Concept Needle - small footprint, fully watercooled build at 7L
- Thread starter robbee
- Start date
Hey Morj, I've long doubted between an 80mm or a 92mm radiator. The full width of the stack of single slot gpu + motherboard with 40mm cooler + offsets and spine came at around 80mm though so it was just a perfect match.
The inner width of the case is 82mm btw, so I would have to add a full 12mm to it to fit a 92mm radiator (with some wiggling room). As an added bonus, the X-flow pattern and multiple fill/drain ports on the Alphacool 80mm radiator are super convenient!
I'm still dreaming of making a 92mm version with room for a ddc pump though, so it may just happen one day.
I'm confident that triple 80mm radiator will do the job for the RX 5700. It's almost the same surface area as a single 140mm. Optimum Tech ran some water cooling tests with a power guzzling i9-10900 and RX 3090 setup, and the results were surprising with a load temp of 66C on an open air bench. You should have more headroom with your GPU.
I think @robbee is using double 80mm, not triple. Alphacool triple 80mm rad is 286mm long. In this design the radiator is top-mount:
Yes it is a double 80mm radiator. I'm not 100% sure about its cooling capacity, but that's part of the adventure 
If it turns out really bad, i can still opt for a push pull config by making the case a bit higher. If that is still insufficient, it's back to the drawing boards. But I'm fairly confident that the radiator will be able to keep a combined 300-350 watts of heat generating components on an acceptable temperature.
If it turns out really bad, i can still opt for a push pull config by making the case a bit higher. If that is still insufficient, it's back to the drawing boards. But I'm fairly confident that the radiator will be able to keep a combined 300-350 watts of heat generating components on an acceptable temperature.
So I got a ryzen 5 3600 on the way and a 1080ti with an EKWB waterblock. Now comes the big question: will a double 80mm rad be able to dissipate all this heat if I undervolt it enough? Should I go for a 4 fan push-pull config? Or will it be a complete failure? I'll find out this weekend or next week!
Would be nice to see what can do a 4 fans push/pull configuration compared to one with 2 fans only.
As I never took the time to check that with my NCase build...
Would be nice to see what can do a 4 fans push/pull configuration compared to one with 2 fans only.
As I never took the time to check that with my NCase build...
I got 2 spare 80mm be quite fans so I'll definitely try it. They're not as high quality as the noctua ones, but they'll show the potential of extra fans
Cool!I got 2 spare 80mm be quite fans so I'll definitely try it. They're not as high quality as the noctua ones, but they'll show the potential of extra fans
I'm currently running an i7 8700K (so more TDP than R5 3600) and an undervolted (not underclocked) 1070Ti (so a bit less TDP than 1080Ti). Total system load under CPU+GPU stress is around 260W, with a dual 92mm rad seems to have some headroom.
So I think R5 3600 with an undervolted+underclocked 1080Ti is manageable with dual 80mm at least for everyday use without heavy gaming or full throttle synthetic tests.
So first results are in! This was rough to get up and running though. Getting the air out of the loop was very hard because of the weak pump. It's a good thing the radiator has so many fill ports.
Now for the numbers. Let's start with the fans and pump. I've made a profile that ramps their speed up depending on the average temperature of my CPU and GPU combined. At an average of 50C or less, the fans spin at 30% and pump at 50%. This ramps up untill at an average of 85C fans and pump spin at 100%.
As for the temperatures, things work and are fine, but not very pretty. I clocked my 3600 to 4.2GHz at 1.10v, the 1080ti was left at stock for now but i'll undervolt that later.
Running Cinebench, the CPU gets up to 76°C so that's fine so I didn't test much further. However, when stressing the GPU, things get a bit more toasty. I ran multiple Heaven benchmarks and these are the results:
So a couple of conclusions:
- Temperatures with 2 fans are a bit hotter than I'm comfortable with. I'll probably go with a 4 fan setup.
- CPU temperature ramps up quickly when the GPU is stressed. I'm pretty sure the CPU cooler is well attached and there's no more air in the pump. I think it's just the fluid getting too hot because the fans struggle to cool down the radiator.
- The noise is somewhat louder than I'd like and the pump noise has an annoying high pitch to it running over 60%. I still haven't made the shell part of the case so I hope that will make things a bit better.
- I'll undervolt the 1080ti next so the temperatures get lower.
