One of 2 SLI card is too hot

[Zveroid]

Hi there!

So I've finished my watercooling loop recently. It looks like this (in order of flow):

1. D5 Pump ALPHACOOL VPP655 PWM
2. 2 videocards GTX 1080Ti Asus ROG Strix (in Barrow BS-ASS1080T-PA waterblocks)
3. 240mm Radiator (45mm width)
4. CPU waterblock (EKWB Asus ROG Strix monoblock)
5. 360mm Radiator (45mm width)
6. 450ml Barrow Reservoir

G1/4 fittings are used

The problem is too hot upper videocard: it warming up to 85C.
I already tried to remount waterblock (it wasn't necessary in the end - thermal pad was ok). Also I've swapped videocards and anyway, upper videocard is always hotter than whole system (approx 55C under load).
I've checked waterblock and didn't find any air there.

I've used Unigine: Superposition benchmark with enabled SLI profile, so both cards utilization is about 100%.

I wonder if someone can give some clue how to handle this.

Thanks.

P.S. Here's photo of the loop:
Here's photo:

 

[mackerel]

Might be a pain with hardline in place, but I'd try running the GPU blocks in series as opposed to the parallel it looks like it is currently. When running parallel, you'd hope it'll balance out but there is a risk of more flow to one block than the other. In series you guarantee the flow through both will be same at the cost of more loop resistance to flow.

 

[Zveroid]

Thanks, mackerel!

I already thought about this. Now they're connected as recommended in Barrow's manual. Now I should think how to connect OUT of 1st water block and IN of 2nd water block cause seems they're designed to have exact ports for flow input (left) and output (right).
Not sure if I'll have time for this in nearest two weeks, but I'll tell about results once I'll do it.

 

[maxfly]

Mackerals suggestion is a good one. also be sure to check the bios and set the pump rpms to 100% or max. as a side note keep a close eye on that pump. i know it isnt a vpp755(extremely high failure rate when it released) but i just dont trust alphacools d5 pumps anymore.
going by the pic you posted, try to tip your whole case to the right and give it a good shake while the pump is running and see if you can get any big bubbles out of it. if you cant see them youll be able to hear them as they escape and run out into the pump/res.

 

[EarthDog]

That pump needs to be set to max?? Wow. I have't run any pump I owned at max ever. Diminishing returns and all over 1-1.5GPM it just makes more noise and likely shortens pump life.

 

[Zveroid]

Mackerals suggestion is a good one. also be sure to check the bios and set the pump rpms to 100% or max. as a side note keep a close eye on that pump. i know it isnt a vpp755(extremely high failure rate when it released) but i just dont trust alphacools d5 pumps anymore.

I've already set pump rpms to 100%. Also I have plans to change my pump to some another one. For example some of these EKWB products, what do you think?
EK-XTOP Revo Dual D5 PWM Serial
EK-XTOP Revo D5 RGB PWM - Plexi

going by the pic you posted, try to tip your whole case to the right and give it a good shake while the pump is running and see if you can get any big bubbles out of it. if you cant see them youll be able to hear them as they escape and run out into the pump/res.

Photo is made from the left side of case. So do you mean tip it to the front (on the right side of the pic) or to the right side of case?
Actually, I did shake it several times in different ways right after I filled loop with coolant.

 

[maxfly]

That pump needs to be set to max??

for testing the barrow blocks and connections. i dont have any experience with those blocks. specifically how restrictive that set up may be. sadly it didnt help.

ek has solid pumps and tops. either of those are good choices. im not saying that your 655 wont last a good longtime its likely a standard laing d5 rebadged for alphacool. the 755s were the d5s that had issues primarily.
yes from the right side of the pic. tipping towards the front of the case. its worth a try, if that doesnt help im thinking that Mackerals suggestion is probably the last best option in troubleshooting. it seems like there is poor flow going thru the first card for some reason. or theres an inordinate amount of heat being generated by it as the primary card. being that there arent many people that have slied 1080tis we dont have much to go off.

 

[GTXJackBauer]

If you're still looking to replace the pump which I don't think you need to atm, make sure you go with EK's G2 PWM pump. It's their newer Generation 2 properly PWM spec'd pumps.

I also use a parallel GPU configuration with the opposite flow and connections with no issue. Not sure what the main issue can be here.

Give us a run down of all your temps while benching, air, water temps along with flow and pump speed if possible as well.

 

[Zveroid]

If you're still looking to replace the pump which I don't think you need to atm, make sure you go with EK's G2 PWM pump. It's their newer Generation 2 properly PWM spec'd pumps.

