Summary
Alex Ziskind spends three days trying to run Linux on an RTX Spark laptop, the MSI Prestige N16 Flip AI Plus with Nvidia’s N1X chip, which shares its GPU design with the DGX Spark but ships with Windows on ARM. The effort involved a JetKVM, a MacBook Air camera pointed at the screen, Claude Code driving the process remotely, blank screens, an overnight freeze and a kernel panic.
The video starts with earlier results running GPT-OSS 120B: 65.4 tokens per second on native Windows, 65.1 in WSL and 61 on the DGX Spark. He then works through distribution attempts. Omarchy produced a flickering screen on his MSI model, Debian lacked drivers for the keyboard, touchpad and GPU, and a customized Omarchy kernel with Nvidia’s open driver finally helped.
A key discovery was that the laptop reserves 96 GB of its 128 GB unified memory for the GPU, a setting only Windows can change. After reducing it to 4 GB, Linux saw 120 GB and ran a 4K 120 Hz desktop with CUDA. LM Studio’s ARM64 Linux build reached about 61 tokens per second unplugged but only about 15 when plugged in, a surprising result the video investigates.
📺 Source: Alex Ziskind · Published October 31, 2026
🏷️ Format: Hands On Build
