The newly released product designs will ship in October alongside products from six existing OEMs.
By Frank Ulom
·
Published on September 3, 2026
·
2 min read
NVIDIA announced today that the new RTX Spark Windows PC will be available in October, providing AI enthusiasts and developers with more creative options.
According to reports, NVIDIA’s partners showcased related hardware at IFA 2026. The newly released product designs will ship in October alongside products from six existing OEMs. For instance, Acer showcased a concept design for its RTX Spark high-efficiency desktop PC, while Lenovo released the Yoga Pro 9n and Yoga 9n 2-in-1 device.
According to official information, RTX Spark will be equipped with a high-efficiency 20-core Grace CPU and up to 128GB of unified memory, as well as an RTX Blackwell GPU with a computing power of up to 1 petaflop (Daily Lamp note: petaflop floating-point operations).
The RTX Spark super chip powers high-performance, thin and light laptops with all-day battery life, as well as compact, energy-efficient desktops, supporting always-on agents . With the new Windows Agent framework, agents can run securely in the background under operating system-level control.
In addition, NVIDIA today announced the launch of the free and open-source Personal AI Router (PAIR) technology , which can intelligently distribute AI inference tasks among multiple computers on a local network. Agent workflows typically break down complex tasks into smaller tasks that can be executed in parallel, but performance can degrade when all inference requests are contending for the same GPU.
NVIDIA PAIR automatically discovers compatible PCs on the local network, routing independent inference requests to systems with available computing power. It works with Ollam and LM Studio, automatically adjusting as devices join or leave the network. This is compatible with Windows, macOS, and Linux, and with RTX 20 series and later GPUs, RTX PRO desktop GPUs (Turing architecture and later), DGX Spark, and Apple M4 or later chips.
