Today's post is going to be about the thermal and noise performance you can expect from an Air Cooled Winter One with perforated panels. Before that, here’s a quick update on manufacturing.
Quick Manufacturing Update
A handful of parts in full production are rolling off the line, and have gone through 2 rounds of QC thus far. They should make it through a third and final round of QC by the end of this week.
With Lunar New Year holidays right around the corner, it’ll be quiet for a week or two, as it is a holiday for most of our manufacturing partners. Full scale production of Batch 1 will resume at each site after the holiday.
A packaging supplier reached out to me, confirming their timeline for March and their ability to deliver.
Thermal and Noise Testing
Because there wasn’t as much thermal and noise data on the perforated panels or air cooled builds, I thought I would put my air cooled Winter One through its paces, and share the process and results with you.
CPU: Ryzen 5950X
Motherboard: Gigabyte B550i Aorus Pro AX
CPU Cooler: Noctua NH-L12S
GPU: RTX 3090FE (stock air cooler)
Memory: 64GB Crucial Ballistix 3600 CL16
Power Supply: Corsair SF750 Platinum
Case Fans: 4 x Noctua NFA14 PWM fans
Front M.2: WD SN850 2TB (using motherboard heatsink)
Rear M.2: WD SN750 2TB (no heatsink)
Case: Winter One V12 - Perforated Panels
Airflow Setup: top/bottom exhaust, side panel intake
Fan Curve: Linear, from 20% at 20C to 100% at 80C, based on the highest temp (either CPU or GPU core)
Ambient Temp: 21ºC
The thermal load used for this test was 5 minutes of OCCT’s power stress test, which includes a GPU + CPU burn test (using SSE / CUDA). I ran the tests for 5 minutes - enough to reach steady state in my build). For data logging, I used HWInfo64. The above protocol and specifications stayed the same for both tests.
For noise measurement, two microphones were placed at a 1ft (12inch / 32cm) distance from the side panel of the case. The UMIK-1 is a calibrated measurement mic, with dBA readings displayed on a laptop (taken using REW). The second device is my phone, used to measure frequency responses during the noise tests. The room is not treated, and noise floor was 28dBA.
Stock Thermals and Noise
5950X: Stock frequencies, no PBO, 120W power limit
3090FE: Stock frequencies, no UV, 350W power limit
Total Heat to Dissipate: 470W
At stock settings Winter One idles at 36ºC on the CPU, and 40ºC on the GPU. Noise levels are 35dBa at a distance of 30cm / 1 foot, at idle.
Idle Noise at Stock Settings - 35dBA
Idle Temperatures, Stock Settings
Under Steady State Load, Winter One is hitting 65ºC on the CPU and 65ºC on the GPU, with all parts at stock settings. These are incredibly tame thermals for an SFF sandwich case. Noise levels are around 45dBa. There are also no high frequency spikes in the noise profile of the fans.
Steady State Temperatures under load, Stock Settings
Overclocked Thermals and Noise
5950X: Stock + PBO-2, VF Curve Editor (-10 offset), 250W peak power limit
3090FE: Custom VF Curve with 2025Mhz @ 950mV, 400W power limit
Total Heat to Dissipate: 625W initially, 575W steady state.
Overclocked, Winter One idles at 36ºC on the CPU, and 36ºC on the GPU. Noise levels remain at 35dBA, at a distance of 30cm / 1 foot.
Idle Temperatures, Overclocked
At the ramp-up of the test, the 5950X with PBO-2 pulls an insane 242W, and hits 90ºC almost immediately. It then starts to lower its power limit to maintain 90ºC die temperatures.
Ramp up and then dial back of power target on the 5950X
At steady state, the 5950X is still running at 90ºC, but thanks to Winter One’s airflow optimization, the L12S Noctua Cooler is managing to dissipate an impressive 170W of thermal load.(For context this a mere 20 W lower than the 190W dissipated by my NH-D14 tower cooler on an open air test bench, with the same 5950X!). Our overclocked 3090FE is managing 67ºC while pulling 400W.
Steady State Thermals, Overclocked.
Noise levels when overclocked are 53dBa. Even at 54 dBa, and 1500 RPM on the case fans, the noise profile remains free of high frequency spikes that would cause annoying whine.
Front / Rear M.2 Drive Temperatures
I'll focus on the rear M.2 temperatures, as this drive does not have a heat sink, and must rely on airflow for cooling, while being sandwiched behind the motherboard, and adjacent to the (toasty) backplate on the 3090FE. The rear M.2 represents the worst case scenario for an SSD in Winter One. It's also good to keep in mind that a typical M.2 SSD’s will thermal throttle at 89ºC. For this test, I copied my steam library from one drive to the other, and back.
Rear M.2 500GB Write Temps: 68ºC
Rear M.2 500GB Read Temps: 56ºC
Rear M.2 Temperatures, Reading, 500GB copy
At this point, it's safe to say that Winter One has achieved it's goal of being an exceptionally well-cooled SFFPC case. Thermals and noise only get better as you use an AIO, or set up a custom loop. Winter One's design philosophy represents a huge leap forward in cooling capability of SFFPCs.
In the next update, we'll take a look at:
Airflow, v12 simulations
First Release of the v12 User’s Guide / Manual
Open the floor to submitting questions for the next Q&A
To see all images from this post, visit the Kickstarter.