Cooling capacity/performance

[Azrael2k8]

Hi,

 

I am worrying if my current rad/fan setup is sufficient for my GPU upgrade(2x 1080 Ti).

 

Here is what I currently have:

 

Case: BeQuiet! Dark base pro 900

Top Rad: EK 420 CE with 3x EK-Vardar F2-140 As exhaust Pull

Front Rad: EK 280 CE with 2x 140 BeQuiet SilentWings 3 as intake Push

Exhaust Fan Back: 1x BeQuiet Silent Wings 3

GPU: Palit 1070 super jetstream

CPU: i7 6700 no OC (water cooled)

GPU Overclock: Core: 2050 Mhz, Memory: 4120MHz under 99% load

 

With fans on max and all lids and doors of the case closed I get 51°C - 53°C with GPU and CPU in the loop while playing BF1 on Ultra 1440p with 80-90fps. I consider this pretty hot. Idle temp is ambient + 10°C.

 

I am about to purchase two msi 1080 ti SeaHawk EK X and not sure if my cooling loop/case is up for it? Will it overheat? Are the SilentWings3 too weak in terms of static pressure? The Vardar are on an acceptable noise level though I don't want to go louder.

 

I would be happy if you could share your opinion.

 

Kind regards

[W-L]

1 minute ago, Azrael2k8 said:

-SNIP-

You have more than enough rad space even if you go with a dual GPU setup. For those temps you have with the current setup is nothing to worry about at all. That case is slightly restrictive in airflow due to the side venting design so you may see a drop if you were to take off the top and front of the case to a certain degree. 

[Azrael2k8]

Just now, W-L said:

You have more than enough rad space even if you go with a dual GPU setup. For those temps you have with the current setup is nothing to worry about at all. That case is slightly restrictive in airflow due to the side venting design so you may see a drop if you were to take off the top and front of the case to a certain degree. 

Ok, good to hear. Would it make sense to switch the SilentWings3 to fans with higher static pressure to improve cooling performance though? Or would this only result in marginal improvments?

[W-L]

1 minute ago, Azrael2k8 said:

Ok, good to hear. Would it make sense to switch the SilentWings3 to fans with higher static pressure to improve cooling performance though? Or would this only result in marginal improvments?

The silent wings 3 are a good pressure fan for using on rads, a higher rpm and static pressure fan can help but also at the downside of noise. You can try and see if the case is making the majority of the impact in your temps by removing the top and front panel, it should help say at around 5C but if it's more in your setup then you know that it is the major limiting factor. 

[Azrael2k8]

Just now, W-L said:

The silent wings 3 are a good pressure fan for using on rads, a higher rpm and static pressure fan can help but also at the downside of noise. You can try and see if the case is making the majority of the impact in your temps by removing the top and front panel, it should help say at around 5C but if it's more in your setup then you know that it is the major limiting factor. 

Thanks for your input. I will give it a try and see how much the case impacts on the cooling performance. I really like the case though since it keeps the noise down and the dust out (at least more than other cases). But that means worse air flow I guess. I will test the temps tomorrow with dust filters removed etc. but as long as I don't come close to 70°C I am happy. I think the 1080 TI are also generating less heat in relation to performance, which might help.

[Azrael2k8]

So I just installed the two MSI 1080 Ti Sea Hawk EK X with an EK Serial DUAL 3 Slot Terminal.

 

However, one of them is getting hot really fast whilst the other has 14°C offset. I attached an Image of the Terminal connection to make sure I am using the correct in/outlets but I think that I did not screw up there. Does anyone have another idea as to why one card would get much hotter even though the hotter one has less load?

 

Edit: I just checked since I saw a small air bubble in one of the GPU Blocks. After setting the pump to 100% and tilting the case there are only tiny air bubbles left. I also want to mention that the GPU that is overheating is rising and falling very rapidly in temperature, could this mean that the connection between gpu and waterblock is bad (i.e. thermal paste) ?

 

[W-L]

43 minutes ago, Azrael2k8 said:

So I just installed the two MSI 1080 Ti Sea Hawk EK X with an EK Serial DUAL 3 Slot Terminal.

 

However, one of them is getting hot really fast whilst the other has 14°C offset. I attached an Image of the Terminal connection to make sure I am using the correct in/outlets but I think that I did not screw up there. Does anyone have another idea as to why one card would get much hotter even though the hotter one has less load?

