NVIDIA unveiled its RTX Spark, an AI “superchip” designed for Windows laptops and desktops, at Computex. The chip offers 1 petaflop of AI computing power and features 6,144 Blackwell RTX cores along with 20 Mediatek Arm CPU cores.
The RTX Spark, which NVIDIA likens to its RTX 5070 laptop GPU but with a reduced power draw, includes a neural processing unit (NPU) capable of 40 TOPS, integral to Microsoft’s Copilot+ initiative. Its GPU can utilize up to 128GB of unified memory and operate within a power range from single-digit watts to 80W. NVIDIA has no plans to release the RTX Spark alongside dedicated GPUs.
NVIDIA CEO Jensen Huang described the RTX Spark as a complete reinvention of the PC, emphasizing its shift towards AI agents rather than human input. “Today, when you think about your phone, the one thing you don’t do with it is make phone calls,” Huang said. “You do just about everything else.”
The development of RTX Spark has been a collaborative effort with Microsoft over several years, which indicates a familiarity with the Prism emulation layer needed for running older Windows applications on Arm-based chips. NVIDIA representatives also stated the company is collaborating with all major anti-cheat providers for popular games.
Pavan Davuluri, Microsoft’s head of Windows and devices, mentioned that Windows 11’s workload profile has been optimized for the RTX Spark to enhance performance and efficiency. “Whether you’re checking your email or running an agent locally to debug code, the Windows scheduler on RTX Spark will ensure you get the best performance and efficiency out of your CPU,” he noted.
The RTX Spark is aimed at content creators and those new to AI development, contrasting with the DGX Spark desktop, which targets large companies and AI developers. Interest in the RTX Spark’s market performance is heightened, as NVIDIA has not released a system-on-a-chip (SoC) for consumers since the Tegra series.
