Nvidia RTX Spark Encrypted Firmware Slows Inference Speed To 35% On Linux
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

MindStudio reported that an MSI Prestige RTX Spark laptop running native Linux generated 13.9 tokens per second on GPT-OSS-120B, compared with 65.4 on Windows and 65.1 in WSL. The report links the gap to Linux’s inability to access a power-management firmware table that keeps the GPU near 790 MHz on wall power; Nvidia has not provided a response or confirmed a fix in the supplied material.

A MindStudio report says native Linux ran GPT-OSS-120B inference on an MSI Prestige RTX Spark laptop at 13.9 tokens per second, compared with 65.4 on Windows and 65.1 in Windows Subsystem for Linux (WSL). The report attributes the gap to Linux’s limited access to the laptop’s power-management firmware, which it says leaves the GPU running at about 790 MHz on wall power rather than roughly 2.2 GHz under Windows.

In MindStudio’s benchmark, the laptop’s Windows setup generated 65.4 tokens per second on GPT-OSS-120B. WSL reached 65.1 tokens per second, while an Nvidia DGX Spark desktop running Nvidia’s Linux build reached 61. Native Linux on the laptop managed 13.9 tokens per second, about one-fifth of the Windows result. These figures are results reported by MindStudio, not independent verification.

The report says the slowdown was also visible in a longer run: Windows completed the workload in about eight minutes, with the GPU near 2.2 GHz, while Linux took about 37 minutes with the GPU around 790 MHz. Temperatures reportedly stayed between 40 and 47°C, leading the report to identify power limits, rather than heat, as the likely explanation. In a separate vLLM test with eight concurrent clients, MindStudio reported 127 tokens per second in WSL and 39.5 on native Linux.

Linux did run on the laptop after configuration and driver work. MindStudio reports that the system supported a desktop at 4K and 120Hz, audio, keyboard and touchpad, as well as CUDA acceleration. But it also describes installation and memory-allocation complications: the laptop initially exposed only 28GB of its 128GB unified memory to Linux, and the memory split had to be changed in Windows before Linux could see more.

At a glance
reportWhen: Reported in MindStudio’s article; publi…
The developmentA MindStudio benchmark report found native Linux inference on an MSI Prestige RTX Spark laptop ran about five times slower than Windows, with GPU clock speeds held far below Windows levels.

Linux Performance Falls Far Behind Windows

The results matter to developers considering RTX Spark laptops for local AI workloads. The hardware is marketed for AI development and offers up to 128GB of unified memory, while commonly used tools such as vLLM and SGLang are designed around Linux. But the reported performance difference suggests that installing Linux directly may prevent users from getting the inference speeds seen under Windows, at least on the tested configuration.

For now, MindStudio presents WSL as a practical workaround: it runs Linux tools within Windows and, in the reported tests, retained nearly the same inference rate as Windows itself. That may let developers use Linux-oriented software without giving up the Windows driver and power-management stack. It is not the same as running Linux directly, and the supplied report does not establish whether results generalize to every RTX Spark laptop, workload, or software version.

Products Worth Considering

Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, 12MP Center Stage Camera, and Wi-Fi 7, Sky Blue

Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, 12MP Center Stage Camera, and Wi-Fi 7, Sky Blue

Apple MacBook Air 15-inch (2025) with M4 Chip, 24GB Memory, 512GB SSD, Midnight

Apple MacBook Air 15-inch (2025) with M4 Chip, 24GB Memory, 512GB SSD, Midnight

Samsung Galaxy Tab A11+ 11-Inch Wi-Fi Tablet, 6GB RAM, 128GB Storage, Gray

Samsung Galaxy Tab A11+ 11-Inch Wi-Fi Tablet, 6GB RAM, 128GB Storage, Gray

As an affiliate, we earn on qualifying purchases.

Firmware And Linux Setup Explain The Gap

RTX Spark laptops, including models from MSI, Asus and Dell, pair Nvidia GPU technology with Windows on ARM and unified memory shared by the CPU and GPU. Nvidia’s DGX Spark desktop, by comparison, ships with Nvidia’s own Linux build. That distinction matters for users who want to run Linux-first AI software directly on a laptop rather than inside Windows.

