Log FormD T1 Full Custom Loop - All Black no RGB

jouz

Trash Compacter
Original poster
Oct 9, 2020
51
92
Still waiting for my T1, but thought I'd to start my log regardless since I've got pretty much all other parts.

Like many, I got inspired by Ali's video's on the T1 featuring a full custom loop.
Building a new rig was long overdue, and I always wanted to do watercooling, so I went for it.
I'll follow a strict no-rgb-led policy.

Parts:

QtyPartComment
1B550I AORUS PRO AX
1Ryzen 5 3600Got a really cheap one two weeks before Zen 3 launch, but might upgrade next year when there is a decent priced Zen 3 8-core.
1Zotac RTX 3080 OCWasn't my first choice, but that's what was available for near MSRP (In hindsight, I'm pretty happy with it)
1HyperX Fury (HX432C16FB3K2/64) 3200MHz DDR4 CL16 (Kit 2x32GB)(yes I sometimes do need 64GB for work related stuff)
2Samsung 970 EVO PLUS M.2 (1x 500GB, 1x 1TB)
1Corsair SF750
1EK Quantum Vector Trinity RTX 3080/3090 Waterblock + Backplate + Rotary Terminal
1Nouvolo Aquanaut CPU + Pump Combo
1Aqua Computer QUADRO Fan & Pump controller. I was surprised how well this works.
1Laing DDC-1T Plus PWM
33x Noctua NF-A12x15 PWM chromax
1EK-CoolStream Classic SE 240
2mEK ZMT Matte Black Tubes (16/10)
10(TODO: Fittings, I got a couple more than I'll likely need)
1FormD T1 (black)(Still waiting)
1J-HACK M2426(Still waiting)


Pictures:

First pics of all the parts on a testbench:


Once I'm happy with the loop in the actual case, I'll replace the tubing with the Matte Black ZMT.
To manage the cables better, I've ordered the J-HACK, and I'll make some custom cables for the GPU, Pump & Quadro.
Only thing that I really don't like is the white end-part on the GPU waterblock. I'll ask EK if I can get a black replacement part (the ones they use on the Acryl verison of this block).

T1 Configuration:
Build is going to be in 3-slot mode, with the GPU offset to 2-slot position so there is room for airflow between the backplates of the GPU & Mainboard (without taking of those backplates).
The rad will be set up with the three NF-A12x15, but I'm planning to set the two top fans up in push mode - from dry testing on the bench this seems to perform much better, and contrary to what's being said elsewhere, I actually much prefer the push-noise over the pull noise generated in this configuration.
Finally, I'm debating to put the case in the "regular" orientation, with the radiator on the bottom, pushing the hot air out the top. I'll do some testing how much this affects performance.
 

jouz

Trash Compacter
Original poster
Oct 9, 2020
51
92
First torture test to get an estimate if I completely messed up by thinking I can cool a 3080 + CPU with a single 240 Rad.
Also the main idea is to repeat these tests after the case is completed and compare :).

For all tests (and for day-to-day use) I'm configuring the Quadro to keep the coolant at a set temperature.
For the first, heavy test I set it to 45°C (~23°C delta) degree using PID loops for both the top & bottom fans and the pump.
45° coolant was chosen as this is what causes the CPU to at least not have worse cooling performance than with the stock aircooler.

I've configured the fans & pump so that the maximun that the Quadro may run them at is still just about an acceptable noise level for me:
  • Pump max is 30% ~ 1650 RPM
  • Top (push) Fans max are 85%
  • Bottom (pull) Fan max is 75% (again, I dislike the pull noise, so I prefer this to run a little slower)

Synthetic Load
This is after running both Prime95 & Timespy Extreme in a loop.
The 3080 is Overclocked (+170MHz) and at 105% power limit to get the "worst case" of heat emission.
Together with the CPU, there is a total of ~400 Watts (according to the hw monitor) that the Rad needs to get rid of.

Here is the result after running this for 1h. The fans usually spin at up to ~75% and 65% respectively (it goes a bit back and forth periodically), I.e. not even reaching the maximum level I've set:

(Note that the Prime95 often gets the CPU to load up to 85W)


So I think there is still some room if I wanted to sacrifice on noise performance for lower coolant temperature, as well as for the to be expected worse conditions once the whole thing is inside a closed case.

Gaming Load
Here a more representative load: playing Warzone with Ultra Settings @3440x1440.
Here, I've also undervolted the GPU to 793mV@1800MHz.

After ~30 minutes, with the temperature target at 45° Fans run between 500 and 600 RPM, basically silent:



If I set the target temperature a little lower, say to 37° (15° delta), the fans run a little harder around 800rpm, which is where they just about start to get noticeable (but very pleasant still):
 
Last edited:
  • Love
Reactions: eedev

jouz

Trash Compacter
Original poster
Oct 9, 2020
51
92
Two more runs to replicate the testing method @eedev used for his case: 30 minutes of Unigine Heaven 4.0 at 100% & 60% fan speed, with time plots.

