I conducted my own cooler comparison of three midsize Noctua air coolers in my 5.9 liter Yiluka YLK 01 case APU build with a Gigabyte X570 I Aorus Pro Wi-Fi motherboard using a Ryzen 7 Pro 4750G APU with 32 GB of Corsair Vengeance 3200 MHz LPX low profile memory with XMP enabled using a 450 Watt EVGA PSU. The case has a stated available cooler height limit of 95 mm. For maximum performance with each cooler I chose:
Noctua NH-L9x65 SE-AM4: A9x25 Fan only & then double downdraft operation adding a A12x15 fan
Noctua NH-L12 Ghost S1 Edition: push-pull downdraft A9x25 & A12x15 or A12x25 fan
Noctua NH-L12S: downdraft A12x25 single fan operation
Each cooler was configured as I would use them in my preferred configuration to maximize their performance in the YLK 01 case. Only the motherboard chipset fan was also in use. All coolers were installed with the cooler fins running from front to back in the case to maximize air flow as the right side of the case is substantially blocked by the vertically mounted PSU.
The Noctua NH-L9x65 SE-AM4 was tested first with only the Noctua A9x25 fan, and then a Noctua A12x15 fan mounted to the case top panel. The top mounted A12x15 fan was affixed with zip-ties and required 3 mm spacers between the fan and the top case panel to eliminate turbulence noise when the fan was mounted directly to the panel. Both fans were pushing air down towards the motherboard. Noctua’s NT-H1 thermal paste was used.
On the Noctua NH-L12 Ghost S1 Edition I replaced the included Noctua NF-B9 PWM with a Noctua NF-A9x25 fan, and top mounted first a Noctua NF-A12x15 and subsequently a Noctua NF-A12x25 fan for push-pull downdraft airflow. The Noctua Ghost Edition is actually 20 grams heavier than the Noctua NH-L12S - will have to see if this makes any difference in cooling performance. The Noctua NH-L12 Ghost S1 Edition came with Noctua’s NT-H2 thermal paste and this was used on both L12 coolers. Noctua sends mixed messages with this cooler - on the one hand they deliver it with their latest thermal paste, but include an older fan with less air flow and static pressure than their A9x25 fan.
The Noctua NH-L12S includes a bottom mounted Noctua NF-A12x15 fan, to which I intended to add an additional Noctua NF-A12x25 fan above the cooler for push-pull downdraft operation. The Noctua NF-A12x25 fan was not too tall to fit above this cooler, but the entire cooler was twisted laterally towards the rear of the case so that the Noctua NF-A12x15 fan would not fit underneath the cooler on this motherboard. The packaging did not show any damage upon arrival. I did not want to join the cooler-bending-to-fit club so I returned the cooler after my initial test with only a top mounted downdraft Noctua NF-A12x25 fan. The A12x25 fan on this cooler being closer to the case top panel produced a bit more noise than Ghost S1 Edition, but substantially less than the noise of a Noctua A12x15 fan mounted directly to the case top panel.
To stress the APU, I used the Blender classroom render for the CPU side and Unigine Valley with sound on for the GPU side running simultaneously. Idle temperatures, maximum temperatures for the CPU, System 1, VRM, highest wattage draw and time to complete the renders were noted.
During Blender classroom renders with Unigine Valley:
The chipset fan ran at 3,901RPM with the dual fan L9x65.
With the L12S with the top mounted A12x25 fan, the chipset fan ran at 3,552 for about one minute and then shut off.
With the L12 Ghost Edition with either the top mounted A12x15 or A12x25 fan, the chipset fan never came on.
Noctua NH-L9x65 SE-AM4: A9x25 Fan only & then double downdraft operation adding a A12x15 fan
Noctua NH-L12 Ghost S1 Edition: push-pull downdraft A9x25 & A12x15 or A12x25 fan
Noctua NH-L12S: downdraft A12x25 single fan operation
Each cooler was configured as I would use them in my preferred configuration to maximize their performance in the YLK 01 case. Only the motherboard chipset fan was also in use. All coolers were installed with the cooler fins running from front to back in the case to maximize air flow as the right side of the case is substantially blocked by the vertically mounted PSU.
The Noctua NH-L9x65 SE-AM4 was tested first with only the Noctua A9x25 fan, and then a Noctua A12x15 fan mounted to the case top panel. The top mounted A12x15 fan was affixed with zip-ties and required 3 mm spacers between the fan and the top case panel to eliminate turbulence noise when the fan was mounted directly to the panel. Both fans were pushing air down towards the motherboard. Noctua’s NT-H1 thermal paste was used.
On the Noctua NH-L12 Ghost S1 Edition I replaced the included Noctua NF-B9 PWM with a Noctua NF-A9x25 fan, and top mounted first a Noctua NF-A12x15 and subsequently a Noctua NF-A12x25 fan for push-pull downdraft airflow. The Noctua Ghost Edition is actually 20 grams heavier than the Noctua NH-L12S - will have to see if this makes any difference in cooling performance. The Noctua NH-L12 Ghost S1 Edition came with Noctua’s NT-H2 thermal paste and this was used on both L12 coolers. Noctua sends mixed messages with this cooler - on the one hand they deliver it with their latest thermal paste, but include an older fan with less air flow and static pressure than their A9x25 fan.
The Noctua NH-L12S includes a bottom mounted Noctua NF-A12x15 fan, to which I intended to add an additional Noctua NF-A12x25 fan above the cooler for push-pull downdraft operation. The Noctua NF-A12x25 fan was not too tall to fit above this cooler, but the entire cooler was twisted laterally towards the rear of the case so that the Noctua NF-A12x15 fan would not fit underneath the cooler on this motherboard. The packaging did not show any damage upon arrival. I did not want to join the cooler-bending-to-fit club so I returned the cooler after my initial test with only a top mounted downdraft Noctua NF-A12x25 fan. The A12x25 fan on this cooler being closer to the case top panel produced a bit more noise than Ghost S1 Edition, but substantially less than the noise of a Noctua A12x15 fan mounted directly to the case top panel.
To stress the APU, I used the Blender classroom render for the CPU side and Unigine Valley with sound on for the GPU side running simultaneously. Idle temperatures, maximum temperatures for the CPU, System 1, VRM, highest wattage draw and time to complete the renders were noted.
| Blender Classroom Render & Unigine Valley YLK-01 in CPUID HWMonitor | |||||
| @21C | Render Time | Wattage | Sys 1 | CPU | VRM |
| L9x65 A9x25 Only | |||||
| BCRR & UV | / 10:34.8 | 89.06 | 51 | 81 | 51 |
| L9x65 A9x25 & A12x15 Idle | 43 | 27 | 30 | ||
| BCRR & UV | / 10:34.38 | 88.76 | 46 | 77 | 44 |
| L12S A12x25 Top Down Idle | 43 | 27 | 32 | ||
| BCRR & UV | / 10:26.23 | 89.21 | 44 | 69 | 43 |
| L12 GE A9x25 & A12x15 Idle | 39 | 26 | 30 | ||
| BCRR & UV | / 10:21.93 | 89.1 | 44 | 71 | 43 |
| L12 GE A9x25 & A12x25 Idle | 38 | 26 | 30 | ||
| BCRR & UV | / 10:21.58 | 89.04 | 40 | 67 | 40 |
During Blender classroom renders with Unigine Valley:
The chipset fan ran at 3,901RPM with the dual fan L9x65.
With the L12S with the top mounted A12x25 fan, the chipset fan ran at 3,552 for about one minute and then shut off.
With the L12 Ghost Edition with either the top mounted A12x15 or A12x25 fan, the chipset fan never came on.
Last edited:
