Why the best reason to care about Nvidia’s RTX Spark is that everything can run on Arm

A familiar Windows problem

Fourteen years ago, Microsoft released a Windows PC powered by Nvidia’s Tegra Arm chip. The Surface ran a stripped-down version of Windows, and Microsoft later wrote off $900 million in inventory.

The key limitation was software compatibility: the chip could not run existing Windows apps, leaving buyers without their legacy programs. Nvidia now says its RTX Spark laptops will handle apps and games, including software with kernel-level anti-cheat.

Windows 11 on Spark

The first wave of RTX Spark laptops is due Friday, October 16th. The RTX Spark superchip is the same chip Nvidia put in mini desktops for AI developers a year earlier, but those systems ran Linux; the laptops run Windows 11. Microsoft’s Prism emulator enables older Windows apps and games to run on the Arm chip. Nvidia senior director of products Justin Walker told journalists, “All your apps, all your games, everything’s going to run great on Spark.”

Compatibility still has hurdles

Other companies have also brought software to platforms it was not originally designed for. Apple shifted its computers from Intel x86 to Arm-based chips beginning in 2020, using Rosetta 2; Valve’s SteamOS demonstrated Windows games on Linux and Arm in 2022.

Qualcomm’s first well-received Windows on Arm machines arrived in 2024, including the Microsoft Surface Laptop 7th Edition. But some games failed or stuttered, and major multiplayer titles such as Fortnite did not run on Qualcomm or SteamOS devices because of anti-cheat incompatibility.

A broader case than AI

The writer says an RTX Spark laptop has substantially more graphics power than a Steam Deck or Qualcomm Snapdragon X Elite, comparable to a midrange discrete laptop GPU such as an RTX 5060. A reviewer had used one for several days and found it mostly indistinguishable from a traditional Windows machine; more detailed early impressions were expected after the embargo lifted.

The laptops are also pitched as local AI workstations, but prices can reach $7,000 and the writer questions whether buyers will accept the tradeoffs. Local token generation is much slower than AI in the cloud, the article notes, while broader Windows and game compatibility may be the more compelling promise.