Swapped my ML120 PROs for Noctua NF-F12 IPPC-3000 PWM (aka why static pressure matters)

Joined
Oct 23, 2018
Messages
1,011

TL;DR
The ML120s were very good, but the Noctuas are way way better, prob due to higher static pressure.


I thought the non-RGB ML120 PRO was a great fan which offered very good radiator performance at a very quiet sound level. Then I tried these Noctuas. Woah.

I'm running a 5000D with a 360x45 rad up top with fans as exhaust in pull, a 360x45 rad on the side with fans as intake in pull, a 120mm rear exhaust, and a trio of QL120s (🤮) as front intakes. With the machine running at full load and all 10 fans maxed (2400rpm for the MLs, 1500rpm for the QLs), coolant temp was a pretty steady 39C. I don't like the coolant to be above 40C, so 39C with everything maxed out was not a great place to be for long term reliability.

When I replaced all 7 of the MLs with the Noctuas and set them at 2400rpm, the coolant is now a pretty steady 37C with the machine at full load. At this speed, the noise level is essentially the same as the MLs. If I set the controller to 39C, the rpms drop to about 2000rpm, the pump slows down, and it actually becomes quieter than the MLs were at the same temp. I think I can live with the temp being 39C now that the cooling system is operating at a much lower duty cycle than before.

With the fans at full bore (2900-3000rpm), the machine is MUCH louder than the MLs, and the coolant comes down to 36C. I think the QLs are a limiter here. They're only listed at 1.55mm-H2O compared to the 7.63 of the Noctuas. I think the side intake fans on the rad are generating enough pressure that the QLs are having trouble acting as intakes. I have a few more of the Noctuas in a box, as well as 6 of the low-profile NF-A12x15. I'm going to try both of those as front intakes and see how it goes, plus I'm going to try using the thin ones as push fans on the top rad (to give it push-pull).

I have a bigger build which is using the even worse performing ML120 PRO RGB on 2x480x60 rads in push-pull. I'm going to be swapping those out for the 3000rpm Noctuas along with all of the other fans. That should really make a huge difference on that build.
 
Well a big limiter is what's your air intake temp? What sort of delta are you aiming for?
 
The intake temp was the same for both setups. That's what makes it a valid comparison. At 2400rpm, the Noctuas produce a 2C greater ∆T.

I don't aim for a specific ∆T since it isn't a meaningful metric in this context. ∆T doesn't make tubing go soft - T does. ∆T doesn't make components run cooler - T does.

∆T is sort of useful if you're trying to use less radiator and want to calculate just how little you can get away with. In this case, ∆T is used in the power dissipation calc. The thing is, why would you ever want to run less than the maximum radiator capacity? Because you enjoy the sound of fans and want to make sure they're loud enough?
 
I guess one thing to note is that the coolant temp I'm referencing here is at the pump (the built in sensor on a D5 Next). That means the coolant has passed through the radiators and then been warmed back up by the GPUs. The ∆T between the pump sensor and the High Flow Next at the outlet of the second rad is even bigger now than it was with the ML120s, but I primarily monitor the pump sensor since that one is hotter and is the one used for system control.

Here's a reference pic of the system. It's out of date now that the fans have been replaced and a Leakshield has been added, but the loop routing and other components are still the same.

1654182618710.png
 
Lol I wasn't knocking your test. Only curious because those coolant temps are quite a bit higher than I'm used to seeing.
 
I think they're high because this really isn't an ideal setup for cooling.

Part of it is that the thermal loading is somewhat high for this amount of radiator. There's about 800W of heat getting put into the coolant during renders. 2x360x45 isn't a lot of radiator for that. It runs much cooler with only one GPU going.

Part of it is that it's a somewhat restrictive loop with three blocks, two rads, two QDCs, and a handful of tight 90º corners. The D5 maxes out at 140l/h, but the coolant temp vs flow rate curve hasn't flattened out at that point.

Lastly, the ideal radiator setup here would have one on the front for intake instead of on the side. Unfortunately, the case is about 10mm too small for that unless a skinny rad were used up top. Another option could be to swap the X-flow side radiator for a regular one. The shorter top end tank might allow for the fans on that to be relocated to the other side (in push) where they would be more efficient. But doing that would mean the routing wouldn't be as clean. Function over form... mostly :)

I'll stick a probe on the front intake tonight to get an idea of what the intake air temp is like. The room is definitely warmer than my living room, but I have an air conditioner partially blowing at it. I'd guess the intake air is around 23C, but I really don't know just yet.
 
Intake air is hotter than I thought. My thermocouple is showing 26-26.5C when placed 1" off of the side intake and 24.5-25C at 1" off of the front intake. As I type this, the Noctuas are at about 2000rpm. Coolant temp at the pump is 37.2C, coolant temp after the second rad is showing 34.4C.

Image from iOS (41).jpg


I'm really looking forward to swapping out the fans on the 1000D build sitting behind it. These fans should make an even bigger difference on that one - I'm expecting almost a 10C drop.
 
Back
Top