Corsair 900D WC Build Log

[Zoinks]

He has to drain and disassemble the loop to swap the blocks, there's no reason I can think of why you wouldn't want to open up the blocks to check for some kind of glaring fault before reassembling the loop and restarting the bleed/leak test process. Just take the extra step while it's all torn down.

I was thinking more along the lines that taking the block apart is something else that could potentially go wrong and cause a problem. Were it me, I would swap the blocks first and only take it apart if the overheating followed the block to the other card.

 

[Zoinks]

What I think(it may be wrong) is that something is preventing solid contact between the block and the card itself.

Certainly a possibility. But it is also possible that the block you have mounted to the card that runs hot has a manufacturing defect such that it will not make good contact with the GPUs no matter what card it is mounted on. Swapping blocks between cards will test both hypotheses, whereas taking the block apart alone will not. This is also partly why I suggested you swap the blocks first, and then take block apart on the card that runs hot only if overheating follows the block to the other card.

Pitch troubleshooting ideas, but please, don't expect me to do them immediately and then move on to something else. That's what I'm seeing, multiple methods of troubleshooting then response after response, you guys seem to rule them out and move on to complex methods when I haven't done simple ones yet. I may have misunderstood what you guys are trying to do, but hey, I'm voicing what I think.

I am focused exclusively on your GPU temps and at this point I haven't ruled anything out. How can I? I'm not the one working on your system. The best people can do here is make suggestions and share their ideas. It is entirely up to you to use that information as you see fit...or not. It is also up to you to exercise due diligence and some requisite care in assembling your system. I think you get all of that.

Cool card(Top 590) is getting regular dual GPU idle temps at 24-28, each GPU, on the card, varying through ambient temps. Flow doesn't seem like an issue considering deltas and fan speed, which is running in a quiet state, about 30-50%.

Hot card(Bottom 590) is getting temps that are weird. One GPU on the card is usually 8-10 degrees hotter than the other. Ex. GPU 1 on the card, 44 C, GPU 2 on the card, 36 C.

And at idle that is fairly common, hitting max clock speeds, not at full load, they EXPONENTIALLY rise in temps, but in the same range of 8-10 degrees. HOWEVER, under GPU benchmarks, these suckers almost hit 100 C in sync after 5 seconds of Heaven(Unigine that is lol). I'm just rambling on about what has happened in my experince with these and it's probably not helping anything and just is extraneous words for you guys to read.

I can't say that I know a lot about the GTX 590. I have no personal experience with them. However, I did take a look at a review of the card on guru3d.com, and with the stock reference air cooler they came up with 37 degrees idle and 79 degrees under load.

http://www.guru3d.com/articles_pages/geforce_gtx_590_review,8.html

So your current temps on the hot card are higher (both idle and load) than a box stock GTX 590 on air.

Secondly, I don't recall seeing any load tests on your CPU. If I'm wrong about that please feel free to correct me. Your CPU idle temps look ok iirc, but if I were working on your system I would like to see load temps as well with Prime95. If CPU load temps are good then the problem is most likely not loop related and you can focus your attention on that hot running video card. I also don't recall you mentioning the load temps on the video card you say runs cool. You can use EVGA Precision X to log your GPU temps and RealTemp to log your CPU temps, or Speedfan to log them both during testing.

Good luck...

 

[Eugenius88]

So you're saying the same GPU that had heat issues before in your first old loop is the same card that is overheating with the new WB on it?



Some bubbles can take days, weeks and months to remove.



No. Figure of speech but I figured you'd understand by shaking your waterblock around like a radiator to possibly loosen anything up in there. Better off taking it apart.



Thanks for understanding and being honest. I've been assisting you and following your thread since the beginning to help you get through this. I noticed alot of laziness so I felt my time was being wasted. That's why I said what I said. Thanks for not taking it personal as it wasn't intended.



