Stalled DICE CASES - FLOW : expandable SFF case for watercooling enthusiast (built in distroplate)

Magz

Chassis Packer
Jun 29, 2019
19
14
Did a bit more testing with fans turned off completely during a cinebench run. And it does become very apparent that the average CPU temp scales relatively linearly as the water temperatures rise, with a relatively stable delta of around 30-35C between water temp and CPU die temp. If 40 is considered within normal parameters, then it sounds like I'm well within the expected zone. I'll try to tweak the fan curves a bit to try and keep under that 45C ceiling where components will start to be in the danger of throttling. The flow rate doesn't seem to make a noticeable difference so I'm keeping the pump speed constant for now.

EDIT: Here it is "stabilized" aiming for 40C max while running Cinebench.

Looks like I will need to boost those fans speeds higher once I connect the GPU to the same loop... which is kind of a bummer because with these settings it's just on the threshold of being inaudible.😕

Was curious as to what temperatures people are getting with their builds (and how they were running the various components).

I am running a 3900x on the Asrock X570 Mini ITX with an EVGA RTX 2080 Super (using the Hydro Copper GPU block).

I was running the fans (Noctua NF A12) @1000 rpm and pump (Alphacool Eispumpe VPP755) @3500rpm and then decided to run the Time Spy stress test and ended up getting 78C as max temperature (idle temp for GPU was 27C). Having never done this before, I assumed something was wrong with my installation of the GPU block so I decided to drain the loop and re-seat the card. In hindsight, I was clearly expecting too much - I wanted the system to be quiet and cool.

Anyway ended up re-running the test and same temps! So I decided to run the fans and pump at full speed and then end up with a GPU temp of 66C. At this stage I am assuming my installation of the various blocks is correct and I need to tweak the fan and pump speed to find the sweet spot of silence and performance. Thoughts folks?
 

xunilon

Chassis Packer
Dec 5, 2019
20
23
Looks fantastic!

EDIT: How are the GPU thermals behind glass? Are those LED lit fans?
Thanks. The thermals are decent! It stays at around 70C in Cyberpunk with a custom fan curve. Better than in an H210 with SFX PSU and vented PSU bracket plus 2x 40mm fans in taking air from the bottom PSU dust filter. And quieter.
 

TehMoonRulz

Cable Smoosher
Dec 16, 2020
10
2
Thanks. The thermals are decent! It stays at around 70C in Cyberpunk with a custom fan curve. Better than in an H210 with SFX PSU and vented PSU bracket plus 2x 40mm fans in taking air from the bottom PSU dust filter. And quieter.
I wonder if putting a small spacer on the glass and case would give it enough extra air to decrease temps?
 
  • Like
Reactions: xunilon

ThisIsForBruce

Trash Compacter
Mar 25, 2020
36
34
As for my temps, they're actually quite good. Better than i would`ve thought honestly. Although my setup is probably much different then most.
I am most likely cooling more than the average using this case also.

Playing Call of Duty for ours gets my water temps up to 39C max with the fan curve sitting at 48%.
The Pump/fan curves i use are custom ones with multiple different variables since everything runs through my Aquero.
I`m running a HWlabs 280 on the bottom and a 3x40mm Alphacool rad in the GPU side of the case.
My 3950x is an amazing core overclocker as im running 4.4/4.3 (CCD1/CCD2) and temps never really go over 80 under load @ 1.3vcore (around 1.210 SVI2 Vin)
GPU is a 2080ti and it literally never hits 50C. Mid 40s under load. OC is 2100/1950 w/ +.005 v.

I`m curious as to what everyone elses temps are running at
 
Last edited:
  • Like
Reactions: kosta

kosta

SFF Lingo Aficionado
Oct 27, 2019
119
70
I opened up my Optimus block and cleaned out the cat hairs that were stuck inside 🤦‍♂️ Also replaced the thermal paste with liquid metal. It made a massive difference, the average temps under load fell by almost 10 degrees. The 5900x doesn't exceed 70 under load any more. I've also bumped up the pump speed a bit and now the CPU sits at around 65-70 during a Cinebench load test. Fans never exceed 900rpm and the water temp stabilizes at around 12C over ambient air. This is with 2x240 HWLabs x-flow rads and no GPU in the loop.

I also tried undervolting the 5900x and I'm kind of bummed about the whole experience. Using Ryzen Master, I was able to run all-core stable at 4.5Ghz and 1.25V which was both cooler and faster than using auto-OC/PBO. But I found no way to undervolt/overclock cores without killing the single core boost capabilities. So eventually I just set everything back to auto. 😕

In terms of the flow rate - I'm getting about 0.5 G/h (105 l/h) with the pump at 55% (2200RPM).

I installed a temperature sensor for air so that I can calculate the delta between air and water and use that as input for controlling the fans, but I quickly realized that the Aquacomputer QUADRO doesn't support virtual sensors in hardware so this method only works while the software is running. I think I'll probably revert to using the water temps alone again as I like to have all the logic running on device.

