Running Linux on an Nvidia RTX Spark laptop works, but according to a hands-on report from mindstudio.ai, it comes with a major performance penalty. Once the machine is plugged into wall power, the GPU gets locked to roughly 780 MHz instead of the 2 GHz-plus it reaches on Windows.
That gap is not cosmetic. A chip that pushes around 65 tokens per second on a 120 billion parameter model on Windows struggles to break 14 tokens per second on Linux — roughly 35% of the expected speed. The cause is tied to power management firmware that Linux currently cannot read.
Nvidia RTX Spark Linux GPU clock bug explained
The RTX Spark platform ships in laptops such as the MSI Prestige N16 Flip AI Plus, the Asus ProArt and the Dell XPS, and uses the same GPU design as Nvidia's DGX Spark desktop unit. The key difference: the DGX Spark ships with Nvidia's own Linux build out of the box, while the laptop variants ship with Windows on ARM.
Getting Linux onto an MSI Prestige was not plug-and-play. A first attempt with Omarchy, an Arch-based distribution, failed because the MSI hardware behaved differently from Asus and Dell units, producing a flickering, unstable display during boot. Debian proved more reliable, but the installer could not connect to Wi-Fi, so a USB Ethernet adapter was required. Out of the box, Debian also failed to recognize the keyboard, touchpad or GPU, since vanilla kernels carry no laptop-specific hardware support.
Memory split and driver workarounds on RTX Spark
Switching back to Omarchy solved the driver problem, because its kernel already bundles Nvidia's open-source driver. Linux then booted properly, but could only see 28GB of the laptop's 128GB of unified memory. The firmware reserves 96GB for GPU use, and that split can only be adjusted from Windows.
These machines support unified memory configurations up to 128GB shared between CPU and GPU — unusual for a laptop and central to why they are marketed at AI developers. Windows on ARM with CUDA support is also new: ARM-based Windows machines have existed for years, but pairing that architecture with Nvidia's CUDA stack only became possible with this chip.