MindStudio identifies a Windows-only component called the Nvidia Platform Controllers and Framework as the source of the power-management difference. According to the report, its power table is encrypted and can be decrypted only by the Windows driver; Linux currently lacks an equivalent component. The report says the laptop’s GPU therefore runs at about 780–790 MHz on Linux when plugged in, compared with roughly 2.2 GHz on Windows. This is the report’s explanation, and Nvidia’s confirmation is not included in the source material.

The Linux installation itself was not straightforward in the tested MSI machine. MindStudio says an initial Omarchy attempt produced an unstable display, while Debian required workarounds for hardware support and network access. The account says Omarchy’s bundled Nvidia open-source driver later enabled a working desktop and GPU acceleration. Adjusting the memory reservation from Windows was also necessary: the laptop initially reserved 96GB for the GPU, leaving Linux able to see only 28GB of the 128GB total.

“Windows produced 65.4 tokens per second. WSL hit 65.1 tokens per second. Native Linux on the laptop managed only 13.9.”

— MindStudio report

Nvidia Response And Broader Results Unknown

The supplied source is a single MindStudio report and does not include an Nvidia response, independent replication, or confirmation that the encrypted firmware table is the definitive cause. It also does not provide a publication date, detailed benchmark methodology, or the exact software and driver versions needed to judge how the results might change with updates.

The report’s headline frames Linux inference as falling to 35% of Windows performance, but its stated GPT-OSS-120B figures—13.9 versus 65.4 tokens per second—are about 21% by direct calculation. The same source describes other tests and power conditions, but the supplied material cuts off while discussing unplugged performance. The precise basis for the 35% figure and the full battery-powered results are therefore unclear. It is also unknown whether other RTX Spark laptop models behave the same way.

Linux Driver Support Is The Key Milestone

MindStudio says a durable fix would require Nvidia to bring the platform controller’s power-management logic into its Linux driver. The source does not report that Nvidia has announced such work or give a timeline for a driver or firmware update. Until one is available, users weighing a native Linux installation may need to compare its reported performance and setup limits with running Linux tools through WSL.

Further testing could clarify whether the results apply across RTX Spark models, Linux distributions, driver versions, and workloads, and whether performance changes when the laptop is unplugged. For now, the published figures describe one reported MSI Prestige configuration; broader conclusions depend on confirmation from Nvidia and independent benchmarks.

Key Questions

How fast was native Linux in the reported benchmark?

MindStudio reported 13.9 tokens per second on GPT-OSS-120B using native Linux on an MSI Prestige RTX Spark laptop. Its reported Windows result was 65.4 tokens per second.

Did WSL also run slowly?

No. In MindStudio’s test, WSL reached 65.1 tokens per second, close to the 65.4 reported for Windows. WSL uses the Windows driver stack while providing a Linux environment.

What does the report say causes the slowdown?

MindStudio links it to power-management firmware that Linux cannot access through an equivalent component. The report says the GPU remained near 790 MHz on wall power under Linux, versus roughly 2.2 GHz under Windows. Nvidia has not confirmed that explanation in the supplied material.

Can Linux run on the RTX Spark laptop?

According to MindStudio, Linux can run after configuration and driver work, with a desktop, CUDA acceleration, and key peripherals functioning on the tested MSI system. The report also describes installation obstacles and says a memory reservation setting had to be changed in Windows before Linux could access most of the laptop’s unified memory.

Has Nvidia announced a fix?

The supplied report does not cite an Nvidia announcement or a release date for a fix. MindStudio says a solution would require Nvidia to add the relevant platform power-management logic to its Linux driver.

Source: mindstudio.ai

You May Also Like

How NVIDIA Kumo Tabular Balances Accuracy And Efficiency In Prediction

NVIDIA has released Kumo Tabular, an open model for table-based predictions. Its benchmark claims lack scores and independent validation in the supplied material.

Your ‘App’ Could Have Been A Webpage (So I Fixed It For You)

Tech experts reveal many mobile apps could be optimized as webpages, offering better performance and accessibility. Here’s what happened.

DevDay 2026: A Recap Of The AI Highlights

OpenAI has a page titled “DevDay 2026 Recap,” but the available information does not confirm event details or announcements.

You Can Defeat The Dream Devourer From Chrono Trigger Using An Int Overflow

A new exploit allows players to defeat the Dream Devourer using an integer overflow, sparking interest in the game’s community and modding scene.