TLDR:
The test causes a combined average load of ~235W from CPU & GPU to be cooled.
  • @60% fans: max ΔT = 14.4°C
  • @100% fans: max ΔT = 11.6°C
GPU stays easily under 40°C in both cases.
CPU runs a bit warmer but stays at around 46°C@100% fans, and at 50°C@60%.

It seems that 235W is a little bit above what COD generated in the test above, but I think it's going to be a bit below what we can expect "next gen" games to generate.
Either way, this further supports that there should still be headroom for once the system is in a closed case, especially when running the PID loop to target a slightly larger ΔT.

Data:

Fans fixed @100% (resulting in ~1727 RPM):



meanmaxmin
Ambient Temp22.0 C22.5 C21.2 C
Coolant Temp32.7 C33.6 C32.5 C
CPU Temp45.8 C55.5 C42.4 C
GPU Temp38.3 C40.0 C30.0 C
CPU Package Power36.9 W46.8 W30.4 W
GPU Package Power197.0 W217.0 W11.7 W
Fan Speed1727 RPM1730 RPM1684 RPM



Fans fixed @60% (resulting in ~1215 RPM):



meanmaxmin
Ambient Temp22.6 C22.8 C22.1 C
Coolant Temp36.3 C37.0 C32.9 C
CPU Temp50.1 C62.9 C45.5 C
GPU Temp42.7 C45.0 C37.0 C
CPU Package Power36.6 W46.8 W31.7 W
GPU Package Power202.0 W221.0 W132.0 W
Fan Speed1215 RPM1226 RPM887 RPM
 
Last edited:

Lytnin

SFF Lingo Aficionado
May 2, 2020
97
200
Did you run into any bottle necking from the cpu paired with a 3080?
Nice part list/build overall... End result should be good.
 

jouz

Trash Compacter
Original poster
Oct 9, 2020
51
92
Thanks!

Did you run into any bottle necking from the cpu paired with a 3080?

So far, definitely no CPU bottlenecks. It's been pretty balanced for playing in ultrawide resolution (i.e. 3440x1440 - which is inbetween 1440p and 4k).
I also don't play any esports games, and don't have a 144+Hz monitor - as long as I can manage Cyberpunk @ 60FPS on this screen I'm happy :)
 
Last edited:

eedev

Cable-Tie Ninja
Apr 23, 2020
154
288
Nice!! Hopefully you'll get the case soon! Very interested in 3080 cooling!
 
Last edited:

Cookie

Case Bender
New User
Dec 1, 2020
2
0
Hey there, nice build. Cyberpunk won't be an issue. Congrats getting a 3080, I had no chance getting one :( Maybe one day if stock ever replenishes.
 

jouz

Trash Compacter
Original poster
Oct 9, 2020
51
92
Finished the (first version of the) build a couple days ago and finally got around to documenting it. Pretty standard 3-slot setup, except I've offset the GPU to 2 slot, which allowed a little more air between GPU and Mainboard (although that space is also used by the ATX cables for now...).






Some notes:
  • I used all regular cables for now. I'm still waiting for the J-HACK M2426 to arrive, which is when I'll reassemble the case and update all cables.
  • I did a test job with that yellow vinyl foil to cover the ugly white GPU cover. I might try something nicer when I rebuild the case
  • All fans & radiator are mounted with screws. I did sand the bottom fan a little bit in the areas where it touches the mainboard and the fittings that go into the radiator, because otherwise the fit was so tight that the fan got compressed to the point that it would start rattling.
  • The DDC Heatsink does cause some bulge in the panel. I'll definitly either sand it down or remove it entirely next time I'm draining the loop.
  • I'm aware the mini reservoir is below the pump. I've just used it for filling and bleeding the loop (operating the case "upside down" in the meantime).
  • I did at first try out the "regular orientation" (radiator on bottom), with a positive pressure (fans pushing air into the case from the bottom), but the performance was so bad, I immediately switched to the recommended pull / radiator on top setup.
  • It's pretty cramped now above the PSU. Still debating whether or not I should change the routing, maybe going from the CPU over the top-back part of the case. Will decide after getting the custom cables.
Performance in closed case compared to the open testbench:

Reporting the mean numbers of the last 5 minutes after ~45 minutes of running heaven, putting the numbers from the previous runs with on the open testbench next to it.

