Real machines, recorded methods
The Lab
Original hardware work with visible evidence status, explicit limitations and no invented benchmark results.
USB-C Adapter Write-Penalty Test
Same SSD, hash-verified: reads hit 539.1 MiB/s through a USB-C to USB-A adapter while writes stayed near 39 MiB/s. Why read-only benchmarks miss it.
What 135 Measurements Actually Show
The cross-dataset readout: 12 W idle, 623 MiB/s writes, the USB port lottery and the zero-failure stability run — every value sourced.
Docker Workload Benchmark
Corrected Nginx concurrency, PostgreSQL, Redis and mixed-load results with failed preliminary runs excluded.
12.57-Hour Stability Test
150 samples, zero HTTP or container-health failures, zero restarts and application-aware recovery evidence.
Jellyfin Intel QSV Lab
Native Jellyfin proof, codec matrix, two- and four-workload results, playback observations and wall power.
Core i3-1215U QSV Codec Support
Measured H.264, HEVC, VP9 and AV1 decode paths, QSV encode results and the unsupported AV1 encode boundary.
EQi12 Power Consumption
Whole-system startup, Windows idle, Docker idle, CPU load, QSV, S3 sleep and shutdown readings.
Ubuntu Live Boot & Linux Compatibility
UEFI USB boot test on stock EQi12 with real lspci/lsblk output: WD SN540 NVMe, Realtek Ethernet, Intel AX200 Wi-Fi all detected.
Network and SMB Lab
Both physical Ethernet ports, Wi-Fi 6, 50GiB SMB, 10,000 small files and integrity checks.
Storage and USB Lab
SMART, sustained internal SSD performance, four USB-A ports, readback hashes and one repeatable slow path.
Mini PC Test Method
How evidence is classified, failed tests are retained and unknown capabilities remain unknown.