Difficulties bleeding system

[grandpatzer]

So Ive been using external PSU and tilting the case in any totation possible and there still must be air pockets in top GPU because it gets hot.

I was suggested that it might be bad block install but I watercooled GPU's invidually(so 1 gpu only in loop) and the temperatures were wery good (42c furmark).

So I recently tried again with serial link and the top gpu got to 65c after 5min of IDLE and my pump is pushing at highest level possible (D5 vario).

I'm not sure how to get the air out, maybe the pump is not good enough for 4blocks + 2rad?

 

[rubix_1011]

So, individually, each card is cooled well? That rules out any issue with individual block mount/contact.

That pump shouldn't have a problem pushing through that setup. Have you tried running flow as suggested on the other thread, CPU, then down through GPUs and through the rads? CPU block is the most restrictive, so your pump should really push through it first.

 

[grandpatzer]

Ah ok did not know that, so the optimal would be Pump -- CPU -- GPU -- Rad -- Res -- Pump?

Does that help me bleed air bubbles from the GPU blocks?

So going first to CPU gives less resistande for pump so I can also run in lower?

I'm also considering just putting a air Mugen cooler on the cpu.

 

[rubix_1011]

CPU blocks are typically the most restrictive component in a loop (unless you are running RAM or MB blocks) so by having your pump push through this first, it should take advantage of the most head pressure being put out by the pump. In theory, it shouldn't really matter, but it seems like your GPU blocks and bridge are causing an issue by flowing how it currently is.

Radiators are usually the lowest restriction components in a loop by comparison.

 

[wolfram23]

Air bubbles should get forced out from the flow of water unless it's a really bad design or the flow rate is too low.

I can't really tell the order you have that hooked up, some of the connections are not visible. Can you detail it?

 

[grandpatzer]

Res ---> Pump --- bottom rad ----> GPU's ---> CPU --- Front Rad --- Res --- Pump

 

[grandpatzer]

As I said one card any card I get 42c in furmark.

when 3cards in IDLE I get 80c when I decide to turn off pc.
I also rearrange cards, this time the hot card is 30c but the new top card is 80c and bottom card 60c.

 

[rubix_1011]

When you run a single card, I'm guessing you run it without the bridge block, correct?

Are you absolutely sure you have the bridge setup correctly and nothing is blocked that shouldn't be?

 

[grandpatzer]

Yes I did it 100% as the manual said.
The Bridge came with 6washers but in manual no say where to put them..

I have interchanged inlet and outlet which is OK according the manual.

manual: http://www.ekwaterblocks.com/shop/EK-IM/EK-IM-3831109855096.pdf

http://www.ekwaterblocks.com/shop/blocks/vga-blocks/fc-connection-parts/ek-fc-bridge-triple-serial.html

 

[grandpatzer]

oh yes also no bridge when one card, just 2 barbed fittings.

 

[rubix_1011]

OK, this is really perplexing.

Does your CPU get hot and build up heat similar to the other cards? I would assume that the CPU would get hot as well, since the first card (bottom remains cool) and progressively the GPUs get hotter in the middle and top. The CPU should also heat up as well...is this the case?

Can you pull the cards with the EK bridge still installed...then blow air through the inlet on the bridge and see how difficult that is and how much air exits the outlet?

If this is fairly easy, and you are reinstalling the cards and EK block, when you go to fill/prime the loop, try this:

Take your CPU outlet tube (that goes to your big rad) and disconnect it. You might need to run another piece of tubing to it that will siphon from a bottle to fill.

Try to siphon water through the loop as much as possible until you have water through your CPU block without a lot of air.

I still think you should change your pump outlet tube to run to your CPU inlet -> CPU outlet to your top GPU bridge inlet. Just make sure you complete your loop if you decide to go this route.

 

[wolfram23]

Must be something with the bridge.

Is it set up in serial or parallel? And the 6 washers... do you mean they are metal? Or are they rubber? If rubber they should go between the bridge and the gpu blocks.

I'd try taking the bridge off and pouring water into each port individually to make sure it's going where it should be and not leaking through.

 

[rubix_1011]

^Yeah, something isn't right with that bridge. He said it's the serial bridge, and I believe it's milled to be serial-only. I think he'd have to have the parallel bridge to run parallel (which would be a better option) but either way, this shouldn't kill flow like this.

 

[grandpatzer]

yes the cpu temps were similar to the middle card so under 40c.

I think the reason for the Top card being 80c is because it was main card, as for the bottom card it's the other 16x PCI-E so maybe it is also in use.

Obviously in IDLE in windows 7 the cards won't use alot of gpu usage should be 0-4% at most gpu load.

I got tired and needed a PC so I removed and only running one card now, I think the flow is really bad in the Bridge and gpu's because with just one card without bridge I can see tiny bubbles move around in the block.

With 3 blocks and the bridge the airbubbles don't move but if I press the CPU - Rad tubing the air bubbles move, also the CPU blocks the tiny air bubbles move.

