I run the D5 on my EKWB Quantum FLT at 34% / 2383RPM

Seems to be the sweet spot where its pretty much inaudible (to me) yet moves enough water to keep things cool

Thats on a 2x Rad +CPU & GPU

i used to run mine between 55 and 75% with 3 rads, CPU, GPU, GPU active backplate.
technically it could have stayed at 55, but accelerating did shave a deggree or two on the GPU.

Now with a DDC, i have the same flows at lower speeds, with less noise. Sometimes there's no need for dual D5 :p

Testing showed that the best flow rate was around the 1gpm (US) or 227 l/h for optimal cooling which can be hard to achieve without a duel pump setup so most people with a single pump set it at max but ideally one needs a flow meter.

the important question is "are your water-cooled components running at a temp that you're happy/comfortable with?"

I have no idea what my flow rate is all I do know is that (and forgetting other variables for the moment) with my (single D5) pump running near as dammit silent at 34%/2383 RPM means the CPU doesn't exceed a delta of 45/46c and the GPU around the 20/21 c delta mark.

Increasing the pump speed drops maybe 1 c off that but is rather noisy.(in comparison)

Zitat von »Bartdude«

Testing showed that the best flow rate was around the 1gpm (US) or 227 l/h for optimal cooling which can be hard to achieve without a duel pump setup so most people with a single pump set it at max but ideally one needs a flow meter.

i wonder what testing it was.
temperatures rise massively below a certain speed but even with a loop as restrictive as mine, it was below 30% PWM on a D5 which was like 70L/h. above 100l/h, up to the maximum flow i can achieve (~220l/h) there's maybe 1°C difference. and that's with a 10900k overclocked and a 3090.

Zitat von »Remayz«

i wonder what testing it was.
temperatures rise massively below a certain speed but even with a loop as restrictive as mine, it was below 30% PWM on a D5 which was like 70L/h. above 100l/h, up to the maximum flow i can achieve (~220l/h) there's maybe 1°C difference. and that's with a 10900k overclocked and a 3090.

I think most people refer to the Xtreme Rigs testing from 2015. You can see the summarized table in this image and the full writeup here.

0.5 GPM is ~113 Lph
1 GPM is ~227 Lph
1.5 GPM is ~340 Lph

This is min maxing, in a real system this is not important, if you do not want to dissipate the largest possible amount of heat with the smallest possible radiator.

yeaaaa... the test is awesome in its own right but you have to take it with a pinch of salt.
They test single radiators with an aquarium heater which is pushing them to their limits (which is exactly what you want for benchmarking).

In our cases where we have that load split between 2 or more rads, the importance of flow drops drastically.
Also, they themselves state that flow has very little impact on performance, and that it only impacts some radiators below the 0.5 gpm mark (https://www.xtremerigs.net/2015/02/11/radiator-round-2015/5/ the spoiler about flow rate data)
That's about 110l/h, with a single rad loaded balls to the wall.

Splitting 4 or 500W on two rads or more, you can go way below that and be totally fine.

It's very easy to test anyway. load the GPU, and reduce flow. The temp will climb extremely slowly for most of the speed range. And there's a moment when it will spike, along with the deltaT between radiator input and output. That's where the flow is too low and starts to kill your temps.
For me it happens below 70L/h (0.25 GPM). my GPU temp rises maybe 2°C between 210 and 75 l/h, then between 75 and 55 it may rise 7 or 8°C.
It will vary depending on your waterblock efficiency of course.

Now if you game with a 12900K and a 3090 on a single 360mm rad, by all means crank the speed up, but you have other problems :p

Yes, the fact that coolant flow rate doesn't have much impact was exactly the point. You asked why people say that and I provided the Xtreme Rigs data that is the basis for that.

I disagree about "pinch of salt." IMO, this is extremely useful data for trying to size a system. They did not test only "balls to the wall." Yes, they tested with fans at 100% load (1850rpm in this case). In addition to that, they also tested with fans as low as 750rpm which is closer to 30% duty cycle on many popular desktop fans.

Looking at the higher dissipation rads in their table, you could choose to go with a single Phobya G-Changer V2 (column 4, row 4) 360 rad with the pump and the fans all running at 100% (1.5gpm/1850rpm) and be able to dissipate 472W with 10CdT. Or, you could slow the pump to .1gpm, slow the fans to 750rpm, and just use two of that rad in parallel to dissipate 400W. Looking at the table, this would then leave the machine with a fair amount of extra capacity for emergencies or increases in loading.

You are correct that it becomes a bit more difficult to calculate when the radiators are in parallel, but they do actually provide the data to allow you to do that calculation yourself. You would need to figure out the interior case temp unless both rads are set up as intakes.

the pinch of salt was about taking time to interpret the test results :p
the entirety of the test is excellent, i have nothing against it at all. But our interpretation of those numbers may not be accurate if we don't read it through.

But yea our typical loop even at low flow will have pretty low dT compared to the tests. most 360 rads are able to dissipate between 4 and 500W of heat. I usually advise people to count 300W per 360 rad just to have headroom and not have a jet engine on their desk.
The tables gives results in watts, but i doubt we'd like to know the water temperatures they get. The hotter it is, the more efficient the radiator becomes, so these 10°CdT may equate to your GPU running at air cooler temperatures.
Again, it's hard to directly apply those numbers to a real use, but at least it allows us to see what are the most efficient rads.