Fans fixed @60%:

TestbenchClosed T1
Ambient Temp22.6°C22.7°C
Coolant Temp36.3°C45.3°C
CPU Temp50.1°C64.8°C
GPU Temp42.7°C56.9°C
CPU Package Power36.6 W41.2 W
GPU Package Power202.0 W212.0 W
Fan Speed1215 RPM1293 RPM

Fans fixed @100%:

TestbenchClosed T1
Ambient Temp22.0°C22.6°C
Coolant Temp32.7°C39.6°C
CPU Temp45.8°C59.0°C
GPU Temp38.3°C50.5°C
CPU Package Power36.9 W41.8 W
GPU Package Power197.0 W204.0 W
Fan Speed1727 RPM1810 RPM

So as you can see the cooling performance deteriorated quite a bit, more so at 60% fan speed (ΔT increase of ~9°C) than at 100% (ΔT increase of ~6°C)
However, also the components are overall hotter: While the CPU was previously ~14°C above coolant temperature, now it is 19.5°C above. For the GPU it went from ~6°C to ~11°C above coolant temperature. I'm guessing this is due to the poorer passive cooling of the system compared to the open bench.

To be honest, I hoped for less of a decrease in cooling performance. On the other hand, these numbers are pretty close to what others seem to get with similar setups (see for example the benchmarks from @eedev ), so it was to be expected. I'd still consider this setup very usable - albeit not quite silent when under heavy loads.

Of course, I'll rerun the same tests after I've cleaned up the cables - maybe there are another 1 or 2°C to gain :).
 
Last edited:

eedev

Cable-Tie Ninja
Apr 23, 2020
154
288
I'm happy you were able to build it!

Sanding was necessary for the DDC Heatsink, I did not do this for nothing!!! 😹

I like the yellow waterblock👍
 
  • Like
Reactions: jouz

jouz

Trash Compacter
Original poster
Oct 9, 2020
51
92
Logged temps, power consumption from Playing ~2.5hrs of cyberpunk last night, to show some real world performance when targeting a 43°C coolant temperature with the Quadro PID control (The top plot shows temps in C, the Middle one shows Pump and Fan speed, the bottom one showing power consumption of CPU and GPU in W). The frequent drops in load/fan speed are during moments when the in game menu is open, and during some less demanding cutscenes.

 
Last edited:

jouz

Trash Compacter
Original poster
Oct 9, 2020
51
92
Small update - Rev 2

Decided to make the switch to quick connect couplings (QDCs) for filling with an external reservoir, discarding that mini reservoir.
I've placed the QDCs on the outside because I've also started contemplating on getting an external radiator down the line (mostly for if I decide to upgrade the CPU).


As you can see, I've cleaned up the routing altogether, used the two screw holes in the front spine to mount the Quadro.
I'm not 100% happy with the routing for the aquanaut inlet, but after contemplating a lot of options (i.e. rotating aquanaut 180deg), this one was as good as any. I do wish we could have the aquanaut in any rotations (i.e. both ports on top) for AM4 sockets.

I've also found a cleaner solution for the ugly white piece on the EK radiators by disassembling it and individually wrapping it in vinyl.

I'm happy with the QDCs, overall they make bleeding the loop much easier, as the small reservoir really did a poor job of capturing the medium to small bubbles later in the process - it worked, it just took way longer. I'm not sure how well cheaper QDCs would perform (in terms of restricting flow & preventing drip), but for the price of these Alphacool ones I'm still not sure if it's worth it compared to the price of the tiny reservoir, especially if you can build a clean loop with the res/fill port at the top (which however is a challenge in the T1).

I've also switched back to the regular casing on the pump for now - no more panel bulging :). As you can see below, removing the ddc heatsink does seem to affect CPU temps, so I'll definitely sand it down and put it back on in the future.

Performance:

I was a little concerned if QDCs would have negative impact due to potentially reduced flow. This seems unfounded, the overall performance actually slightly improved (at the same pump speed) - I'm guessing thanks to better airflow by making more room under the push/pull fans.

As usual, 1h run of heaven, again added the values from previous setups to the table:

Fans fixed @60%:

TestbenchRev 1Rev 2
Ambient Temp22.6°C22.7°C21.1°C
Coolant Temp36.3°C45.3°C41.6°C
CPU Temp50.1°C64.8°C64.3°C
GPU Temp42.7°C56.9°C54.7°C
CPU Package Power36.6 W41.2 W40.5 W
GPU Package Power202.0 W212.0 W214.0 W
Fan Speed1215 RPM1293 RPM1242 RPM

Fans fixed @100%:

TestbenchRev 1Rev 2
Ambient Temp22.0°C22.6°C21.4°C
Coolant Temp32.7°C39.6°C37.4°C
CPU Temp45.8°C59.0°C60.1°C
GPU Temp38.3°C50.5°C49.5°C
CPU Package Power36.9 W41.8 W39.9 W
GPU Package Power197.0 W204.0 W202.0 W
Fan Speed1727 RPM1810 RPM1755 RPM


@60% fixed fan speed, coolant went from ΔT=22.6°C to ΔT=20.5°C, the GPU got cooler by that amount as well. The CPU temperature however stayed the same - I think the pump heatsink helped a little with CPU temps! Similar trends when looking at the results @100% fixed fan speed.
 
Last edited:
  • Like
Reactions: FreshGnar and eedev