-------------------------------------------------------------------------------------------------------------------------------
Cerberus-X 2025 : Experimenting the 1-Flow setup (BTX revival ?) (05th August 2025)
-------------------------------------------------------------------------------------------------------------------------------
The 1-Flow Layout
When I got my Noctua D12L I tested it, at first, in the vertical orientation (airflow) in my Cerberus-X
However, results were mediocre in terms of cooling & noise.
Cerberus-X late 2024 : Cooling the AMD Ryzen 7 9800X3D (19th December 2024)
Since, I discovered & adopted the Double 120mm Fans Bracket (thanks to the RTX 5080 Ventus) and the GPU with double airflow passthrough, i.e. the RTX 5090 FE.
The idea behind this common airflow is not new as it was promoted by Intel with the BTX (RIP..

).
Inside the Cerberus-X, it is pretty easy : Intake from the Bottom (tower stand) / Left (vertical stand, as I’m using) and Exhaust at the Top/Right.
While using the RTX 5090 FE, there is no longer a wall inside the case, making this unified airflow a viable possibility.
Here is the final result (I’m amazed by the natural light)
Here is the full setup, as there were some updates :
- CPU : AMD Ryzen 7 9800X3D (AGESA PI 1.2.0.3e), PBO Curve Optimizer -15
- Motherboard : Asus ROG STRIX B850-E GAMING WIFI
- CPU Cooling :
- Heatsink : Noctua D12L with offset mounting bars NM-AMB15
- Cooling fan : Noctua NF-A12x25r
- Thermal paste : Arctic MX-6
- Contact frame : Thermal Grizzly
- RAM : Corsair Vengeance 2x24Go 7600 MTS (XMP profile). Booting time is also excellent.
- GPU : NVIDIA RTX 5090 FE connected at PCIe 8x Gen5
MSI Afterburner Profiles :
- Stock (575W, PL 100%) : Thermal Bench
- Undervolted (510W, PL 100%) : Core with Curve 825-910 @2700Mhz, GDDR7 +3000Mhz
- Overclocked (575W, PL 100%) : Core with Curve 825-End +300Mhz, GDDR7 +3000Mhz
- PSU : Corsair SF1000 with its standard cables.
- SSD#1 : Samsung 9100 Pro 2 To (Gen 5), connected to the CPU at PCIe 4x Gen5, Motherboard heatsink
- SSD#2 : Samsung 990 Pro 4 To (Gen 4), connected to the CPU at PCIe 4x Gen5, Motherboard heatsink
- SSD#3 : WD SN850X 2 To (Gen 4), connected to the CPU at PCIe 4x Gen5, Motherboard heatsink
- Intake-Bottom Fans : 2xNoctua NF-A12x25 Chromax/black (~1150 rpm / 50% PWM) + Inlet side spaces NM-IS1-12 Chromax.black Sx2, Intake
- Power charged-Bracket Fans : 2xNoctua NF-A12x25 (~1350 rpm / 60% PWM) with Akasa Fan Bracket AK-MX304-12BK
- Exhaust-Top Fan : 1xNoctua NF-A12x15 Chromax (~1350 rpm / 60% PWM)
- Exhaust-Rear Fan : 1xNoctua NF-A9x14 Chromax (~1950 rpm / 60% PWM)
- Front i/o : 2xUSB Type-A, 5Gbps
Gaming Performance
Regarding the performance, let’s start with the gaming/GPU thermals, under a constraint run of 3DMark SpeedWay of 20 minutes (please check the music in the background, really good)
I’ll compare with the FormD T1 as the methodology & testing conditions (ambient temperature 27°C, Nvidia drivers) are similar :
RTX 5090 FE (575W Stock) Core Max T°, GDDR7 Max T°, Fans Speed Max
- Cerberus-X (D12L 1-flow) : 74°C / 92°C / 1600 rpm
- FormD T1 : 80°C / 98°C / 1840 rpm
Without any surprise, the gaming performances are top-notch for this Cerberus-X D12L 1-Flow setup.
CPU Cooling Performance
What about the CPU cooling test?
As usual I’m using Cinebench 2024 Multicore & IBT 2.54 to test the CPU cooling capabilities.
The Cerberus-X is completely closed and I’ve set up the RTX 5090 FE @1600 rpm.
The latest AGESA seems to have slightly altered the CPU behaviour on Cinebench 2024 Multicore as I got a score of 1319 instead of 1360 (Max) or 1349 (June 2025)...despite having no thermal throttle to report!
In terms of temperature, on Cinebench 2024 I got a peak temperature of 92.5°C at the start, time for the fans to kick in. Afterwards the R7 9800X3D T° is hovering around 87-88°C and it is boosting constantly at 5225Mhz, while using 129W.
Even in Intel Burn Test 2.54 (more power angry than Cinebench 2024 multi), my setup was able to avoid any thermal throttle, even if it was close. Max T° 94.6°C, Boost at 5225Mhz, using 139W.
As hoped, the RTX 5090 FE is playing well its role of airflow passthrough!
Conclusion
Compared to the Cerberus-X with the C14S, this D12L 1-Flow setup is offering a clear advantage on the noise as I was forced to push the 140mm fans speed to 1200 rpm to exhaust all hot air from the RTX 5090 FE.
The D12L 1-Flow setup has clearly a smoother airflow versus the C14S setup.
Regarding the CPU temperature difference between the C14S/D12L I’m not expecting a huge difference as my initial review. Direct comparison would be difficult due to the AGESA PI 1.2.0.3e that seems to have altered some performance.
I’ll continue my investigations and let you know!