NVIDIA Launches PAIR Tool to Share AI Computing Power Across Devices
NVIDIA has launched a beta version of a tool named NVIDIA Personal Artificial Intelligence Router, which is called PAIR. This tool allows AI agents to use computing power from devices that are connected to the same home network rather than depending on just one computer the AI can spread the work across multiple devices.
It is made for people who use AI models that take jobs and split them into smaller parts. When many of these jobs are sent to one computer at the same time they often stop and wait in a line. PAIR sends these jobs to devices that are available, on the network, which can help make the waiting time shorter.
How does NVIDIA PAIR work?
NVIDIA PAIR does not take the place of programs like Ollama or LM Studio which’re the ones actually running the AI models. Instead, it works like a traffic manager. It looks at which devices are on the network and not being used. It checks if those devices have the AI model installed and if they are not busy with other things like playing games. Then it sends each job to the possible machine and brings the answer back, to the app that asked for it.
It is important to understand that PAIR does not bring together the power of graphics cards into one single device. Each task continues to work on one machine. The new idea is the chance to do tasks at the same time, on different machines. This stops the tasks from waiting in a line.
Who can use it?
PAIR currently works on Windows, macOS and Linux systems. It supports NVIDIA GeForce RTX 20-series graphics cards and newer NVIDIA RTX PRO workstation cards NVIDIA DGX Spark systems and Apple devices that have M4 chips or newer.
Why is PAIR important?
NVIDIA states that this setup is useful for “multi-agent” AI systems, where a primary AI assistant creates several smaller auxiliary agents to handle different parts of a task. In a test conducted with five of these auxiliary agents, a single laptop took about 18 minutes to complete the task. When the same task was split across three devices using PAIR the time it took dropped to nine minutes—half of what it took. NVIDIA says this test was a single demonstration and does not mean the same results will happen on every setup. The company adds that PAIR is meant to work in the background. Devices can. Leave the shared group whenever they want. For example, if a laptop goes into sleep mode or someone starts a game that uses a lot of graphics power on their PC PAIR adjusts automatically. Moves the task to another device.