Noted. I'm not sure when I going to replace my pump, because I have another things to try yet.

I also use a parallel GPU configuration with the opposite flow and connections with no issue. Not sure what the main issue can be here.

Probably it's trouble of exact waterblocks. Barrow is chinese WB. They're descent but they're still chinese =)

Give us a run down of all your temps while benching, air, water temps along with flow and pump speed if possible as well.

Here's my temperatures during play my mostly hot (really) game - WoW on max. As you can see, most of temperatures are around 60 degrees but single videocard is about 80.


Unfortunately I don't have water temperature sensor. Air in my room is about 35 degrees.

I still didn't have a time to replace parallel connection to serial - will have it this week. If it will help, I'll think about remaining parts cooldown.

 

[GTXJackBauer]

Wow isn't so demanding even at full graphics. Something isn't right with those temps and can't put my finger on it atm. Keep us up to date on any future changes.

As for barrow, I think it's a knockoff of Bitspower from a chinese company. With that said, I'll always be loyal to Bitspower for their quality over the many years over barrow any time of the day, no matter the price.

 

[Zveroid]

Wow isn't so demanding even at full graphics.

Actually, wow has 2 graphics tabs. And with really max options, it warming system more than Unigine Superposition benchmark.

As for barrow, I think it's a knockoff of Bitspower from a chinese company. With that said, I'll always be loyal to Bitspower for their quality over the many years over barrow any time of the day, no matter the price.

I guess, I wouldn't take any chinese stuff if I was have any experience with water cooling before. But this is my almost first assembly. Most probably I'll replace barrow with something else, when next generation of videocards will be released.

 

[freeagent]

The top card will always run warmer with sli. The heat from the bottom card rises up. You could try adding a fan.. have the tips of the blades a few mm higher than the top card so you scrape the heat from the pcb, and the rest of the fan should direct the warm air to the back. It doesn't look like there is too much air flow in your case. After looking at your pic again, I don't see much room for an aux. fan. Also, I see your gpu temps are under 70c loaded? You have nothing to worry about. If I run 2x 580s in my case, the top card hits 100 no problem lol. But she doesn't crash so it must be ok haha.

 

[GTXJackBauer]

20c difference between 2 GPUs in SLI no matter the config is, is just too high under custom water imo or at least it defeats the purpose of spending such added funds. You'll usually get a 10c difference on average between two GPUs in serial. Parallel brings them a bit closer to each other.

Obviously something isn't right. Free could be right and it could be the lack of air flow. If you're swaping both GPUs and you're getting the same temps on the top PCIe, it's obviously not the waterblocks.

How many video cables do you have connected to the top GPU? Try connecting them to the 2nd GPU and have you tested this with your side case open? Try that and come back to us with the results but make sure you're running everything on stock settings.

 

[freeagent]

I think 20c is about what I saw with my two gtx 580s. It was disgusting. Just running heaven and the top gpu was at 100c after about 10 minutes. I didn't care about the heat from the gpus too much, but I wasn't used to feeling warm air coming from the psu. My thingy said 620w was being drawn. Those Fermi cards can really pound back the jiggawatts. Had I been running my X58 with those two cards that psu would have been at its limit.


Anyways.. I tend to ramble.


Quiet systems usually don't do too well overclocked, as stock can be a brute depending on what you run. You definitely want to get some air moving through your case. Looks like you have room in the back for a 140mm fan? Id go with a high speed high cfm pwm fan for sure Its a decent place to begin. You may see some results by removing your pci covers too.

 

[GTXJackBauer]

100c on Fermi's, even though their cap was somewhere in the 110c-120c, was a bit high but my 480s in SLI were screaming at a 'comfy' 85c-90c. The environment around me changed fairly quickly and I needed to keep my self hydrated. The whole 747 jet noise was another thing. My PSU was pumping around 800-850w. I would even hit 900w on a 950w PSU while folding or benching. It was brutal. I kinda miss it for just a second till reality smacks me upside the head of what I went through. lol

I just realized the rear fan is set as intake. What are your fans configured in the whole system? It should be front/bottom intake and top/rear exhaust or at least have the rear as exhaust. It might be getting too warm in there.

 

[Zveroid]

How many video cables do you have connected to the top GPU? Try connecting them to the 2nd GPU and have you tested this with your side case open? Try that and come back to us with the results but make sure you're running everything on stock settings.

I have 2 cables in my top GPU. I will try to connect them to bottom one. But Unigine shows 100% utilization of both GPU's, so I'm skeptical about this.

