Air Cooling Optimisation Experiment and Results

[direkteffekt]

Background

 

I recently switched my PC back over to being fully air cooled after being burned with some GPU returns and being left with a GPU waterblock for a card that I no longer owned. That and some other bad luck left me to decide that custom water was no longer worth the hassle.

 

I got a new air-cooling focused case and CPU cooler and switched back. The key hardware is:

• Case: Corsair 5000D
• Case Fans: Noctua NF-F12 PPC
• CPU: Ryzen 9 5950X
• CPU Cooling: Noctua NH-D15 Chromax Black
• GPU: MSI RTX 3090 Suprim X

 

My Problem

 

After switching back and running the system for a while, I found myself a little annoyed at the temperatures and the sound levels that my GPU was reaching. I started with trying some curve optimization and trying to manually adjust fan curves, but I felt like I wasn't quite getting what I wanted. After a little bit it occurred to me how ridiculous it is that with the design of cases these days, the GPU has only 1 of the front case fans pushing air directly into the cooler, while there are effectively 2 fans feeding air to the CPU which uses half (or less) as much power.

 

My Solution

 

My solution to the problem was to 3D print a number of ducts for various components of the PC to try and optimise distribution of the front case fan's air flow and to reduce the re-circulation of hot air inside the case.

 

Fist up were the GPU airflow components:

 

In addition to GPU airflow components I designed a duct for the CPU cooler, to take air from the last remaining front intake fan:

 

The Result

 

The resulting PC looked like this:

 

To try and confirm that my experiment was a success, I did my best to test the improvements offered by the components as they were added in. For the purposes of these tests I had the following configuration:

 

CPU: Stock settings, no PBO

CPU Load: Looped Cinebench R23 in Stability mode

 

GPU: Custom curve 1605 MHz @ 825 mV Core, 9000 MHz RAM, 35% fixed fan speed (Approx 1100 RPM). Results in about 300 W power draw

GPU Load: Stationary in Deep Rock Galactic space rig at spawn point

 

Case Fans: Fans fixed at 960-970 RPM

 

With the fixed fan and clock speeds it should give as good an idea as possible of the performance increase. I tested in four stages, first without any ducting, then with the GPU duct, then adding the GPU box and finally installing the CPU duct.

 

Each test was run for about 20 minutes to allow everything to get up to temperature. The ambient for all tests stayed around 23C according to my desktop thermometer. All temperatures were the averaged values for the last minute of testing based on HWINFO logs.

 

Test 1 - No Duct:

• CPU: 65.7 C
• GPU: 79.8 C

Test 2 - GPU Duct Only:

• CPU: 66.9 C
• GPU: 68.0 C

Test 3 - GPU Duct & Box:

• CPU: 66.9 C
• GPU: 65.4 C

Test 4 - All Ducts Installed:

• CPU: 64.2 C
• GPU: 66.1 C

 

Conclusion

 

Based on what I was trying to achieve (lower noise and lower temps) I consider the experiment a success. Ducting the air from the lower fans to the GPU provided a huge increase in cooling capacity at fixed fan speeds (about a 12 C drop). Installing the GPU Box dropped an extra 2 C.

 

A side effect of the more directed air to the GPU resulted in additional recirculated air making it to the CPU cooler, which slightly increased temperatures (although given this was not strictly temperature controlled could have been an effect of ambient temps), but installing the CPU duct offset this change with an additional improvement in cooling to the CPU.

 

From these results, while all steps may not be necessary I can recommend providing some level of GPU ducting to those who are interested in achieving better cooling performance and noise levels to go with it.

[Ralfi]

This reminds me of a duct a Youtuber made for their GPU, which also worked well.

 

 

Good work!

 

Someone should really start manufacturing these, although it'd be very hard to come up with ducts as every case/GPU/CPU cooler's different.

 

[direkteffekt]

12 minutes ago, Ralfi said:

This reminds me of a duct a Youtuber made for their GPU, which also worked well.

 

 

Good work!

 

Someone should really start manufacturing these, although it'd be very hard to come up with ducts as every case/GPU/CPU cooler's different.

 

I actually watch his fan showdown videos. Didn't see that video.

 

But yeah, I don't think you could realistically manufacture them given the number of different variables that need to be taken into account. What might be possible is a dynamic 3D model where people could input their own measurements for their own cases and hardware that would generate a 3D printable model that people could get made.

