NVIDIA announced its official entry into the PC market during its keynote at Computex 2026 and GTC Taipei. CEO Jensen Huang revealed the RTX Spark, which is designed to power a range of premium laptops and small form factor systems arriving this fall. Desktops and workstations running Windows, powered by the GB300, are also slated for release.
The RTX Spark chips, formerly codenamed N1 and N1X, are designed for lower-power mobile form factors. Functionally, they are similar to the GB10 SoC that powers the existing DGX Spark mini-AI workstation.
The architecture is built around a two-chiplet SoC. A full-featured RTX Spark features 20 CPU cores (10× Arm Cortex-X925 performance cores and 10× Arm Cortex-A725 efficiency cores), connected to a Blackwell-architecture GPU with 6,144 shader cores. This setup parallels the performance of a discrete GeForce RTX 5070. The CPU complex links to a Mediatek-designed CPU + I/O chiplet via a fast silicon bridge interconnect offering 600 GB/sec of bandwidth.
These chips utilize a unified memory architecture, allowing the CPU and GPU to share the total system memory. Top-end configurations can support up to 128GB of LPDDR5X unified memory connected via a 16-channel interface. All RTX Spark SoCs will support PCI Express Gen 5 + Gen 4 connectivity.
Lower power N1-based RTX Spark models are expected to feature either 12-core (8P+4E) or 10-core (7P+3E) CPU configurations, with 2,560 or 2,048 shader cores, respectively. These configurations top out at 64GB of memory using an 8-channel interface, with TDPs in the 45W range.
NVIDIA is partnering with major companies including Dell, ASUS, Lenovo, HP, and Microsoft to bring these systems to market, with arrivals expected this fall. The initial target market is expected to be premium, creator, and AI-focused systems.
In addition to laptop and small form factor systems, NVIDIA also announced larger desktop and workstation systems. These larger machines will be powered by the RTX Spark and the GB300, featuring up to 748GB of coherent memory and up to 20 petaFLOPS of AI compute performance.
Regarding software and features, NVIDIA demonstrated its collaboration with Adobe, showing RTX Spark-optimized versions of Photoshop and Premiere with AI-enhanced capabilities. The company also detailed DLSS 4.5 Ray Reconstruction, a new AI pass that reconstructs full-resolution ray-traced effects. This feature is also coming to Blender 5.3. Additionally, NVIDIA announced up to 4x frame-generation for video using ComfyUI with the RTX Spark.
Looking forward, NVIDIA presented a two-year roadmap slide, pledging new PC-focused SoCs based on its latest architectures for laptops, desktops, and workstations, with subsequent chips slated for 2028 and 2030.