Now for the numbers. Let's start with the fans and pump. I've made a profile that ramps their speed up depending on the average temperature of my CPU and GPU combined. At an average of 50C or less, the fans spin at 30% and pump at 50%. This ramps up untill at an average of 85C fans and pump spin at 100%.
As for the temperatures, things work and are fine, but not very pretty. I clocked my 3600 to 4.2GHz at 1.10v, the 1080ti was left at stock for now but i'll undervolt that later.
Running Cinebench, the CPU gets up to 76°C so that's fine so I didn't test much further. However, when stressing the GPU, things get a bit more toasty. I ran multiple Heaven benchmarks and these are the results:
| 4 fans push-pull 100% | 4 fans push-pull profile | 2 fans push 100% | 2 fans push profile | |
| CPU | 78 | 80 | 80 | 85 |
| GPU | 72 | 75 | 75 | 79 |
So a couple of conclusions:
- Temperatures with 2 fans are a bit hotter than I'm comfortable with. I'll probably go with a 4 fan setup.
- CPU temperature ramps up quickly when the GPU is stressed. I'm pretty sure the CPU cooler is well attached and there's no more air in the pump. I think it's just the fluid getting too hot because the fans struggle to cool down the radiator.
- The noise is somewhat louder than I'd like and the pump noise has an annoying high pitch to it running over 60%. I still haven't made the shell part of the case so I hope that will make things a bit better.
- I'll undervolt the 1080ti next so the temperatures get lower.
So I've been tweaking voltages of the 1080ti a bit and the results have been great. I ran another Heaven benchmark with a 4 fan setup with my fan profile, and the GPU was at 66C and the CPU at 72C. Fans were spinning at 60%, pump at 70%. GPU boosted to 1911MHz at 0.943v where before it required almost 1.1v to get to that point. I'll do some more testing with 2 fans and post some pictures of my current build later.
My 2 cents: get a temperature probe for the coolant (water) temperature, hook up the fan curve to this probe reading. It will save you some noise spikes and also will help to monitor the overall cooling solution heat soaking much better (i. e. is the coolant temperature stable or rising during the tests). If you don't have a temp sensor connector on your motherboard, you can use QUADRO by Aqua Computer.
Thanks a bunch for the tests!
Good to see the push/pull setup cools a little bit better!
Good to see the push/pull setup cools a little bit better!
I wonder now how adding panels to the case would influence these temp values!
This is the current state of the project. I also received the M.2 riser but I'll need to cut a new version of the spine if I want to use it. For now, I'll start working on the shell and try to make it work like this. We'll see later about the riser.
First thing I'll do is bend the acrylic. I'm in the process of making a DIY bender and will get back to you when that is somewhat more progressed. Be back later with more!
First thing I'll do is bend the acrylic. I'm in the process of making a DIY bender and will get back to you when that is somewhat more progressed. Be back later with more!
That's actually a good advice. Unfortunatly, Asrock boards don't seem to come with temperature sensor inputs so I'd have to resolve to the Quadro, which seems like a lot of 'overhead' in a case this small. At the moment, I'm trying to avoid anything molex or sata powered unless absolutely required. I might try this one day as it intrigues me and the software looks very comprehensive. Thanks!
I've been tinkering for a couple of hours now trying to get the GPU to stay into place without being pushed out by the riser and have come up with a 3D printed solution that might actually make this whole thing prossible. Here's some pictures to show the GPU side and just how thin the whole thing is.
Don't mind the cable mess on the bottom, it'll get hidden by the shell when that's done.
Honestly, back when I ventured into custom watercooling, most of the annoyances I ran into were related to motherboard fan/pump control, and a aquacomputer's system would have solved all of it.
But back then all that was available was their full blown Aquaearo, which was both large and quite expensive.
The quadro seems like a no brainer to anyone doing custom watercooling, especially if you want to push the thermal/acoustic limits of your loop, doesn't take much space either.
After reading this gorgeous project post I can't but wonder how it's going?
Hey there! I'm afraid this project has been stalled for a couple of reasons. Firstly, the performance of the radiator didn't really cut it without setting the fans and pump to noise levels I wasn't comfortable with and secondly, the current generation of gpu's is completely incompatible with this case so I focussed on my latest case, the Basalt (which is more or less finished atm, feel free to take a look!)
That's great to hear! Would love to see some more if you'd like to share your progress on here.
Similar threads
