Linux on Nvidia RTX Spark laptops hits 35% inference speed due to firmware lock
Benchmarks show native Linux on RTX Spark hardware runs at roughly 14 tokens per second, while Windows achieves 65 tokens per second.
Key points
- GPU clock speeds drop to 780 MHz on native Linux compared to 2.2 GHz on Windows.
- The slowdown stems from encrypted power management firmware that only Windows can read.
- WSL maintains full performance by leveraging the Windows driver stack for Linux tools.
Benchmarks published on Mind Studio show native Linux on Nvidia RTX Spark laptops achieves only 13.9 tokens per second on the GPT-OSS-120B model, whereas Windows reaches 65.4 tokens per second.reported The headline of the report describes this as slowing inference speed to 35% on Linux.reported
Evidence: reported
Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linuxaifleet-hn-front
Evidence: reported
Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linuxaifleet-hn-front
What changed
A report on Hacker News highlighted that installing Linux on RTX Spark laptops like the MSI Prestige N16 Flip AI Plus results in significant GPU throttling when connected to power.reported The GPU locks to roughly 780 MHz on Linux instead of the 2.2 GHz it reaches on Windows, for reasons tied to power management firmware that Linux currently can’t read.reported
Evidence: reported
Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linuxaifleet-hn-front
Evidence: reported
Nvidia RTX Spark encrypted firmware slows inference speed to 35% on Linuxaifleet-hn-front
How it works
The root cause is an encrypted power table managed by Nvidia’s Platform Controllers and Framework, a Windows-only component. Because Linux cannot decrypt this firmware, it cannot access the full power delivery logic between the CPU and GPU. Consequently, the open kernel driver cannot boost the GPU clocks beyond the conservative default limit.
What the evidence means
Memory allocation presents another barrier, as the laptop reserves 96GB of unified memory for the GPU by default. Only Windows exposes settings to change this split, meaning Linux users cannot easily access more RAM for model loading. WSL serves as a workaround, maintaining full GPU boost while running Linux-first AI tools like vLLM.
Cost and market position
The RTX Spark platform uses the same GPU design as Nvidia’s DGX Spark desktop unit, but laptop variants from MSI, Asus, and Dell ship with Windows on ARM. The DGX Spark ships with Nvidia’s own Linux build out of the box, whereas every laptop variant ships with Windows, creating a divergence in default operating system support.
Why it matters
- Product teams relying on native Linux for AI workflows face a significant performance penalty on RTX Spark hardware.
- The encrypted firmware lock prevents standard open-source driver optimizations, forcing reliance on WSL for full speed.
- Hardware procurement for local LLM development is constrained by OS-specific power management limitations.
- The inability to adjust memory splits in Linux restricts
Updated
Numbers in context
Circle area is proportional to the magnitude; labels retain signs. Figures may describe different measures. Read their contexts above.Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Reading receipt
- GPU clock speeds drop to 780 MHz on native Linux compared to 2.2 GHz on Windows.
- The slowdown stems from encrypted power management firmware that only Windows can read.
- WSL maintains full performance by leveraging the Windows driver stack for Linux tools.
