SLI bridge - What's the best way?

[wrenaudrey]

As said in the title, which of these two is the best way of connecting two cards together?

From my sig, i have used XSPC Razor 780 block. So its natural to go with that block again for the 2nd 780 that's going in my rig.
Both cards will be WC'd, not how, but what's the best way out of these two?


*Not my images -- from Google Images by typing "xspc razor flow bridge sli"*

This setup?

Or this setup?

Thanks all in advanced

 

[Dragolu]

I guess your asking serial vs parallel. Ive used both setups in the past and as long as the pump is decent it doesn't make a big difference preformance wise. Aesthetically im partial to the parallel.
Parallel will decrease restriction (pressure drop) but flow through each block will be slower. It may make a difference if your loop is very complex or you have other high restriction blocks but it doesn't make a difference for most.

 

[Mysterae]

Dragolu has it pretty much covered. Before I tested it thought serial would have been much better, but it's negligible. One card is always about 2C higher than the other, regardless of the arrangement. I think the warmer card performs more duties for SLI.

Serial:

Parallel:

Decide with aesthetics .

 

[The_Governour]

Nice! What temps are you getting on load with the XSPC WB?? I want to get just the same one as I think the led option is very nice!

 

[Mysterae]

Nice! What temps are you getting on load with the XSPC WB?? I want to get just the same one as I think the led option is very nice!

Which one . Rather than derail this thread I'd suggest you head over to my build log (link in sig) as the answer isn't that simple. They are good blocks though! Remember that with the chrome version of the Raystorm you won't get the acrylic mount (therefore the ability to mount leds), you'll have to get that separately. I got mine from ebay.

 

[Scoob]

In my experience, Parallel vs. serial on the GPU makes little temperature difference - assuming the loop has sufficient capacity of course.

However, it DOES make a difference in regards to FLOW. In my setup, I have a great pump, and both the rad and CPU Block are very high flow. The GPU blocks, when in serial, DID limit flow somewhat as they were the most restrictive point in my particular setup. Switching the GPU's to parallel flow removed that limitation, improving CPU temps slightly and the pump seemed a little quieter.

Again, this is a small difference, but I felt having a high-flow CPU block, a high flow (1080) rad, a great pump and using the larger tubing, would have been a little wasted if one part of the loop was restrictive.

My aim with my build was to get excellent flow so the pump would not be stressed in any way, as well as be near silent but still be able to cool a HOT overclocked system more than adequately. I'm still pleased with the results over two years on.

I helped a friend with a similar build, but using three GPU's, and it's still working very well there too.

Cheers,

Scoob.

 

[wrenaudrey]

Hello again! and im sorry for reviving this dead thread.
Thanks for all your advice on the serial or parallel matter.

I went with parallel in the end due to aesthetics.

Finished the loop yesterday and only found the time minutes ago to play some BF4.
I always have monitoring programs opened when im gaming and i noticed that once i started playing, the temps on the second(bottom) card is unusually high.

Here are some pics:

This is the setup.

This is the screen shot of the unusually high second GPU

On idle, they are both great.

Any suggestions what the problem might be and how to solve it?

 

[SparkleDJackson]

that block must be fitted wrong bud. both my 780 6gbs sit at about 25 idle and barely hit 45 when cranked. i have more radiators like bu still. did you do a test fit? big air bubble maybe. i know my parallel setup produces the odd air bubble i have to bleed every once in a while.

 

[Scoob]

Possibly you have an air lock in the lower block - I'd be running the loop but not the PC and shaking things around a bit to free it it.

Equally, it could well be the fitting of the block as Sparkle suggests, though that's a bigger job to check and rectify.

Regardless, I'd NOT be stressing that system at all with those temps on the lower GPU. Upper is fine, but you shouldn't be seeing more than a 2c difference between them really.

My loop is plumbed similarly, but to an external rad, though I use individual SLI connectors rather than a block like you. Are you 100% sure your SLI block supports parallel flow? When I was looking initially, some of the EK ones did not. I'm sure you got the right part, but a quick sanity check doesn't hurt. Still, even if that block was flowing sequentially, it'd not explain those temps.

Best of luck & do post back!

Scoob.

 

[wrenaudrey]

Im really puzzled with this one.
Im not entirely sure if its a stuck air bubble because i've shaken it, tilted it sideways on bothways around 75deg and tilted it backwards and forwards around 75deg too.

Shall i try going to serial first?
And if that fails, re do the bottom block?

 

[Scoob]

One thing to remember about parallel flow is that ONE of your GPU blocks could be almost entirely blocked yet the rest of the loop would appear fine.

If you're confident that you've given shaking it a fair go then, yes, I think it's time to drain and pull the loop apart. Double check the SLI link, make 100% sure it allows the flow you think it does - I have seen such links with unexpected restrictions, i.e. they'll parallel flow when connected in one way, but not in other. If you're happy with that, I'd say carefully strip down the block from the GPU & see if something is amiss there.

It is possible, though unlikely, you've left the plastic backing on one of the thermal pads - IF your particular block uses them for the core GPU - or simply that the paste is poorly applied (air trapped) or something simply isn't screwed in right and applying pressure to the GPU as it should. I've used blocks where you screw in little uprights first - like the ones that go into your case to support your motherboard - one of these, cross threaded would lift the block slightly even though the screw would appear to be firmly in place.

Stripping down the entire block should highlight the issue hopefully.

Best of luck!

Scoob.

P.S. Late here, so I'll check the thread tomorrow.

 

[wrenaudrey]

One thing to remember about parallel flow is that ONE of your GPU blocks could be almost entirely blocked yet the rest of the loop would appear fine.