This explains alot. 3 things are possible. Most obvious is contact. Your temps changed when you tightened your backplate as that connects and tightens the waterblock in front of the GPU. Make sure the screws are tight enough where you stop once they don't need to be more tightened.

You also said some of the TIM didn't spread. Is this on the GPU? This is bad contact than. You have contact issues. Make sure your cooling pads are in the right places as well.

The block can also have debris in it and bubbles can also be the culprit. It can be all those 3 combined. If you're scared to take the waterblock apart and clean it than you should get someone else to do it for you and make sure the washers line up when putting it back together. Test it for leaks before powering everything on again.

I think you figured out how problematic watercooling can be. Well you're almost at the finishline. You've learned alot. Finish up when you have time for it but do it right and don't miss any steps.

Ok, I just discovered something relieving, even though generally it shouldn't be. I saw a bunch of little bubbles comes out into my res once I turned on the computer. Temps dropped 3 C. Ambient is actually higher than yesterday too. Yesterday = no heater on. Today = heater on. Temps still rise quickly so there may be that big bubble still lodged in there.

I also talked with my friend whose dad did his custom water cooling loop for him. That was his first time too. He said how he was getting high temps on his GPU. So his dad was priming the loop like a mad man and got it to normal. I'm not sure if he was getting constant high or an exponential high like me. I'll figure out this weekend. I'll try my best to prime this 80lb sucker. This question is most likely a stupid question, but can you successfully get bubbles out when fill port is closed? I tried searching it up but didn't find anything, I feel like you could though.

Sorry to be disorganized BUT YES! Remember the thread I had with my 590 getting up to 100 C by opening up chrome? Yeah, I didn't prime it at all. Even though it was a small loop, it seems really likely that was the culprit.

ALSO
TIM didn't spread when I had when I had the alphacool block, but what I'm starting to think now, is the bubbles blocked the heat distribution so that's why high temps were there. To be clear, I have XSPC blocks now, they're lovely and better than the Alphacool/EVGA ghetto setup.

Nothing personal taken Jack, I can accept I'm lazy, I hate it, but that's how I am, sadly.

 

[Eugenius88]

Certainly a possibility. But it is also possible that the block you have mounted to the card that runs hot has a manufacturing defect such that it will not make good contact with the GPUs no matter what card it is mounted on. Swapping blocks between cards will test both hypotheses, whereas taking the block apart alone will not. This is also partly why I suggested you swap the blocks first, and then take block apart on the card that runs hot only if overheating follows the block to the other card.



I am focused exclusively on your GPU temps and at this point I haven't ruled anything out. How can I? I'm not the one working on your system. The best people can do here is make suggestions and share their ideas. It is entirely up to you to use that information as you see fit...or not. It is also up to you to exercise due diligence and some requisite care in assembling your system. I think you get all of that.



I can't say that I know a lot about the GTX 590. I have no personal experience with them. However, I did take a look at a review of the card on guru3d.com, and with the stock reference air cooler they came up with 37 degrees idle and 79 degrees under load.

http://www.guru3d.com/articles_pages/geforce_gtx_590_review,8.html

So your current temps on the hot card are higher (both idle and load) than a box stock GTX 590 on air.

Secondly, I don't recall seeing any load tests on your CPU. If I'm wrong about that please feel free to correct me. Your CPU idle temps look ok iirc, but if I were working on your system I would like to see load temps as well with Prime95. If CPU load temps are good then the problem is most likely not loop related and you can focus your attention on that hot running video card. I also don't recall you mentioning the load temps on the video card you say runs cool. You can use EVGA Precision X to log your GPU temps and RealTemp to log your CPU temps, or Speedfan to log them both during testing.

Good luck...

Good card hits 60 C load, I hear that's normal for 590 and 690.
MAX Average for CPU at 3.8ghZ: ~65 C pretty good considering when I rendered with my all in one back on the alienware almost hit 90 C while rendering at 4.1ghZ lol

 

[Eugenius88]

Something I forgot to mention, whole case is open, both panels are off, fyi, if that's any importance.

