TL;DR
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
MindStudio reported 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 linked the gap to GPU clocks held near 780 MHz under Linux and a Windows-only power-management component; Nvidia has not provided a response in the supplied material.
A MindStudio report found that native Linux inference on an MSI Prestige RTX Spark laptop ran at 13.9 tokens per second on GPT-OSS-120B, far below the 65.4 tokens per second measured in Windows on the same laptop. The report attributes the slowdown to Linux being unable to access a Windows-only power-management component that allows the GPU to reach higher clock speeds, a problem relevant to developers considering these AI-focused laptops for Linux workloads.
In the reported test, the laptop’s GPU ran at about 780 MHz under native Linux, compared with roughly 2.2 GHz under Windows while plugged into wall power. MindStudio said the GPU’s temperature remained between 40 and 47°C during testing, pointing away from thermal throttling as the cause in those runs. The measurements are the report’s results; the supplied material does not include an independent replication.
The same comparison showed 65.1 tokens per second in WSL, which runs Linux within Windows, and 61 tokens per second on Nvidia’s DGX Spark desktop. In a separate vLLM test with eight concurrent clients, MindStudio reported 127 tokens per second in WSL and 39.5 under native Linux. It also said a longer run took about eight minutes on Windows and 37 minutes on Linux.
The Linux installation itself was possible but required configuration. According to the report, the tester used a Linux kernel with Nvidia’s open-source driver, adjusted the laptop’s GPU memory reservation in Windows, and then obtained a working desktop, CUDA acceleration, and peripheral support. Without changing that reservation, Linux saw only 28GB of the laptop’s 128GB unified memory; after the Windows-side setting was reduced, Linux could access 120GB, the report said.
Linux Costs Developers GPU Throughput
The results suggest that Linux compatibility does not guarantee Windows-level GPU performance on this laptop. That distinction matters to AI developers: tools such as vLLM and SGLang are commonly used in Linux environments, while the RTX Spark laptops are marketed for workloads that benefit from substantial shared CPU and GPU memory.
For users who need Linux-based tooling today, the report presents WSL as a practical workaround: its tested throughput was close to Windows because it uses the Windows driver stack. But that is not the same as running Linux directly on the machine. The reported native-Linux results also leave open whether the limitation affects other RTX Spark laptop models or configurations to the same extent.
Products Worth Considering
Samsung Galaxy Tab A11+ 11-Inch Tablet, 6GB RAM, 128GB Storage, Expandable, Gray
Apple MacBook Air 15-inch with M5 chip, 16GB Memory, 512GB SSD, and Liquid Retina Display
Samsung Galaxy Tab S10 Lite – 6 GB RAM, 128 GB storage, up to 2 TB microSD, 10.9‑inch LCD, Gray
As an affiliate, we earn on qualifying purchases.
Windows Controls the Power Path
The RTX Spark laptop platform appears in systems including the MSI Prestige, Asus ProArt and Dell XPS, according to the source report. These machines use Nvidia GPU technology and offer unified memory configurations of up to 128GB. Unlike Nvidia’s DGX Spark desktop, which the report describes as shipping with Nvidia’s Linux build, the laptop models ship with Windows on ARM.
MindStudio identifies Nvidia Platform Controllers and Framework as a Windows-only component that manages power delivery between the CPU and GPU. It says the power table is encrypted and that the Windows driver can decrypt it, while Linux currently lacks an equivalent path. This is the report’s explanation for the lower Linux clock speeds, not a statement from Nvidia in the supplied material.
Setting up Linux also involved hardware-specific obstacles, including installer network and peripheral support, according to the account. The laptop’s memory allocation posed a separate issue: the report says Windows exposes the setting that changes how much unified memory is reserved for the GPU. These setup limitations are distinct from the clock-speed issue, though both affect the experience of using the laptop with Linux.
““Windows produced 65.4 tokens per second” on GPT-OSS-120B, while “native Linux on the laptop” managed 13.9.”
— MindStudio report
Nvidia Response and Scope Remain Unknown
The supplied material does not include comment from Nvidia, a company confirmation of the firmware explanation, or details of a planned Linux driver update. It also does not provide the report’s publication date, full benchmark methodology, or enough information to establish whether other RTX Spark laptop models show the same clock behavior.
The source text begins to describe results when the laptop is unplugged but cuts off before giving them. The effect of battery power on Linux performance is therefore unclear from the available account. The reported figures should be treated as results from the described hardware and tests, not as a universal performance estimate for every Linux setup.
A Linux Driver Fix Is the Key
According to MindStudio, resolving the power limit would require Nvidia to bring the platform controller’s power-management logic into its Linux driver. The source does not say Nvidia has committed to doing so or give a release schedule. Until a fix or further testing is reported, buyers considering Linux use can compare their needs against the report’s results and weigh native Linux against WSL.
Further tests across other RTX Spark laptop models, Linux distributions, driver versions and power conditions would help show how broad the issue is. For now, the established point is narrower: MindStudio measured much lower native-Linux throughput on one MSI laptop configuration and linked it to GPU power management that Linux could not access.
Key Questions
How fast was native Linux in the reported test?
MindStudio measured 13.9 tokens per second on GPT-OSS-120B using native Linux on an MSI Prestige RTX Spark laptop.
How did Windows and WSL compare?
The report measured 65.4 tokens per second in Windows and 65.1 in WSL on the same model and laptop, compared with 13.9 under native Linux.
What does the report say causes the slowdown?
It attributes the gap to GPU clocks staying near 780 MHz on Linux when plugged in, and to a Windows-only Nvidia power-management component that Linux cannot currently access. Nvidia has not confirmed that explanation in the supplied material.
Can Linux run on the laptop?
Yes, according to the report, but installation and configuration required workarounds. The tester reported a working desktop and CUDA acceleration after adjusting the memory reservation in Windows and using a compatible Linux kernel and drivers.
Is there a confirmed fix?
No fix or release schedule is given in the source material. MindStudio says a remedy would depend on Nvidia adding the platform power-management logic to its Linux driver.
Source: mindstudio.ai
