20 degree difference on 2x 580's in loop

[Squigly]

Hey all

Needing some advice here, I am finding that my 2nd card in the loop is running 20 - 25 degrees hotter than the first card in my loop, Even at idle there is a 4 - 5 degree difference.

My setup is RES>DUAL PUMPS>RAD1>CPU>RAD2>2 GPU's>RAD3>RES

I am using EK blocks for the 580's and have made sure they are seated nicely and applied some fresh TIM, even changed order of the cards in the loop and the 2nd always gets hotter.. I thought it was due to the 1st card heating the water to much but from most of what I read it shouldn't raise the temp by that much since the water reaches equilibrium after a while. I was originally using a Serial SLI bridge from EK for the loop and then ordered a Parallel bridge hoping it would help.... it didnt...

With air cooling I can understand the top card getting hotter as the heat rises, but this is the lower card or 2nd card in the loop... I have also shaken the crap out my case to make sure there are no air bubbles.. but with the 2x D5 pumps the flow rate pretty much pushed all air out anyways.. Any ideas peeps??

 

[George S]

What pump/s are you using? (Brand and model)

Are you using Coolant? Perhaps the blocks are clogged?

 

[P1000]

Tried swapping the blocks to see if the problem moves? My guess would be it is some problem between the block and the GPU chip... Possibly between the GPU chip and the heatspreader(if it has one)

 

[Airnews39]

Do both cards have a IHS? (heatspreader) there will be a small gap between the block and the card if it doesn't. I reckon this is your problem.
EK makes a separate block for the 580's without the heatspreader.

Check this thread out regarding the heatspreader issue... Already RMA'd once, can I please get a GTX 580 with a IHS

 

[milkshakes]

@george s:he said he is using d5's (more than enough power).
it can only be clogged blocks or a faulty gpu...
open up the block(s) and have a look.
if none of the above...gremlins...

 

[George S]

@george s:he said he is using d5's (more than enough power).
it can only be clogged blocks or a faulty gpu...
open up the block(s) and have a look.
if none of the above...gremlins...

I Would still like to know waht pump he is using, I have the MCP355 and it just amazes me how strong it is every time.

I do think it is a clogged block IMO.

 

[milkshakes]

forgive me,i thought people into watercooling know that d5 is short for laing d5,wich is known by many names : Danger Den "DD D5",Swiftech "MCP655" and Koolance "PMP450"

 

[Archer]

Two different gtx580's? One of them happen to be evga? I half remember them dropping the IHS

 

[Airnews39]

It seems that the lack of IHS is pretty varied and there seems to be no hard & fast rule about which cards have them and which don't. My EVGA 580 has the IHS.

 

[Squigly]

Hey all
the GPU's are identical, they are both Gainward GTX 580 3GB Phantom's (so the non reference design) I am useing the full EK blocks (specifically designed for the gainward non reference cards) on them so all memory etc is covered by the blocks, the cards are only 4 - 5 months old and the blocks around 2 months so i doubt they clogged.. at least i would hope not yet.. Granted I do not have any inline filters in the loop.. but really clogged so quick??

as for the pumps, they are the swiftech MPC655 (aka D5 as Milkshakes pointed out above)

I had the 2nd card handling the Physics in my SLI setup and tried swapping that to the primary card (BTW i am only running a single monitor at the moment, overkill yes i know.. but there are CUD's to be done *grin*) But even with the primary card handling the physics there was still a 20 degree difference.. and since the blocks are running in parallel its not a case of water heating to much.. I had sent the cards back some time back for them to be checked and i got an all clear on them.. may have just been a lazy supplier that didnt want to take the hit though....

EDIT:
Sorry just to clarify, the cards did have IHS, but the waterblocks are designed to fit directly to the card not the IHS.

This is the exact block I am using on both cards

 

[Airnews39]

Have you swapped each block to the other card?
At least it will help to determine if the block or the card is the problem.

 

[Squigly]

was thinking that may need to be my next step... will give that a try over the weekend.. need to drain the loop to get the cards out (slaps forehead for not ordering those quick disconnects.. )

 

[Airnews39]

Quick Disconnects are the bomb! Loving having them in the loop!