Here is the "idle" situation:


After 30 minutes of Cinebench:


This is how I configured the sensor:


I tried to smooth out the small changes in air temps using the low pass and average filter but it feels like a bit of a hack. If someone knows a better technique for simulating a hysteresis for air temps, I'd love to hear it!
 
Last edited:
  • Like
Reactions: Fitchew

ThisIsForBruce

Trash Compacter
Mar 25, 2020
36
34
I opened up my Optimus block and cleaned out the cat hairs that were stuck inside 🤦‍♂️ Also replaced the thermal paste with liquid metal. It made a massive difference, the average temps under load fell by almost 10 degrees. The 5900x doesn't exceed 70 under load any more. I've also bumped up the pump speed a bit and now the CPU sits at around 65-70 during a Cinebench load test. Fans never exceed 900rpm and the water temp stabilizes at around 12C over ambient air. This is with 2x240 HWLabs x-flow rads and no GPU in the loop.

I also tried undervolting the 5900x and I'm kind of bummed about the whole experience. Using Ryzen Master, I was able to run all-core stable at 4.5Ghz and 1.25V which was both cooler and faster than using auto-OC/PBO. But I found no way to undervolt/overclock cores without killing the single core boost capabilities. So eventually I just set everything back to auto. 😕

In terms of the flow rate - I'm getting about 0.5 G/h (105 l/h) with the pump at 55% (2200RPM).

I installed a temperature sensor for air so that I can calculate the delta between air and water and use that as input for controlling the fans, but I quickly realized that the Aquacomputer QUADRO doesn't support virtual sensors in hardware so this method only works while the software is running. I think I'll probably revert to using the water temps alone again as I like to have all the logic running on device.

Here is the "idle" situation:


After 30 minutes of Cinebench:


This is how I configured the sensor:


I tried to smooth out the small changes in air temps using the low pass and average filter but it feels like a bit of a hack. If someone knows a better technique for simulating a hysteresis for air temps, I'd love to hear it!
Take that custom Air/Water delta sensor reading and use it as the data source for one of the Software Temp Sensor outputs under the Sensor tab.
Then you can use that Software Temp Sensor as the data source for your controller/fan curve. You can then adjust the severity of the curve using the offset slider.
I have an air temp sensor directly behind the exhaust of my 280 radiator on the bottom of the case, so it should be the hottest exhaust in the system. It does not fluctuate fast, usually is moving as fast as the water temp is.
Make sure you`re using a proper 10k temp sensor or else its readings will be off as i don't think there is a way to change the resistance reading it uses. Same for flow sensor, but that reference reading can be changed so if you want a truly accurate reading it needs to match the sensor correctly. I personally don't care, as i only use the number as a reference for the Aquaero`s controller. Up to you 🤷‍♂️
Hopefully I understood what you`re trying to do with that reading... lol
 
  • Like
Reactions: kosta

kosta

SFF Lingo Aficionado
Oct 27, 2019
119
70
Take that custom Air/Water delta sensor reading and use it as the data source for one of the Software Temp Sensor outputs under the Sensor tab.
Then you can use that Software Temp Sensor as the data source for your controller/fan curve. You can then adjust the severity of the curve using the offset slider.
I have an air temp sensor directly behind the exhaust of my 280 radiator on the bottom of the case, so it should be the hottest exhaust in the system. It does not fluctuate fast, usually is moving as fast as the water temp is.
Make sure you`re using a proper 10k temp sensor or else its readings will be off as i don't think there is a way to change the resistance reading it uses. Same for flow sensor, but that reference reading can be changed so if you want a truly accurate reading it needs to match the sensor correctly. I personally don't care, as i only use the number as a reference for the Aquaero`s controller. Up to you 🤷‍♂️
Hopefully I understood what you`re trying to do with that reading... lol
Thanks! That's what I did, but it seems like the soft sensor curves only work when Aquasuite is running. They are not part of the profile that gets uploaded directly onto the hardware (QUADRO). Regardless, it works pretty well, there's a dozen ways to control the ways, having a bit of analysis paralysis. :)
 

ermac318

King of Cable Management
Mar 10, 2019
655
510
Thanks! That's what I did, but it seems like the soft sensor curves only work when Aquasuite is running. They are not part of the profile that gets uploaded directly onto the hardware (QUADRO). Regardless, it works pretty well, there's a dozen ways to control the ways, having a bit of analysis paralysis. :)
That is correct, Virtual Sensors/Software Sensors only work when Aquasuite is running with the Quadro. Mine is on the way and I plan to just use the water coolant temp (as my ambient is fairly stable) with either a Temp Target or Curve setting.
 

kosta

SFF Lingo Aficionado
Oct 27, 2019
119
70
That is correct, Virtual Sensors/Software Sensors only work when Aquasuite is running with the Quadro. Mine is on the way and I plan to just use the water coolant temp (as my ambient is fairly stable) with either a Temp Target or Curve setting.
One note on the temperature target mode that simulates PID behavior. I was somewhat disappointed with it because it only starts operating once the temperature target is breached. Until that point, the fans just spin at minimum speed you set.

