Nvidia’s RTX Spark laptops pack 1 petaflop of AI performance

Nvidia CEO Jensen Huang says the company’s new laptop platform delivers 1 petaflop of AI performance. That’s a figure that used to describe machines filling entire rooms. Now it fits in a backpack. The pitch is simple: run serious AI models on your own device instead of renting someone else’s servers. What Nvidia actually built The hardware is called RTX Spark, and its centerpiece is what Nvidia calls a “superchip.” It pairs a Blackwell RTX GPU with a Grace CPU that Nvidia designed in collaboration with MediaTek. On paper, the specs are hefty. The GPU carries up to 6,144 CUDA cores, the parallel processors that do the heavy math for graphics and AI. The CPU side brings 20 cores. The headline number comes with a footnote. The platform hits up to 1 petaflop of FP4 AI performance. FP4 refers to 4-bit floating point, a compact number format that trades some precision for speed. It’s popular for running AI models efficiently, but it’s not the same as raw, high-precision compute. Memory might be the more important spec. RTX Spark offers up to 128GB of unified LPDDR5X memory, shared between the CPU and GPU rather than split into separate pools. Nvidia says the setup can run large language...