 

[Eugenius88]

At this point it has to be a bubble. Bubbles keep popping into my res after I start up the pump. I'll try this weekend to work this out, preferably Saturday evening.

 

[JeremyCT]

It needs to be bumped and turned while running if you're still getting bubbles. The shock/bump is important as that's what knocks the bubbles loose, kinda of like tapping on the side of a bottle to get the bubbles moving.

Tilt it one way, bump bump bump, tilt it the other way, bump bump bump. See what happens. It's safest to do this with only the pump running and the system not running.

Based on your comments on the TIM not spreading though, I don't have high hopes of that being the eventual solution.

 

[Eugenius88]

I did my black belt test, two days of it with success and I got my driver's license this week.
The black belt test has be bruised and sore and hopefully I can physically recover to prime my loop tomorrow. If I didn't mention before, this computer is almost 80 lbs.

 

[Eugenius88]

Well, the TV and Advil were my friends today. Oh well, I'll prime the loop once every joint in my body isn't sore. Cya guys then.

 

[Eugenius88]

Wow what a week. My awesome dog of only 6 years had to be put down and the day after I got hand-foot-mouth disease(still affected by it). I haven't used my computer since, but off and on I've been trying to bleed the loop, violently shaking it I'm some directions. I see very vague ways on how to bleed your loop and I still have bubbles.

Should I leave the case on its side, or both and run the pump? If so how long and in all 6 possible directions?

 

[JeremyCT]

Bleed it with just the pump running, you'll have the best results. Shake it, tap blocks, change position. Sometimes, despite all this, you'll have a bubble work loose a few days or weeks later. It's just the way it goes sometimes.

In my experience, running the pump at full speed dislodges bubbles the best, but it creates very small bubbles that don't rise in the reservoir. I like to shake the system with the pump on full, then turn the pump way down to let the bubbles rise in the res.

You'll find your own method that works for you via experimentation.

 

[Eugenius88]

Ugh, over a week, shaking this case side to side and tapping the blocks, temps rise 5 C overall. should I open up my fill port and let air escape?

 

[GTXJackBauer]

Ugh, over a week, shaking this case side to side and tapping the blocks, temps rise 5 C overall. should I open up my fill port and let air escape?

You could and plug it back in.

Also, take a few nice shots of the whole rig with a few close ups so we can see how it looks now.

 

[Eugenius88]

Some furious tapping tubing and temps went down 7 C lol, my loop is so odd when it comes to the air bubbles. One thing is, I still get constant air bubbles, and one card still exponentially heats up. I'll set up shots Sunday preferably and open the fill port, I'm booked these two days and I just wanted to do a quick update.

 

[GTXJackBauer]

Some furious tapping tubing and temps went down 7 C lol, my loop is so odd when it comes to the air bubbles. One thing is, I still get constant air bubbles, and one card still exponentially heats up. I'll set up shots Sunday preferably and open the fill port, I'm booked these two days and I just wanted to do a quick update.

Glad to hear your temps went down. Is your pump at 100% speed? How does the reservoir look? Are bubbles being sucked into the pump? How does the pump sound?

Make sure you don't leave the port open when the pump is on.

Take some shots of the rig so we can see the final product and give us shots of benchmarks or some info on it. Prime95, Valley, ambient, delta, CPU temp, GPU temp on idle and load. That will tell us if your system is running up to par.

 

[Eugenius88]

Glad to hear your temps went down. Is your pump at 100% speed? How does the reservoir look? Are bubbles being sucked into the pump? How does the pump sound?

Make sure you don't leave the port open when the pump is on.

Take some shots of the rig so we can see the final product and give us shots of benchmarks or some info on it. Prime95, Valley, ambient, delta, CPU temp, GPU temp on idle and load. That will tell us if your system is running up to par.