I will put the serial bridhe on shelf, maybe one day I will test out the rerouting now I have spent alot of days with this and tired with it

 

[grandpatzer]

Come to think of it is it possible the water does not go throught the GPU blocks at all?

so it goes: Pump ---> Rad --- Bridge(not gpu blocks) --- CPU -- Rad --- res.

That could explain why air bubbles move in cpu but seems like the blocks have no flow.

As for the bottom card being 60c and middle card 30c maybe the water goes through the middle card because I had no Crossfire bridge so only the top card should have gpu load.

 

[Motopsychojdn]

Maybe redundant by now, (din't quite know how I missed this thread)
with the system off, rock the case to and fro, can you hear ANY sloshing?
make sure res is 100% full for this or you will hear the res, not anywhere thats got a lock in it,
I would also look at taking the Gpu link out, and tubing the cards up 'normally' (you'll need more barbs if you have none spare lying around)
just to rule out the link being the problem.
Moto

 

[rubix_1011]

I'd suggest powering the loop with the PSU jumpered until you are certain you have flow through all the cards.

Are you sure you weren't sent a parallel bridge on accident, but have flow routed as-needed for the serial block? This would be a huge issue and likely explain the majority of your issues, if so.

This should be simple to find out- the parallel bridge would have a straight shot from the top all the way through. If you had a pipe cleaner or something long and straight- you would be able to stick through the top and see it in all 3 holes going down without any re-routes. A serial block would have U-bends for every single junction.

Also, you should consider changing flow order to go through CPU first, then down to GPUs. I don't agree with this making a lot of difference, but it might help with any restriction issues.

 

[toolmaker_03]

hay there still having problems with this new build I see. the order of components can help like they have already stated, go from most restrictive to least. but that doesn't explain the air in your system or the pumps inability to pull air through the system. it should become a negative pressure system when filled and sealed meaning that if there is a leak it will pull air into the system while at the same time releasing water out of it. I have never really considered the bio res. as a reservoir because it is more like tubing than a res. but to each his own on that point. this won't cause your problem, it sounds more like you have a slow leak. let me ask you this, does the pump run faster when you first fill the system and then it slows down after a couple of weeks, and at the same time it seems to have more air in it. if so this may be hard to find since you won't see air rushing in at the point of origin, like you would see on a bad leak.

 

[grandpatzer]

Thanks for all the suggestions, I took out the bridge couple of days ago and got tired with it because it took several of my days and hours spending energy on this

If I ever decide to reinstall that bridge I will all these suggestions.

As for me getting wrongly parallell is no difference as the fitting installations still are correct in my setup for parallell as serial according manual.

 

[rubix_1011]

Yes, but with parallel the path of least resistance would be straight up through the bridge and not through the blocks themselves.

 

[wolfram23]

Also, if parallel is working properly, you'd get 1/3 the flow rate through each GPU compared to the inlet/outlet flow rate. Personally I think serial is the way to go, even if the water is slightly warmer by the 3rd card.

 

[rubix_1011]

Yes, but you'd have lower overall restriction with parallel vs serial...so it's a matter of what is best for each loop. I just can't figure this one out...something doesn't mesh well and I think it starts with not having the CPU first from the pump outlet...but without more information or seeing it in person, it really is hard to tell for sure.

 

[toolmaker_03]

Also, if parallel is working properly, you'd get 1/3 the flow rate through each GPU compared to the inlet/outlet flow rate. Personally I think serial is the way to go, even if the water is slightly warmer by the 3rd card.

I agree serial is the way to go, but your having issues that don't make since if the configuration of the card to card water flow is the issue. it is more like the first card slows the rate of flow through the rest therefore lowering the ability of the water to cool the cards, as you have stated the middle card is not used as much in this configuration. so that does explain the temps, so what is it about that first card that is so restricting to the flow of water?(they do make large pumps with insane pressure ability for systems that are restrictive buy design) this may be a simple solution to the issue.

 

[Lutfij]

i'm going to skip all the posts above and post my finds no that i'm being arrogant or anything

i went onto frozencpu and gave it a try to locate the culprit bridge grandpatzer is mentioning of.

a quick search revealed that the bridges come in two flavors. Serial and parrallel for either crossfire or sli. My question to grandpatzer is, which one dya have> serial or parallel?

if any of them they should be setup accordingly.
Advise #2 - run system in reverse order than whats shown in pics. and yeah run it according to bridge specs. if parallel - then parallel, if serial - run serial.

advice #3 - to bleed the cards, you'll need to run only the pump by paperclipping the psu atx plug rock the case back and forth while you see the air bubbles coming out.

Boiler1990 - on another thread of yours mentioned that the EK blocks can retain some air and yet have water flowing through the blocks.

compared to one card you'll have more air to push out in those three cards.

* something doesn't go well like what rubix mentioned - too many thread dealing with the system running hot... :/ or is it installation error?