TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
A MindStudio report found that native Linux on an MSI Prestige RTX Spark laptop generated 13.9 tokens per second on GPT-OSS-120B, compared with 65.4 on Windows and 65.1 in WSL. The report attributes the gap to GPU clocks staying near 790 MHz on Linux while plugged in, linked to power-management firmware Linux cannot currently read. The results come from one reported laptop setup; Nvidia has not provided a response in the supplied material.
A MindStudio report found that native Linux on an MSI Prestige RTX Spark laptop produced 13.9 tokens per second on GPT-OSS-120B, far below the 65.4 tokens per second recorded on Windows. The report links the gap to GPU clocks staying near 790 MHz on Linux while the laptop is plugged in, which it attributes to power-management firmware unavailable to Linux.
In MindStudio’s reported test, Windows reached 65.4 tokens per second on GPT-OSS-120B, while Windows Subsystem for Linux (WSL) reached 65.1. The report said Nvidia’s DGX Spark desktop recorded 61 tokens per second, compared with 13.9 on native Linux on the laptop. These are results from the report’s setup, not independently verified results or a broader product-wide benchmark.
The report said the laptop GPU ran at roughly 2.2 GHz on Windows but around 790 MHz under native Linux while plugged in. A longer run reportedly took about eight minutes on Windows and 37 minutes on Linux. MindStudio said temperatures remained between 40 and 47 degrees Celsius, and argued that the performance limit was not caused by thermal throttling.
Linux did run after configuration, with a desktop, CUDA acceleration, and working display, audio, keyboard, and touchpad. But installation required hardware-specific workarounds, and the report said the laptop’s memory allocation also had to be changed in Windows before Linux could access most of its unified memory. WSL avoided the reported clock limitation by using the Windows driver stack.
Linux Performance Lags Windows
The results matter to developers considering these laptops for local AI workloads. The RTX Spark systems are marketed around GPU acceleration and large unified-memory configurations, while common AI development tools such as vLLM and SGLang are designed primarily for Linux. If the reported clock behavior applies more broadly, installing Linux directly could sharply reduce inference throughput, even when GPU acceleration is available.
For now, the report presents WSL as a practical workaround: it allows Linux-oriented tools to run while retaining the Windows driver and power-management stack. That is not the same as native Linux, and it may not suit every workflow, but in the cited tests its throughput was close to Windows. Buyers who need native Linux should treat the report as a warning to test their exact laptop, firmware, and workload rather than assume that CUDA support guarantees comparable performance.
Popular right now
KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
Ninja Foodi 6-in-1 DualZone Air Fryer, 8 qt, Grey (DZ201)
As an affiliate, we earn on qualifying purchases.
Firmware and Setup Constraints
The RTX Spark laptop platform, found in systems including the MSI Prestige, Asus ProArt, and Dell XPS, pairs an Nvidia GPU with an ARM-based Windows system and shared CPU-GPU memory. MindStudio contrasted those laptops with the DGX Spark desktop, which ships with Nvidia’s Linux environment. That difference in software support is central to the reported performance disparity.
MindStudio said a Windows-only component, the Nvidia Platform Controllers and Framework, manages power delivery between the CPU and GPU. Its report says the relevant power table is encrypted and that the Windows driver can decrypt it, while Linux lacks an equivalent implementation. The article also describes a separate memory issue: the laptop initially reserved 96GB of its 128GB unified memory for the GPU, and the allocation could be adjusted only in Windows. These explanations are the report’s account; the supplied material does not include Nvidia documentation or a company response confirming them.
“Windows produced 65.4 tokens per second. WSL … hit 65.1 tokens per second. … Native Linux on the laptop … dropped to 13.9.”
— MindStudio report
Scope of the Benchmark
The supplied report does not establish whether the same clock behavior affects every RTX Spark laptop model, Linux distribution, firmware release, or power configuration. Its figures describe a particular MSI Prestige test setup, and the material does not provide independently reproduced benchmarks or Nvidia’s response.
The report also says the direction of the performance issue changes when the laptop is unplugged, but the supplied text cuts off before giving the unplugged results. It is therefore unclear how Linux compares with Windows on battery, or whether firmware or driver updates have since changed the behavior. The exact technical mechanism behind the encrypted power table is described by MindStudio but is not corroborated by a separate source here.
Nvidia Driver Support Needed
The report says a lasting fix would require Nvidia to add the platform power-management logic to its Linux driver or provide another Linux-compatible way to access the power settings. No such fix or release date is identified in the supplied material. Until Nvidia or laptop makers address the issue, users seeking Linux tools can consider WSL, while recognizing that it runs on Windows rather than as a native Linux installation.
Further testing across laptop models, Linux distributions, firmware versions, and both plugged-in and battery use would help establish how widespread the limitation is. Prospective buyers should check for current driver and firmware updates and verify performance on their intended workload before relying on the reported numbers as representative of every RTX Spark system.
Key Questions
How much slower was native Linux in the reported test?
MindStudio measured 13.9 tokens per second on native Linux with GPT-OSS-120B, compared with 65.4 on Windows and 65.1 in WSL on the same reported laptop setup.
What does the report say is causing the slowdown?
MindStudio attributes it to GPU clocks staying near 790 MHz on plugged-in Linux, compared with roughly 2.2 GHz on Windows. It links that difference to power-management firmware and a Windows-only controller component, but the supplied material does not include Nvidia confirmation.
Does GPU acceleration work on Linux?
Yes, according to the report, after configuration. MindStudio said the laptop ran Linux with CUDA acceleration, but reported that GPU clock speeds and inference throughput remained far below Windows in its plugged-in test.
Is WSL a workaround?
In the reported tests, WSL reached 65.1 tokens per second, close to Windows at 65.4. WSL runs a Linux environment on top of Windows and uses the Windows driver stack, so it is not a native Linux installation.
Has Nvidia confirmed a fix?
No Nvidia response or confirmed fix appears in the supplied source material. MindStudio says a fix would require Linux support for the platform’s power-management logic, but gives no release timeline.
Source: mindstudio.ai