What that means is that it will always overshoot the target quite badly and then have to work really hard to bring it back to desired state by ramping up the fans very aggressively. As a comparison, the PID I have in my coffee machine will detect if it's going to overshoot the target and compensate for that. I was hoping that Aquasuite will also start slowly ramping up the fans before it hits the target temp to try and control the rate of increase but I did not observe such behavior.

If you find a way to make it behave that way, please do share how you configured it! I didn't go into the more advanced manual settings.
 

Magz

Chassis Packer
Jun 29, 2019
19
14
So I have done some rudimentary testing using an OXO meat thermometer sitting in the distro (as I don't have a water temp sensor). Each benchmark was run around 3 times back to back and temps are from the final run (Cinebench was run using the multicore bench). Took a 30 min break between each test to make sure the water temps were back to what it was when idle. Also just to highlight, am using the glass side panels. All temps in Celsius and rounded up / down to the nearest degree.

So basically pump speed makes no difference from what I can tell (or in some cases just increased the Delta T - but suppose that's within the margin of error). Running the fans at full speed obviously yields the best results but that isn't an option as I don't want a jet engine next to my ear. So somewhere in the middle is just fine I think. I should really do the tests by setting the fans to 1200 and 1400 rpm to see if that works out any different. But currently keeping pump and fan speeds low as I prefer it to be quiet and I can live with a 15 degree Delta T.

 
  • Like
Reactions: kosta

ermac318

King of Cable Management
Mar 10, 2019
655
510
i think there's likely to be changes based on feedback so probably but i think it will be a bit later
Here's what Dice has said on Discord:
"There will be a lot of change, first of all I want to get rid of sheet metal if I can, because of those tolerance issues. Most if the cases working fine but in some of them things just doesn't line up well. There will be no extra large expands, because the need for them is really small. The expands will be improved, without the need for any extra parts. There will be mesh side panels, larger distro, and a lot of other things. Just didn't want to share until I finish every detail. In the terms of logistics/profit the V1 was a disaster, I did t really think about it in the first time but to manage that many types of parts is really difficult so in the future I want to focus on the V2 to simplify things a bit."

I would imagine that v2 will be similar in concept but a vastly simpler case, focused on the most common use case for the Flow: watercooling.
 

ThisIsForBruce

Trash Compacter
Mar 25, 2020
36
34
Hope everybody's builds are going along alright.
Here is a few pictures i took of the finished build.
Running a 3950x @ 4.4/4.3 GHz. 1.206Vcore with a Asus 2080ti on an EK waterblock and backplate.
32GB Hynix CJR cheap no name with an ebay heatsink that i modified to fit dual rank memory. Runs 3333MHz CL14-17-17-28 @ 1.45v. 1:1 CR, obviously.
Runs whisper quite for normal use, only hits 50% fan curve at max water temp which is 38C.
 

kosta

SFF Lingo Aficionado
Oct 27, 2019
119
70
Hope everybody's builds are going along alright.
Here is a few pictures i took of the finished build.
Running a 3950x @ 4.4/4.3 GHz. 1.206Vcore with a Asus 2080ti on an EK waterblock and backplate.
32GB Hynix CJR cheap no name with an ebay heatsink that i modified to fit dual rank memory. Runs 3333MHz CL14-17-17-28 @ 1.45v. 1:1 CR, obviously.
Runs whisper quite for normal use, only hits 50% fan curve at max water temp which is 38C.
You have glass panels on both sides? What's the direction of your airflow?
 

TheBigG

Trash Compacter
Jun 3, 2019
46
14
did anybody got a response from diceboii since 20th december? He wrote me at that date that he will send out replacement parts now, but I didn't got any replacement parts or a reply since then
 
Last edited:

kosta

SFF Lingo Aficionado
Oct 27, 2019
119
70
I've asked a bunch of questions on the forum but got no replies. You could try discord but I haven't had any success there either.
 

TehMoonRulz

Cable Smoosher
Dec 16, 2020
10
2
I'm running into a strange issue with my d5 ()
I set up the whole loop and the pump wouldn't spin so I took everything apart and now just have a contained loop on the distro plate.
I cleaned the pump out (lots of black gunk...makes me think possible return?) and it would start and stop. If I loosened the d5 bracket it would start and stop and if I kept tightening it the pump wouldn't spin at all.
Also having issues bleeding air out of the pump area of the distro.

And advice would be appreciated.
 

TehMoonRulz

Cable Smoosher
Dec 16, 2020
10
2
I want to add that Ive taken the pump off and ran it in a mug with distilled water and with the impeller submerged it runs continuously just fine.
 

TehMoonRulz

Cable Smoosher
Dec 16, 2020
10
2
Following up on this as well. Ive had two pop off. One for the PSU brackets and another for the bottom case cover.
One of mine just fell off too and its between the distro plate and the frame rail so can't screw down one of the four screws that holds the top. How are they attached - anyone know?

Won't bother re-attaching mine as it would require the whole system to be dismantled.