Our test setup focuses on four components small enough to fit into every chassis format. This allows us to compare the impact case size and cooling have on a modern system. Each of the four measured components is connected to a REED Instruments SD-947 Data Logging Thermometer with K-Type sensors that have been placed as closely as possible to the benchmarked areas of the CPU, GPU, M.2 SSD, and motherboard chipset. A fifth sensor is used to monitor the room temperature for an accurate delta T result. While the data provides us an accuracy of one decimal point, it is still dependent on human placement of the sensor as well as small variance in things like automatic disk activity in Windows for example. The systems are air-gapped, so that no automatic software updates can trigger an unwanted workload on any of the components.
For the cases, the out-of-the-box fan setup is used, connected to the motherboard, with the board adjusting fan speed automatically. This puts the onus of noise and performance on the case manufacturers and motivates them to actually care about thermal performance instead of simply using active cooling as a marketing bulletin point.
On the software side, the test setup includes Windows 10 Pro with Furmark and Prime95 both running at the same time to put the maximum load on both the processor and the graphics card.
We let the chassis sit at idle until temperatures reach equilibrium, but never for less than 30 minutes. Next, our stress-testing load is started, and it continues to run until temperatures are stable. The temperature result for each section of the case is the highest value measured during the test run. All reported temperatures are normalized to 20°C room temperature.