... and then we install them in a case with glass panels, dust filters, RGB fans with smaller diameter blades because they take space to put those leds on the frame, and the case ends up in the corner of the room on the desk in its own bubble of hot air :p

I set mine to 67% I have 252 L/h with 420 rad and just my 3960X, the ∆T is less than 1 ℃ normally

I did try that but found it better to have all 3 rads working with fans at very low speed at idle, but .. the reslt depends on your fans obviously.
Some Arctic and BeQuiet can spin at stupidly low speeds. I'm testing some bequiet light wings high speed, and these buggers wake up at 7% PWM at 270 rpm lol
even without going that low you can still have a dead silent box with working fans.
A little trickle of air through 3 rads VS slightly more air through just one.
I also found that when the fans wake up from 0 RPM, the change is a lot more noticeable than a slight speed increase across all fans.

But again.. depends on what fans, what case etc... have fun tinkering

If I recall correctly i think the original testing was done over at Skinnee Labs but unfortunately that no longer exists

Zitat von »Remayz«

I did try that but found it better to have all 3 rads working with fans at very low speed at idle, but .. the reslt depends on your fans obviously.
Some Arctic and BeQuiet can spin at stupidly low speeds. I'm testing some bequiet light wings high speed, and these buggers wake up at 7% PWM at 270 rpm lol
even without going that low you can still have a dead silent box with working fans.
A little trickle of air through 3 rads VS slightly more air through just one.
I also found that when the fans wake up from 0 RPM, the change is a lot more noticeable than a slight speed increase across all fans.

But again.. depends on what fans, what case etc... have fun tinkering

I have 3 x Arctic P12s on each of my two radiators. As I sit here typing this they are currently idling at 169rpm.

Another thing to take into consideration is reservoir size as a larger mass of water helps to buffer the temperatures (I'm running with an EKWB Quantum FLT 360)

Now to the actual question, I have my D5 NEXT pump set to sound the most comfortable to me, in my current system that is 75% or 4235 rpm.
I end up with 112 l/h in a system with 4 420x45 radiators.
If the goal is the lowest possible noise nothing beats radiator area. My Noiseblocker NB-eLoop B14-PS fans spin at about 500 rpm.
The only thing I hear in my room is my Synology NAS.

Zitat von »alexkyriak«

Zitat von »WinstonWoof«

the important question is "are your water-cooled components running at a temp that you're happy/comfortable with?"

Geat reading this debate, guys. I think this quote above for me is the key question. I've only ever built two PCs in my life, the first one is watercooled and runs Aquaero. The new PC I've just built runs Aquaero also, obviously, and this one I'm setting up to be as silent as possible during normal operation.

So, I'm not chasing absolute minimum temps this time, whereas on the first PC I had (somewhat noobishly perhaps) set up the profiles to chase minimum temperatures. I was "worried" if the CPU was running above 30-35 degrees idle. Which in terms of the normal operational capacity of the chip is obviously extremely low.

So whilst I get a kick out of making the system run as cool as possible, this time I'm interested to run it as silent as possible overall for normal use. The reason for this post/thread initially is partly to see how others approach this aspect of cooling. And also because with the profile I'm trying to develop, the loudest bit of kit in the PC is the pumps, so am trying to run them at 30%-35% to make them "disappear". Currently playing with having a fan profile on my three radiators to only have one set of radiator fans running normally (the other two at 0 RPM), until the water temp hits a certain threshold, at which point the other two sets of radiator fans start up from 0 RPM to start helping out.

Dont forget you can use Playground to proactively autoswitch profiles (amongst other things) when a specific program is opened.

https://www.youtube.com/watch?v=NieGE9AMkas

NB not shown in my demo above but you could add a delay step to either output so that the Profile doesn't switch until after X amount of time

I built my system for optimzied performance and silence. I have 2 360mm rands, a 3090 GPU Block and a EK Velocity 2 block on the 5950x

I started with a single D5. I had to run it at 70% to keep flow rate decent (temps still dropped a few degrees when pushing it to 100%), but I was never happy with this, the D5 whines too much at anything past 60%

Then I tried two D5's. As others have said, the pitch seems to increase when running them at the same speed, and honestly I wasn't seeing the performance I expected, still having to run both at 70% to get the flow I wanted.

So I figured my loop needed more pressure from the pump, so I tried a DDC. Worked great... but it was too noisey even at 60%

So then I tried 2 DDC's (ek 3.25s). Still to noisey, but man the flow was great, just couldnt get the pitch under control even at low level.

So then I tried a D5 and a DCC 3.25 from ek. WORKED AMAZING, D5 at the resivour and DDC halfway in the loop pumping into the rads. the D5 stays at 40% and the DDC stays at 40% (2100 rpm, the % is weird with these pumps), it works great... except after a while the flow rate drops and its like the pumps are working against each other. No clue why this happens, but I tore the loop apart and built a test bed in the garage and ran it again in different comfigs. No matter what I did the d5 + DDC was the best combo BUT after 30 minutes the flow rate would sink.

So then I tried an old Swiftech MCP50x I had in place of the EK DDC. PERFECT! So thats what I'm running now, D5 at 40, MCP50 at 40, 100% silent, amazing flow.

My current guess is the DDC 3.25's propeller is eventually causing issues when paired with the D5 and somehow fights against the flow, by coming out of sync or something with the flow and movement. The MCP50x has a way beefier looking propeller that seems to have no issue muscleing through things. I also noticed the DDC would whine in weird ways, like it was fighting then not fighting over and over against. I might try 2 MCP50's in the future, just hard finding a new one to buy, and right now I only have 1.

Also big plus, the MCP50 runs way cooler than the DDC3.25