 

If there was enough data could even have selections of hardware with known measurements (cases, GPUs, which of the rear PCI-e slots the card sits in etc.).

[jones177]

Good idea.

My Corsair 5000d/3080 ti sort of failed. 

 

I bought a Corsair 5000d mainly because my cases were not deigned for a hybrid GPUs and I was on auto notify for a lot of them. I ended up getting all air cooled cards.

 

I first tested the 500d with a 2080 ti and found that 3 120mm fans in the front did not move any air below the GPU so it ingested it own hot air. Temps were around 75c in games so not a complete disaster yet. 

Then I tested it with an MSI 3080 ti and it gamed in the 80s.  I replaced the rear exhaust fan with a Noctua and added 3 more 120mm Corsair fans on the side and it still did not move enough air from blow the GPU.  I added a 60mm Noctua below the GPU and  was back down to 75c.

 

The build was not getting the MSI Gaming X Trio but an EVGA FTW3 Ultra 3080 ti that uses 50 watts more at stock so it ended up back in the 80s. At that point I gave up on the case and used it in a Strix 3080 build.

 

The EVGA FTW3 Ultra 3080 ti ended up with a Lian Li o11 dynamic that intakes air below the GPU.

 

 

 

 

[Jr2_]

Very cool! All you need now is some blowiematrons, and you've made a server

 

also that cool visual style gotta help with the temps, right?

[direkteffekt]

16 minutes ago, jones177 said:

Good idea.

My Corsair 5000d/3080 ti sort of failed. 

 

I bought a Corsair 5000d mainly because my cases were not deigned for a hybrid GPUs and I was on auto notify for a lot of them. I ended up getting all air cooled cards.

 

I first tested the 500d with a 2080 ti and found that 3 120mm fans in the front did not move any air below the GPU so it ingested it own hot air. Temps were around 75c in games so not a complete disaster yet. 

Then I tested it with an MSI 3080 ti and it gamed in the 80s.  I replaced the rear exhaust fan with a Noctua and added 3 more 120mm Corsair fans on the side and it still did not move enough air from blow the GPU.  I added a 60mm Noctua below the GPU and  was back down to 75c.

 

The build was not getting the MSI Gaming X Trio but an EVGA FTW3 Ultra 3080 ti that uses 50 watts more at stock so it ended up back in the 80s. At that point I gave up on the case and used it in a Strix 3080 build.

 

The EVGA FTW3 Ultra 3080 ti ended up with a Lian Li o11 dynamic that intakes air below the GPU.

 

 

 

 

Believe it or not, I actually started out with an O11D. I changed it out because I couldn't fit a D15 in it when I was getting away form water cooling. Definitely has a better layout for providing optimal GPU airflow out of the box; although this particular GPU is so wide it didn't leave much room beside it for air to move out against the side panel  though.

[jones177]

42 minutes ago, direkteffekt said:

Believe it or not, I actually started out with an O11D. I changed it out because I couldn't fit a D15 in it when I was getting away form water cooling. Definitely has a better layout for providing optimal GPU airflow out of the box; although this particular GPU is so wide it didn't leave much room beside it for air to move out against the side panel  though.

It is a major draw back with the case.

My two setups are connected to TVs and only used for gaming so not a big deal to me.

The computer I use for this sort of thing is the i9 9900k with a D15. I can't hear it at all at 3ft away but I can hear the 5900x with the AIO that is 10ft away. 

 

[thrasher_565]

1 hour ago, direkteffekt said:

Believe it or not, I actually started out with an O11D. I changed it out because I couldn't fit a D15 in it when I was getting away form water cooling. Definitely has a better layout for providing optimal GPU airflow out of the box; although this particular GPU is so wide it didn't leave much room beside it for air to move out against the side panel  though.

it be nice to see this compared to a bottom intake case like the 011 mini (with 120mm fan in the back) the 011 d is flawed because of that or compared to the lianli lancool 2

 

i have a test witch i been lazzy about in my 011d air about front intake vs side intake but still need to do the test haha. i manged to get an Scythe Kotetsu Mark 2 to fit in the 011d air... but its an aio built case not air flow case.

 

but your test just shows how the front intake is not the best for gpu cooling witch i been say for some time now.