If you're confident that you've given shaking it a fair go then, yes, I think it's time to drain and pull the loop apart. Double check the SLI link, make 100% sure it allows the flow you think it does - I have seen such links with unexpected restrictions, i.e. they'll parallel flow when connected in one way, but not in other. If you're happy with that, I'd say carefully strip down the block from the GPU & see if something is amiss there.

It is possible, though unlikely, you've left the plastic backing on one of the thermal pads - IF your particular block uses them for the core GPU - or simply that the paste is poorly applied (air trapped) or something simply isn't screwed in right and applying pressure to the GPU as it should. I've used blocks where you screw in little uprights first - like the ones that go into your case to support your motherboard - one of these, cross threaded would lift the block slightly even though the screw would appear to be firmly in place.

Stripping down the entire block should highlight the issue hopefully.

Best of luck!

Scoob.

P.S. Late here, so I'll check the thread tomorrow.

Im gonna do redo the waterblock and gpu.
Fill and bleed with DI water and test it out.

If the problem persists, ill try serial.
If that doesnt work out, idk what the heck would be the solution.

Ill update when i have any progress.

 

[JR23]

If it were me I would make a point to swap the cards over when you put it back together. If the bottom card is still hot then you know its a flow issue, if the top one becomes hot then it's a block problem and if they both cool down I think it would be safe to assume it was a poor seat/trapped air.

Just looking at it the SLI link makes sense although I don't like how the bottom card uses a different exit on the side. In both an aesthetic and thermodynamic context that really offends me.

JR

 

[wrenaudrey]

Hello again! Just an update here

I drained the loop and reseated the then bottom card and also as JR suggested swap the then top and bottom around.

I 97% bled the loop and ensuring that there is little to no air on the GPU blocks.
I went through the whole of Baku(1st mission on BF4) and these are the temps:

Bearing in mind that the top GPUz was the bottom one which had the problem of that ridiculous 90-95deg playing temp. And the bottom GPUz was the top one before i took it apart.

Regarding the temps, i guess i dont really mind it being <70deg on everything as i do now have one additional card that's making heat. And also all the fans are running on 5v.
But it is still something im slightly uncomfortable about as previously when i only had one card, the temps on that never went >60.

On idle, they're great, more or less neck and neck.

So what're your thoughts guys?

 

[SuB]

Those temps are daft, no way should it be getting above 50 let alone 60 mate.

You have plenty of rad to cool those sufficiently, something is wrong here still I'd say

I have the same blocks on my gpu's and they do not go above 40 at the moment (not overclocked) and I'm on dual 360 rads that are a little thicker than yours, but still 60 + seems too high to me mate.

When you cut load to the cards, how fast do they cool down?

 

[Scoob]

Yes, something really isn't right with your loop - those temps are a little insane for water cooling. My GPU's never go more than 16c over ambient, so, they get into the mid-40's once the room is REALLY hot from excessive gaming.

SuB has a good point. If the cards cool fairly quickly once off-load it suggests your loop is dissipating the heat ok, and the block just isn't getting the heat into the coolant effectively. If they DON'T cool rapidly once off load, it suggests the loop is failing to disperse the accumulated heat.

Previously, you had one "bad" GPU temp, now both GPU's are performing poorly heat-wise.

When you say "reseated the card" are you talking about the block here - i.e. removing and re-fitting - or just physically changing slot? If you HAVE re-seated the block I'd suggest you have some sort of flow issue. Your CPU temps seem a little high, and the fact that the max for all core is exactly 64c strikes me as odd.

Not sure what to suggest, if you were close to me I'd pop round and give you a hand - it's difficult to diagnose issues like this remotely.

Hopefully others have some more suggestions.

Scoob.

 

[Kaapstad]

I would ditch the parallel flow for starters.

From your pic you don't have any fans on the case side of the top rad, maybe you could swap them around. You have got the fans on the top Rad set to exhaust ?

It might be more efficient to have the loop go straight from the CPU to the GPUs and then round the system.

Are the fans on the bottom Rad intake or exhaust ?

 

[Mysterae]

If you are at a loss there is something else to try, but it's a bit of a ball ache - test the system with only one gpu in your loop. That'll mean taking one of your cards out of your case completely.

As others have said, a watercooled gpu shouldn't reach 50C under normal game load.

 

[Warchild]

To be honest. if your system is on idle as you say. Even your CPU temps like a little on the high side. They should be nesting at around 30C avg across all cores too.

Sounds silly but I wonder if you have a miner infected on your system. I had it back when I was using fans and noticed that once I opened a game, my fans went crazy working much harder than they used to.

Aside from that check the seating of the waterblock on the GPU. And dont add too much TIM as this can be just as bad as not enough.

 

[Z3r0]

first of all is just an opinion...i m not so prrof in water cooling..i did some systems before but i like to think the things before i do something. Now in my opinion i dont see the reason for paralel bridge..and thats because if you was enter by the top of first card and also exit from top..well yes i was understand your concern about "how it looks". But you have lest say astandard way of your hose so i think you can try serial and see what is hapenning.

Now the 2nd ideea is that reservoir. Tell me something..the reservoir is giving liquid to the pump and the pump go into the rad from bottom?. I know that reservoir..i had myself same...You can try to upside down it and go from top into your cpu. You have 2 g 1/4 on his botton so thats mean you have place to fill it up. In my opinion the pump is draining the liquid very fast and your return from cpu is near the port where your pump is taking liquid..so it i is possible that the presure into your system is not 100%. Dont understand me wrong...the ideea is that the system dosent have enouf presure to pull air bubles around the loop. It have the preasure to run but if is not 100% sometimes the air bublees remain in same place.