100c on Fermi's, even though their cap was somewhere in the 110c-120c, was a bit high but my 480s in SLI were screaming at a 'comfy' 85c-90c. The environment around me changed fairly quickly and I needed to keep my self hydrated. The whole 747 jet noise was another thing. My PSU was pumping around 800-850w. I would even hit 900w on a 950w PSU while folding or benching. It was brutal. I kinda miss it for just a second till reality smacks me upside the head of what I went through. lol

I just realized the rear fan is set as intake. What are your fans configured in the whole system? It should be front/bottom intake and top/rear exhaust or at least have the rear as exhaust. It might be getting too warm in there.

My fans are: rear intake, front and top pull exhaust.
Actually, I had rear and top exhaust and front intake before, but air inside the case was pretty hot, because hot air from fronpulled inside it. Now temperature in case close to temperature in the room.
I guess, my bigger problem is case construction. My PC case is Thermaltake C23:
View attachment 199711

My front panel has been cut manually and now have some space for air flow, but my top panel wasn't modded yet though. Tomorrow I'm planning to cut hole for fans on top panel and check if it will be better.

So tomorrow I have free weekend and I will try:
1. Move cable from top GPU to bottom
2. Cut hole for fans on the top panel of the case.
3. Change fans direction (front intake, top and rear exhaust)
4. Finally, change GPU connection from parallel to serial.
Also I already bought water thermal sensor, so I'll can to measure water temp during those experiments

 

[GTXJackBauer]

Now that I'm looking at the case and the air flow configuration, I believe it to be a air flow issue. I feel that is a horrible case design (unless TT copy's someone else's lol) for air flow imo and have seen similar cases meet similar fate. Maybe for a single CPU + GPU loop at most you can get away with that kind of design but when you're playing with another GPU and more heat surface (radiator), it will need to get a hold of that fresh air, in other words you need a case without those silly small side pocket panels that only give it a slit of fresh air from its side. At the moment, you're only getting fresh and dusty air from the rear single fan. Over time you will cake your stuff with enough dust, it could trap that heat. You'll want to create positive pressure as atm, you got negative going on.

Are there filters in the bottom, front and top compartment and if so, were they removed when you switched to exhaust?

At this point, I don't think you should even switch your GPU configuration tbh. I completely think its a air flow issue. Again, try with the side case off and see what you get after some tests, etc.

 

[Zveroid]

Filters weren't removed. It was like on the photo first. During PC rebuilding I switched fans to exhaust and made cutout on the front panel (about 30cm height). I also tried to remove these covers at all (so fans were totally opened) - temperature difference between GPU still was about 20-25 degrees. But I guess, I should do it again to make sure.

 

[Zveroid]

Ok, guys, I found the problem. First of all, thanks to GTXJackBauer for advice to move cable to another videocard - it was starting point to check problem correctly.
I moved cable to bottom GPU, but it wasn't work while cards was in SLI mode. So, I've disabled SLI in NVIDIA control panel and tried to start benchmark with cable connected to different GPUs. Top card was obviously hotter than bottom.
After several tests with opened caps and serially connected GPUs (it didn't help BTW), I decided to try swap GPUs in the loop again, and find out that I've confused, when tried to do this first time: bottom GPU, not top, was hotter after swapping. I just stupidly done test wrong first time and made wrong conclusion that always top card is hotter.
So I've re-installed waterblock on problem GPU, and temperatures difference is about 1-2 degrees now (GPUs connected in parallel). I'm not sure where exactly problem was, but assume that it was because of my decision to use standard NVIDIA thermal pad, that was used for stock air cooler. I have replaced it with thermal pads from waterblock package.
Also I've made hole in the top panel, and now my GPU temperatures after hour of continuous Unigine Heaven benchmark are about 55-60 degrees (air temperature is about 30 degrees).

I understand that it's still too hot for water cooling, but it's ok for me because they're not even close to temperatures I had with air cooling before. I guess, I will replace my PC case during next upgrade (it will be after 1180Ti release, I guess) to something with possibility to install 2 fat 360 radiators and good airflow, but I finished for now.

So thanks everyone for your participation, you're all really helped me

 

[GTXJackBauer]

Glad you got the issue's resolved. Sorry for the confusion as we're only good at with the info given to us. I thought maybe it was a bad seat on the GPU via waterblock but once you said that no matter the GPU at the top always ran hot, etc. Since that's thrown out the window and has been fixed, yes a better case for more rads will surely lower your temps. Maybe by another 15c-20c on the GPUs.