, fans, water pump and some other accessories from the same companies, only bought them separately due to complexity of their systems. If they do it with a motorcycle radiator picked up from the junk yard nearby, they should be praised for the efforts.

Since I happen to have a Pentium D and an HP motherboard laying around, I'll get it on water to prove my point, on the cheap, of course.

[Bisley_Bob]

Re: Alex is going watercooled

I'd imagine cooling lasers is probably different to cooling a CPU. I still cant quite believe that you said thinner tubing is better, way to restrict the flow as well as split it up.
Please do proove me wrong with evidence that a y-split loop is better. You have yet to link to any "expert-designed water cooling systems".

[=98=]

Re: Alex is going watercooled

I have to go with Bob here. I'm planning on 1/2" piping to increase flow. Thinner piping is bad

[AvalancheDiode]

Re: Alex is going watercooled

I'd imagine cooling lasers is probably different to cooling a CPU. I still cant quite believe that you said thinner tubing is better, way to restrict the flow as well as split it up.
Please do proove me wrong with evidence that a y-split loop is better. You have yet to link to any "expert-designed water cooling systems".

Yes, there is some difference. More water, more heat and more dangerous.



This is a picture of a linear accelerator. The water cooling is for very strong electric magnets. The tubing seems like 1/4"ID polyethylene, and each magnet draws more electricity than your gaming rig. The trick here is to use narrow piping to enhance flow speed.

If you insist there's a difference in computers:



SLI cards are connected in parallel to ensure both are cooled at the same level. This is essentially the same to the picture above. They need to control the performance of each magnets so that each behaves the same to all others. Same in this case. SLI cards working at different temperatures is bad. If you want to make best of it, you must have both card operating at same condition.

Alex:

What comes in doesn't disappear. It's called mass conservation. Unless you are doing things related to first picture, you're bound by this law. Actually wide tube slows flow speed. By shrink it, the water jets out faster, even though the liter/min remains the same. It doesn't really matter if you use an outboard cooling system. Some people use large piping as internal reservoir. Zalman's nozzle accept 10-8mm tubing, so eventually you are limited by nozzle size. Large != better when you want to move liquid faster. It's the gph (gallons per hour) that matters. If you have a pump with low gph, then it will not work no matter what tubing size you choose.

Of course there's a limit about how narrow you can go, and it usually depends on how good your pump is.

[Bisley_Bob]

Re: Alex is going watercooled

That electromagnet's cooling system looks very different to what you would have in a PC, it's got way more points of cooling than the two that alex wants to cool. Obviously with that many "blocks" (or whatever they are) in series by the time the water gets to the last one it'll be really hot, not to mention that if each magnet draws more than my rig they will be dumping a hell of a lot more heat into the water than a CPU. And yes, with that many parallel circuits i perfectly see the sense in necking down the hose to keep the coolant flowing fast. None of this apples to a PC in which one CPU and one GPU need to be cooled.

OK, you posted a picture of parallel SLi cards, if you go back to the vadim link i posted earlier they've put their cards in series, tit for tat really. By placing the cards in series there may be a 1 or 2C difference in temperature, big deal, there will be other factors at work too which will make the cards differ slightly in temperature. But overall they will be running cooler as they will get a greater coolant flow per unit time. I'll say again, with high coolant flow each block will not change the temperature of the water significantly, certainly nothing like the heat that magnet loop is having to deal with.

With regards to tubing size, a lot of the enthusiast blocks, like the D_Tek which alex intends to buy, will have been made with 1/2"ID tubing in mind, so won't be necking down the flow like a zalman block would (their resterator kits use much thinner tubing). It's more efficient to put liquids through larger tubes so bearing in mind the pumps used in PCs aren't huge beasts (like i'd imagine the ones you're farmiliar with are) everything should be done to facilitate flow, that means nice wide tubing and a nice simple loop without unnecessary things like y junctions and the extra tubing that goes with implementing them.

Put it this way, lets imagine your pump pumps out 1000 litres per hour. You put your CPU and GPU in series, that means both blocks get 1000L every hour. Then you decide to do the y-splitter thing, now both blocks get 500L each per hour, sure the water reaching the GPU may be 1 or 2C cooler, but it gets half the coolant flow, not to mention the CPU which has massively lost out.

[AvalancheDiode]

Re: Alex is going watercooled

Put it this way, lets imagine your pump pumps out 1000 litres per hour. You put your CPU and GPU in series, that means both blocks get 1000L every hour. Then you decide to do the y-splitter thing, now both blocks get 500L each per hour, sure the water reaching the GPU may be 1 or 2C cooler, but it gets half the coolant flow, not to mention the CPU which has massively lost out.

500L/hr pump is good enough for average system with one CPU and one GPU. Actually, more than enough. 1000L/hr system is triple overkill. I just bought a 117gph fish tank pump from some pet supply and y-connectors as well as tube shrink from hardware supply for less than $20. We'll see how it goes in about a week. Since most run system with radiator/fan, and I plan to do it with a milk bottle in water tank (or just large water tank). I should be able to do it with less flux.

Phenom 9950 has a TDP of 140W, and it heats 500L water by 0.22C in an hour. Even if you reduce the amount of water by half, it's still only 0.44C. This is only in static case. Dynamic heat transfer analysis is way harder. The limitation of liquid cooling is how effective heat transfer from the target to water, and effective heat dissipation from water to the enviroment, not the amount of flux in this case.

[Bisley_Bob]

Re: Alex is going watercooled

The 1000l/hr was an example as it was a simple number, i just plucked it straight out of my head. Anyhow, if you are upping the voltage whilst overclocking (which alex will), you want maximum cooling performance, so halving the flow is a ridiculous idea. No overclocker just wants "enough for an average system".

Good luck with your budget loop, make a worklog thread on here, i'd like to see pics, how you put it together, performance etc..