Pump's on 100%.
Micro bubbles are getting sucked in
Pump sounds good
I left the pump on with fill port open when I first primed the loop way back when, what happens then?
Prime 95 Temps

Idle temps

Valley temps after 17 seconds, notice GPU 2 and 3(The GPUs are read differently from Valley to EVGA Precision)
In this case GPU 2 and 3(Card that overheats are later referred to as 3 and 4)

GPU 1

GPU 2

GPU 3

GPU 4


Ambient as stated previously 18 C, and delta at one point on one GPU and CPU was 5 C and generally is around that range on idle.

 

[GTXJackBauer]

Pump's on 100%.
Micro bubbles are getting sucked in
Pump sounds good
I left the pump on with fill port open when I first primed the loop way back when, what happens then?
Prime 95 Temps
View attachment 136599
Idle temps
View attachment 136600
Valley temps after 17 seconds, notice GPU 2 and 3(The GPUs are read differently from Valley to EVGA Precision)
In this case GPU 2 and 3(Card that overheats are later referred to as 3 and 4)
View attachment 136601
GPU 1
View attachment 136602
GPU 2
View attachment 136603
GPU 3
View attachment 136604
GPU 4
View attachment 136605

Ambient as stated previously 18 C, and delta at one point on one GPU and CPU was 5 C and generally is around that range on idle.

Overall it's looking good. Your idle and load temps on the CPU @ stock and OC'd looks great.

As for your GPU temps on load, I would honestly redo the block on that card and reapply the TIM. There is seriously something going on there. My guess is the contact is bad or inside the block. I know you don't want to drain, prime and bleed the loop all over again but you have to when it comes to this type of cooling.

Btw, what pumps are you running again? Update your signature with the watercooling components.

 

[RnRollie]

At "the other forum" there is a guy having problems with his 690 : one core going waaaaay high (100 °C). It looks as if the thermal pads are the reason: they are preventing good contact on the GPU die. He switched to "softer" pads on the "hotter side" and got better results. He said he was going to order thinner (& better) thermal pads (0.5mm) and try those. If that solves the problem the cause is either something on the card itself, like a chip or vrm sitting higher or an imperfection or warp/bend in the block.

Now, the suggestion of swapping the cards/blocks is a good one, if the problem "follows" the block, it's the block

As for Rad dance with the whole case: courage It helps if you screw a barb in the highest point of the loop -that being a rad/bleed plug or the res/fillport- and attach a lenght (1-2 feet) of tube to it. See if you can attach the top end of the tube somewhere above the whole contraption - a piece of rope or elastic will do fine. Top up till you have a little bit of water in that tube - about level with the barb.
Since you have to get rid of bubbles with the pump running (avg to max speed) this allows more freedom of movement & far greater tilting angles and the air can get out through this tube.
Once you think you got rid of most air (the waterlevel should have dropped if you had a few "big" bubbles), switch off and prepare to unscrew this additional barb/tube while taking care to not have any spills -lots of paper towels

I sometimes used this approach on reservoir-less builds : using TWO T-lines: one at the bottom, before the pump, and one at the highest point in the loop
It might be a bit unsightly all this surpluss tubing, but it DID vent the air fast, if not faster as when using a reservoir. And topping up or draining went really well also.

 

[JeremyCT]

Wait, the 690 blocks use pads instead of paste between the block and the GPU? Really? I thought pads were only used at the VRM MOSFET and the RAM chip interface points.

 

[Eugenius88]

Well I wouldn't think that's the problem since I did it with the other card, honestly I think I just want to grab another evga classified 590, the dell one is ridiculous technology wise. It overuses clock speed compared to the EVGA one, during idle mainly. But I don't know. I feel like the card itself may be the issue.

Jack, I'm on break and just finished black belt testing, got me my 1st degree yay! So that means I have time and I will do it. It's a D5 Vario.

Rollie, I'll flush the block afterwards and make sure contact is good enough, if it happens again, bye bye Dell 590