AI-rewritten: This is a summary of an article from Tom’s Hardware, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

Nvidia CEO Jensen Huang announced the RTX Spark platform at Computex 2026 with the goal of reinventing the PC. The platform pairs the GB10 Superchip with Windows on Arm to enable local AI agents. We tested Microsoft’s Surface Laptop Ultra, which features a 20-core CPU, 6144 CUDA cores, and 128GB of RAM. While the system excels at local AI tasks, gaming performance remains limited due to shared memory bandwidth constraints and emulation overhead.

We compared the Surface Laptop Ultra against a MacBook Pro with an M5 Max SoC and a Ryzen AI Halo system using the AgentPerf-local benchmark. The Surface Laptop Ultra finished the Qwen 3.8-27B test in roughly 25% less time than the MacBook Pro, despite the Mac’s higher memory bandwidth. The older Ryzen AI Max+ 395 required twice as long to complete the same task. For the lighter Qwen 3.6 35B A3B model, the Surface Laptop Ultra completed the test in eight minutes, beating the MacBook Pro by 25% and significantly outperforming the Ryzen system.

Gaming results were mixed due to architectural limitations. Most games are not natively compiled for Arm, and the platform’s 300 GB/s shared memory bandwidth is lower than dedicated mobile GPUs like the RTX 5060 Mobile. In Gears of War: E-Day, GPU utilization averaged only 85% at near-1080p resolution with frame generation enabled. The system experienced occasional hangs during map transitions, indicating that the current design faces significant challenges for immersive gaming experiences.