Dual Loop, single reservoir?

I currently have the following:

Swiftech XT cpu block on a i7 950
2 x ek gtx470 block
1 x 360 XSPC RX rad
1 x 120 XSPC RX rad
1 x EK 150 reservoir

All of the above is on one loop. However the pump is struggling with flow and I have bought another identical pump with the intention of running in series. However, that has it's own issues/problems as I have noticed recently on these forums.

A dual loop is not ideal as either the twin gtx470's or the cpu would just end up being cooled by the single 120 rad.

So, I have thought about having a dual loop but using the same reservoir. The more I have thought about it, the better it seems. I would have a pump pushing the water through the cpu block and triple rad and a seperate pump pushing the water through the two gtx470 blocks and the single rad. Flow rates should be good.

With the water mixing in the reservoir, the water would reach an equilibrium wouldn't it? This would give me the cooling power of both rads over both loops? Say the water was getting hotter from the two gtx470's loop, once it hit the reservoir it would mix with the cooler water from the cpu loop. Or am I missing something?

Would this work or have I gone mad?

Note: Replacing the single rad is not an option as it would require case modding which I don't want to do.

It sounds good in theory... In practical application however, you'll need someone with more experience than me to say yay or nay... But as I said, the theory is sound

Ive had a dual loop into one reservoir and it works well due to the water travelling the whole loop.

You can buy a XSPC res that does this or maybe you could get a double top, ive used both and they both work well.

How are you planning to use the EK reservoir?

It's the Ek 150 advanced so has about loads of inlets and outlets. Just going to put the two returns running into the the top and feed the two pumps from the bottom.

The force of the water should mix it quite well in the reservoir.

Two pumps fed by a small res sounds like you could get issues.
Most would be durring filling, and removing bubbles will be a nightmare.
With D5's you could fill / bleed the loop on setting one then crank them up for day-to-day use.

But with DDC you'll need to clamp, hold the tubes or pulse the pumps to stop the res empting in under 2 sec's.

I'd highly recoment fully filling the res after it's bleed (with a fill line) this should stop any way of bubbles getting draged back into the loop.

The fill line will also act to increase the volume of the res.
but I'd guess that two pumps will move water faster out of the res than gravity can refill it.

You can buy replacement tubes from ek - keep your existing top and bottom just get a larger middle section.

shadowscotland said:

You can buy replacement tubes from ek - keep your existing top and bottom just get a larger middle section.

Click to expand...

The pumps are both laing d5 vario so should be easy on setting 1/2 to bleed.

Not sure what you mean by the tubes. I can never figure out whether I should use them or the anti turbulance widget thing. Never really figured what purpose the tubes were for?

And 150ml isn;t really that small of a res, well at least compared to my old swiftech mirco res

The loop holds over a litre of water so always had to try and fill the loop as much as possible first when I had just one pump and then run the pump on slow and top up the res frequently.

And also, why woul;d bubble removing be any harder than just one? I could always bleed/fill one loop at a time? Or am I missing soemthing (again)?

I have a dual loop single res using the old rev 1 XSPC dual ddc res. Works a treat cooling a i7 970, two NF200's on the mobo and 4 GTX470s.

Rads are 1 PA120.3, an RX240 and a TFC120

Glad to hear that Biffa. Wasn't sure if anybody had done it before.

So you reckon I will be good to go and better than going for one loop in serial with two pumps then?

I'd show you a pic but my photobox is out of action till the 21st

But I have a dual DDC res, so its got two pumps on it, but a single shared res of liquid, the newer version of the res I have has a split res, one for each pump.

Anyway I run one pump straight to the GPU's and then up to the 120.3 and back to the res. The other pump goes to the CPU, 120 rad, then mobo, then 240 rad then back to the res.

Seems to work ok, I get sub sub 60 on all gpu's and cores, at the loads I put the system under thats pretty damn good I reckon

I was taking about the body of the EK res - the 60mm acrylic tube.
NOT the little 16mm OD tube that can be screwed into threaded ports inside the X2 reservoir.

But with D5 on slowest you should be fine.

And you right, for a minute I was thinking you would be powering up both pumps at the same time.

Using a single reservoir will lead to some heat transfer between the two loops, but they wont be at the same temperature unless the reservoir is big / stirred. As such the best arrangement of inlet and outlet from the reservoir may need some trial and error e.g. if the flow rates are parallel and in the same direction there won't be much mixing but the flow rates should be at their highest. It's also possible that the heat transfer is enough without any mixing.

@Biffa, could you list the idle/load temps for that arrangement?

I'm probably moving from two pumps in series to two in parallel with a shared reservoir.

Jon: Trial and error is exactly what it took I think I tried every which way

Temps:
CPU: low 33°C - high 65°C
GPUs: low 37°C - high 61°C

Cheers man, it's really useful to have real-world numbers. Even more so when they demonstrate that a controversial approach is successful; I'm genuinely happy that it's working so well for you.

Once my replacement mobo comes back I'll post some figures up as well. It will be interesting to see how much an improvement on the single loop/pump that the system was in.

I think I tried every combination of pump/loop route/fan orientation/radiator order that I could think of.

I'm also running the gpu blocks in parallel as well which reduces the restriction in the loop caused by the blocks.

I've also got Scythe GT 1850's and 1450's on the radiators, push on the 120.3 and push/pull on the 120 and 240.

Biffa said:

I think I tried every combination of pump/loop route/fan orientation/radiator order that I could think of.

I'm also running the gpu blocks in parallel as well which reduces the restriction in the loop caused by the blocks.

I've also got Scythe GT 1850's and 1450's on the radiators, push on the 120.3 and push/pull on the 120 and 240.

Click to expand...

I have often wondered about running the gpu in parallel but can never get my head round why it should work. Water will want to take the shortest, easiest path so I always imagine that one gpu block would not get as much flow as the other one?

Or is this yet another thing I am wrong with?

You would think so but temps are stable across all 4 cards.

That being said I don't think its worth it unless you have three or more cards.

Biffa said:

You would think so but temps are stable across all 4 cards.

That being said I don't think its worth it unless you have three or more cards.

Click to expand...

reason?

I would advise a single loop with dual 3.25 pumps with a dual top.

Greebo said:

reason?

Click to expand...

Because it decreases the flow speed across the blocks, what you end up doing is balancing the restriction against the overall flow, now I don't have experience with trying it on say two cards, but two cards shouldn't really offer enough restriction to worry about, but I know with 4 cards and the blocks I use it provided a slight increase in cooling performance.

As you can imagine the number of variables of 3 rads, two pumps, a shared res and 1 cpu block, 1 chipset block, 3 mosfet blocks and 4 gpu blocks meant that I was scratching my head a lot of the time wondering why one part of the system was much hotter than the rest.

I ended up much more worried about the cpu temps than anything else, I felt confident early on that I could provide 10-20° better temps on the gpu's than the air coolers they came with, but I struggled with cpu temps for quite a while.