 

Edit: I just checked since I saw a small air bubble in one of the GPU Blocks. After setting the pump to 100% and tilting the case there are only tiny air bubbles left. I also want to mention that the GPU that is overheating is rising and falling very rapidly in temperature, could this mean that the connection between gpu and waterblock is bad (i.e. thermal paste) ?

 

If there is an air bubble make sure it gets out of the block that can have a fairly large impact, as for the connections since it's a dual triple a lot block that looks to be correct for the inlets and outlets, @EK Luc should be able to confirm that as I'm not 100%.

https://www.ekwb.com/shop/EK-IM/EK-IM-3831109868072.pdf

 

[Azrael2k8]

1 hour ago, W-L said:

If there is an air bubble make sure it gets out of the block that can have a fairly large impact, as for the connections since it's a dual triple a lot block that looks to be correct for the inlets and outlets, @EK Luc should be able to confirm that as I'm not 100%.

https://www.ekwb.com/shop/EK-IM/EK-IM-3831109868072.pdf

 

Thank you again for your reply. In the meantime I checked again. There are still some small (about pinhead size) air bubbles in the GPU at the top (next to cpu), which is also the one that is getting hotter. I will try to remove both cards and see if I can get the air out of there when they are not plugged into the mainboard. I really dont want to drain my system again...

[Azrael2k8]

25 minutes ago, Azrael2k8 said:

Thank you again for your reply. In the meantime I checked again. There are still some small (about pinhead size) air bubbles in the GPU at the top (next to cpu), which is also the one that is getting hotter. I will try to remove both cards and see if I can get the air out of there when they are not plugged into the mainboard. I really dont want to drain my system again...

Not really sure if I should reply to my own post but I will just do so.

 

I tilted the case, removed the gpus without draining the system and tried to remove the air bubbles. I did get all major ones out of there. But again the small ones remain. I think I will have to troubleshoot this by draining the system. Btw I could watch the waterflow a bit better now, it looks like enough water is travelling through both waterblocks.

 

What would be the best way to troubleshoot this? I think I will be able to do so at sunday. I was thinking of the following options:

 

1. removing the serial terminal and remounting the old connector to the overheating card and check if the issue persists
2. Change Graphic card postions in case water flow is the issue
3. Directly reseat the waterblock of the overheating card and reapply thermal paste

[For Science!]

25 minutes ago, Azrael2k8 said:

-snip-

Hello!

 

Would it be possible to be a side-on profile of your whole system, and another picture side-on but closer up to the GPU?

 

Also, if you have any pictures of the PCB during the installation of the blocks, that could be helpful too.

 

[Azrael2k8]

12 hours ago, For Science! said:

Hello!

 

Would it be possible to be a side-on profile of your whole system, and another picture side-on but closer up to the GPU?

 

Also, if you have any pictures of the PCB during the installation of the blocks, that could be helpful too.

 

Hi,

 

thank you for your interest. I attached two pictures one with the full system and a close-up of the GPUs. I sadly do not have pictures of the PCB whilst installing, but could make some if I drain my system tomorrow. The Waterblocks were obviously pre-installed by MSI/EK.

 

The Res used to be full but the tilting yesterday moved some air bubbles into it.

 

The Water flow currently is like this:

1. Pump -> front rad
2. front rad -> lower GPU
3. lower GPU -> upper GPU
4. upper GPU - > CPU
5. CPU -> top rad
6. top rad -> res
7. res -> pump

 

Do you have any insights as how I should go about trobuleshooting if I drain my system in the next days?

 

[For Science!]

2 hours ago, Azrael2k8 said:

-snip-

Hi unfortunately, I couldn't find anything obvious about your GPU section that would have caused the temp. discrepancy. I did however notice that according to your flow direction, your CPU block input/outputs are reversed - although that should not be the cause of your observations.

 

I would expect the temperatures to rise equally, assuming both your cards are at the same clock frequency and SLI is enabled in the drivers. Maybe you can just make sure by displaying the clock frequency, memory frequency and utilization on your on screen display.

[Azrael2k8]

1 hour ago, For Science! said:

Hi unfortunately, I couldn't find anything obvious about your GPU section that would have caused the temp. discrepancy. I did however notice that according to your flow direction, your CPU block input/outputs are reversed - although that should not be the cause of your observations.