 

[Squigly]

Right, finally got some time to drain the box.. Seems my last TIM refresh didn't go as well as I had hoped...
View attachment 5399View attachment 5400

Seems that one tiny little corner of the GPU was my issue, cleaned off the card and block and re-applied.. Now only getting a 8 - 10 degree difference on the cards under load (down from 20)... amazin what just a few mm of TIM can do....

 

[milkshakes]

wow,it was just that...didnt see that coming.

 

[Archer]

Are you sure you've got the parellel bridge setup properly? I saw on XS a guy had plumbed it incorrectly and was also having temp issues. As for the TIM, I doubt it was that. I think you may have not applied enough pressure (tightened screws), but hey, as long as you've got an improvement, who cares what fixed it

 

[Squigly]

Yea when i first saw it i was doubtful that that much TIM coverage would have had a 20 degree impact... But glad there was an improvement!! Im pretty sure I have plumbed them correctly.. well at least to the installation guide, and it seems to make sense when I looked at the flow..

View attachment 5410View attachment 5411

 

[milkshakes]

looks correct to me...

 

[Fluke]

Any kinks in the tubing?

 

[Squigly]

Nope, kink free

 

[Alpha_Renji]

what temps are you getting btw? Keen to know

 

[Squigly]

The 1st card or top card hits 65 - 70, the 2nd card or bottom card hits 85 - 90

 

[Alpha_Renji]

Ah ok, interesting, for what it's worth, here's my temps on my WC 480's.

I have the 480's with XSPC blocks and I OC with 1.138v (afterburners max voltage) and my temps are both 53'C load (under my aircon) (60'C without aircon on). Cards never differ more than 3'C under load. Also, I use a 2x rasa kits in dual loop (gpu/cpu), only 750l/h pumps.

Without OC'ing I hardly see temps higher than 50'C.

With those pumps and blocks, something tells me you should be getting better temps :|

 

[milkshakes]

i agree,i recently put my gpu under water (figuratively) and almost the entire loop is the same temp...

 

[Airnews39]

My 580 hits mid to low 50's on a hot day & it's OC'd @ 1.138v

 

[Squigly]

That's what I was thinking.. the Gainward Phantom was supposed to be a "cooler" card with its non reference design power management rah rah.. but even under air I found the cards used to hit 97 rather quick when under load.. from what i could find in some other forums these temps were stated as "normal". I decided I wanted to water cool them and was expecting a better drop in temps, I guess the card hitting 65 is a somewhat decent drop. Also if I disable SLI and run a single card then it normally stays just below 60. But that second card seems to heat way to much with SLI enabled.

I am currently running the setup in a single loop to try and keep piping to a minimum, and the CPU never really breaks the 70 mark. But the fact that I am running a parallel bridge on the GPU blocks and getting different temps, specially on the lower card is what is baffling me and i find after about an hour of gaming the water tubes are fairly warm to the touch

Oh yea, before I forget.. this is all with stock clocks on the GPU's... me thinks perhaps that 2nd card is a bit poke'd..

 

[Oj0]

Swap the cards around and see what happens. If the bottom card is once again hotter I'd look at the block, if it's now the top card I'd look that its heat-spreader is flat.

 

[Squigly]

right... finally swapped the blocks on the cards, also separated my CPU to in its own loop now, just works out simpler to drain since I'm working on the cards.. on the plus side with separating the loops my CPU now only hits 58 clocked at 5Ghz (was low 70's)

with the blocks swapped I am still getting a 20 degree difference (card 1 55deg, card 2 75deg) so its definitely not the blocks, its the same card that heats.. so me thinks the 2nd card is definitely somewhat fubar!!

Think my only option now is to drain... again .. re-fit the air-cooler and take the card back and try get it swapped.. what u guys think?

 

[Alpha_Renji]

GL with the swop Custom WC voids warranty last time I checked :| Best hope they don't notice that

 

[Oj0]

Unless you're sending it with the water block attached as a freebie for them to keep they're not going to know you ran water cooling