At a Microsoft event in San Francisco on 7 October, Nvidia and Microsoft set out hardware intended to run AI models locally on Windows machines. RTX Spark laptops opened for preorder the same day, with availability from 16 October and compact desktops in November. Systems are coming from Acer, ASUS, Dell, HP, Lenovo, Microsoft, MSI and Gigabyte. Microsoft built its Surface Laptop Ultra around the part.
Nvidia gives RTX Spark a Blackwell GPU of up to 6,144 cores, a Grace CPU of up to 20 cores, a 600 GB/s link between them, up to 128GB of unified memory and one petaflop of FP4 performance. Separately, Nvidia put DGX Station for Windows into preview, a deskside machine on the GB300 Grace Blackwell Ultra superchip with 748GB of coherent memory and up to 20 petaflops of FP4. DGX Station previously ran only Linux.
The pitch is that a model, and the agent built on it, can stay on the machine rather than going to a cloud. Two cautions. All the performance figures are Nvidia’s own numbers, quoted at FP4, a low-precision format that flatters throughput and tells you little about accuracy. And neither the laptop prices nor a DGX Station price or ship date were published, so the cost of keeping work local is still unknown.
Source: Nvidia, NVIDIA, Microsoft kick off a new beginning for Windows PCs with RTX Spark and AI agents.