 

I would expect the temperatures to rise equally, assuming both your cards are at the same clock frequency and SLI is enabled in the drivers. Maybe you can just make sure by displaying the clock frequency, memory frequency and utilization on your on screen display.

Hi,

 

thank you for inspecting it. I remember that I swtiched the CPU input, output the last time I rearranged my loop and turned the top rad around. I will try to fix that the next time I upgrade my cpu, which will be in the near future.

 

I updated Msi afterburner to show clock and memory frequency and attached the screenshot of heaven benchmark after 2 minutes in to the program, The difference in heat was again really rapidly. like 20 seconds into the benchmark and the gpu1 would be at least 10°C hotter. I also attached a screenshot of my nvidia sli setting. It is in german but the positioning is the same so "maximize 3d performance" is selected.

 

What would be the best to troubleshoot this? Change GPU position or only install the hotter gpu alone and see if it performs the same when solo?

 

Btw what could be the causes for this? I would think one of the following:

 

1. Faulty GPU/PCB getting hot for no reason
2. Bad connection between Waterblock and GPU
3. Waterflow not reaching the second GPU (unlikely)

Would you agree with that?

 

Kind regards, 

 

 

[For Science!]

1 hour ago, Azrael2k8 said:

-snip-

Option 2 sounds most likely to me, but you can make sure by swapping the position of the 2 GPUs. If you swap the cards around in your loop and find that the same card is hot, maybe it is poor contact between the block and the card.

 

However since your card is a pre-installed block maybe it is better to contact MSI and ask about what is the best option. Disassembling the block may void your warranty (no idea about this), and so you should ask the vendor. 

[Azrael2k8]

On 15.7.2017 at 5:18 PM, For Science! said:

Option 2 sounds most likely to me, but you can make sure by swapping the position of the 2 GPUs. If you swap the cards around in your loop and find that the same card is hot, maybe it is poor contact between the block and the card.

 

However since your card is a pre-installed block maybe it is better to contact MSI and ask about what is the best option. Disassembling the block may void your warranty (no idea about this), and so you should ask the vendor. 

I just swapped the cards and noticed I missed to install the orings between serial terminal and the blocks and then realized that the serial terminal only came with two orings. I will order some more but install the two that I have to the card that was having the heat issue. And since it is not leaking I don't think that is an issue now but nevertheless I will install the missing ones as soon as they arrive.

 

I performed one benchmark (benchmark 2.png) before the swap and the other after the swap (benchmark 3.png) and attached the results as pictures. I also relocated the CPU inlet/outlet connections so that it will now go GPU -> top rad -> CPU (correct inlet) -> res.

 

The temperature difference has been cut to half. The card that was cooler is now getting hotter as well. I think 5-7°C difference is acceptable however I do not understand why the other GPU is getting hotter now as well. The screenshots also contain the graphs over time. They are in german though.

 

Any idea what is causing behaviour like this?

[For Science!]

32 minutes ago, Azrael2k8 said:

-snip-

As disturbing as is the lack o-rings, as you correctly pointed out, as long as it is not leaking it probably is not affecting the flow. Your temperatures quite hot even considering it is a 1080Ti, These are my temps for a dual 1080 configuration; both cards (left and right) rise and fall equally and stay about 60 degrees under load (I have 240 + 360 mm radiator space). 

 

It really makes me think something dodgy is happening with the pre-installed waterblock, but I don't know

The only last thing would be that if your pump speed is not 100%, try that. I know that in my system at 90% pump speed for the D5 I barely scrape 3.8 L/min flowrate, so if you are operating at a much slower pump speed, the flow may be a bit slower than ideal.

 

[Azrael2k8]

56 minutes ago, For Science! said:

As disturbing as is the lack o-rings, as you correctly pointed out, as long as it is not leaking it probably is not affecting the flow. Your temperatures quite hot even considering it is a 1080Ti, These are my temps for a dual 1080 configuration; both cards (left and right) rise and fall equally and stay about 60 degrees under load (I have 240 + 360 mm radiator space). 

 

It really makes me think something dodgy is happening with the pre-installed waterblock, but I don't know

The only last thing would be that if your pump speed is not 100%, try that. I know that in my system at 90% pump speed for the D5 I barely scrape 3.8 L/min flowrate, so if you are operating at a much slower pump speed, the flow may be a bit slower than ideal.

