Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linux
Compiled by KHAO Editorial — aggregated from 1 source. See llms.txt for citation guidance.
◌ Single Source
Yes, but it comes with a serious catch.
Key facts
- That turns a chip capable of around 65 tokens per second on a 120 billion parameter model into one struggling to break 14 tokens per second, for reasons tied to power management firmware that Linux
- Disabling that controller in Windows reproduced the slowdown directly: tokens per second dropped from around 83 to about 31, and clock speed fell from roughly 2 GHz down to the same 786 MHz seen
- The gap showed up again with vLLM running eight concurrent clients: WSL reached 127 tokens per second while native Linux reached 39.5
- Windows finished in about 8 minutes while holding the GPU near 2.2 GHz
Summary
Dual-booting Linux works on the MSI Prestige RTX Spark laptop, with a full desktop, GPU acceleration via Nvidia’s open driver, 4K output at 120Hz, and working audio, keyboard, and touchpad once configured correctly. A hidden Windows-only component called the Nvidia Platform Controllers and Framework manages power delivery between CPU and GPU, and it has no Linux equivalent, which is the root cause of the slowdown. GPU clock speed collapses to about 780 MHz on Linux when plugged in, compared to roughly 2.2 GHz on Windows, because the firmware’s power table is encrypted and only the Windows driver can decrypt it. Benchmarks on GPT-OSS-120B show the gap clearly: Windows hit 65.4 tokens per second, WSL hit 65.1, the Nvidia DGX Spark hit 61, and native Linux on the same laptop managed only 13.9.