At Computex 2026 Nvidia unveiled the RTX Spark, a so‑called “superchip” aimed at Windows laptops that need heavy AI, graphics and battery life. The package pairs a 20‑core ARM‑based Grace CPU with a Blackwell RTX GPU housing 6,144 CUDA cores, fifth‑gen Tensor cores, up to 128 GB unified memory, a 600 GB/s NVLink‑C2C link, and targets ultrathin notebooks that claim all‑day endurance.

Pre‑release Geekbench data from June 2025 shows the N1X variant scoring 3,096 single‑core and 18,837 multi‑core points. Apple Silicon M3 Max in a 16‑inch MacBook Pro posts 3,128 single‑core and 20,969 multi‑core, while the newer M5 reaches 4,224 and 17,465 respectively. Nvidia has not released benchmarks, so the comparison remains tentative. With fewer cores than Apple’s 16‑core M3 Max, the RTX Spark trails by roughly two years of silicon progress.

Nvidia markets the chip as the “personal AI computer,” promising on‑device large language models and 4K video generation. For Windows notebook makers, the RTX Spark offers a unified architecture but must close a sizable performance gap before it can rival Apple’s tightly integrated silicon. NVLink may attract power users, but the product still lags behind the current Apple benchmark.