 

Thanks for going through all this trouble for me. I just contacted the guy who sold me the 1080 ti and he tested the card without a terminal block but direct connection with tubes (ugly though) and only had a 2°C heat difference. I guess I will give that a try as well since it is really confusing me that changing the GPU position had that effect.

 

Btw I attached my bios settings for the water pump. I set it to 100% as soon as I had the issues with the new GPUs. My old 1070 was fine with low pump rates but two gpus might need that flow speed.

 

I will report back with the test results when I connected the two cards directly.

[Azrael2k8]

Spoiler

 

Well my Corsair 1000 just blew up. Depending on how much is damaged and when my replacement psu arrives i will need to postpone any testing for the gpus...

[For Science!]

16 minutes ago, Azrael2k8 said:

  Hide contents

-snip-

oh no! any particular reason for the death? any chance of liquid getting in from the top fan that was mounted upwards? I mount my PSU fan down for this potential.

[Azrael2k8]

1 hour ago, For Science! said:

oh no! any particular reason for the death? any chance of liquid getting in from the top fan that was mounted upwards? I mount my PSU fan down for this potential.

I covered it with plastic to prevent Water from dripping in there during my work at the tubes. Either it blew up due to an honest malfunction or the stupid serial terminal did leak overnight. I did not find water in the PSU but i guess Corsair will let me know why it failed.

 

In hindsight i really should have placed the PSU fan down...

[Azrael2k8]

So I just finished installing a ghetto Sharkoon 600W PSU for testing. I don't know if it will be able to power both cards but for now I will test both cards individually.

 

As for damages. Apparently everything survived except the fan controller and maybe the LEDs. This means that for now the front rad wont have any working fans but I guess thats fine. I am happy nothing else died (though I have not tested the second GPU but assumed if one is alive the other one should be too.

 

As for the original issue. I will benchmark both cards individually with the exact same setup and see if there is a temp difference. Results will follow. 

 

Edit: I am currently testing the "bad" GPU and it already is at 72°C after one run in heaven. I doubt that this is good but I will install the other one as well just to be sure. Btw MSI replied to me that removing the waterblock will not void warranty, so before I send it in I will give it a try to replace the thermal paste.

 

Edit 2: See results of the other "good" GPU basicly 10°C colder... That does not like an acceptable margin of error to me or is it?

 

Edit 3: According to MSI 10°C are acceptable differences and are caused to different power consumptions for each chip meaning they vary in voltage and therefore release different amounts of heat. He also assured me that changing thermal paste will have no influence on the guarantee since I am in germany.

[Azrael2k8]

Holy... I just set the bad GPU up with new thermal paste. It is running at 48-52°C in Heaven instead of 70-72°C. I mean wtf. Plus I attached two pictures of what the thermal paste was like before and after. I am not really sure if I did a good job but apparently it worked. What do you think about MSIs approach of TP application? Isn't that very little and also clumsy applied (see that spot on the left)?

 

Now I am only missing a good psu and 2 o-rings and I will reinstall that terminal and see if the temps changed there as well.

[For Science!]

Looks a bit dodgy what msi did, maybe the compound was old? please see attached what Asus did on my air cooler and what I did for my waterblock.

 

As long as it's not too little and you are using a non conductive paste your method is probably fine too.

 

before

 

 

after

[Azrael2k8]

17 minutes ago, For Science! said:

Looks a bit dodgy what msi did, maybe the compound was old? please see attached what Asus did on my air cooler and what I did for my waterblock.

 

As long as it's not too little and you are using a non conductive paste your method is probably fine too.

 

before

 

 

after

How did you get the pins next to the GPU (no idea what they are called) so clean? I am usually afraid to damage that so I leave them covered with the old paste and make sure to clean the actual GPU as much as possible. Is that wrong?

 

Btw I am able to push the temperature to 58°C on the bad CPU with constant benchmarking but that is far away from any real world gaming load i guess and still way better than before.

[For Science!]

1 minute ago, Azrael2k8 said:

How did you get the pins next to the GPU (no idea what they are called) so clean? I am usually afraid to damage that so I leave them covered with the old paste and make sure to clean the actual GPU as much as possible. Is that wrong?

 

Btw I am able to push the temperature to 58°C on the bad CPU with constant benchmarking but that is far away from any real world gaming load i guess and still way better than before.

I used a Q-tip (cotton ear bud) soaked in isopropanol and carefully scrubbed around them. I don't think its that critical to do it though. 

 

I think 58 degrees for a benchmarking is fine.