Nvidia took center stage at Computex 2026 to reveal its highly anticipated entry into the consumer chip market: the RTX Spark. This innovative chipset combines a powerful and efficient ARM CPU with an RTX GPU.
The RTX Spark superchip features an Nvidia Grace CPU with up to 20 cores, paired with an Nvidia Blackwell GPU boasting 6,144 CUDA cores and up to 128GB of fast LPDDR5X unified memory.
Built on Nvidia's GB10 architecture within the DGX Spark platform, this chip is designed for enterprise-class mini-PCs that run a custom version of Ubuntu Linux. The GPU effectively replicates the RTX 5070 (non-Ti) with an equivalent number of CUDA cores and utilizes TSMC's advanced 3nm fabrication process.
While the RTX Spark is explicitly tailored for content creation and gaming, Nvidia CEO Jensen Huang emphasized that the primary focus lies on running AI agents locally on personal computers. These AI agents, representing the evolution of large language models (LLMs), can execute complex tasks on demand, requiring substantial processing power and memory. Here, the 128GB of unified memory plays a crucial role, contributing to the RTX Spark's impressive 1 petaflop AI output capability.
The RTX Spark is tailored for sleek yet powerful laptops, expected to debut this fall. Nvidia is collaborating with industry leaders such as Acer, Asus, Dell, Gigabyte, HP, Lenovo, MSI, and Microsoft, which plans to launch its most potent device to date, the Surface Laptop Ultra, featuring this System on a Chip (SoC).
While exact pricing details are not available, the RTX Spark is anticipated to be on the expensive side. For context, Nvidia's DGX Spark ranges from $3,500 to $4,700 and can be equipped with up to 128GB of RAM. When paired with a premium laptop featuring an OLED display, the final cost could be substantial.
Source
