Join for quick access. Enter your email below and we'll send confirmation plus sign you up to our newsletter.
By submitting your information, you confirm you are aged 16 or over,
have read our
Privacy Policy
and agree to the
Terms & Conditions. Geographical rules apply.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
You are now subscribed
Your newsletter sign-up was successful
Join the club
Get full access to premium articles, exclusive features and a growing list of member rewards.
An account already exists for this email address, please log in.
Nvidia PAIR utility joins every GPU in your home into a cluster for agentic AI tasks — tool uses spare cycles to keep agent swarms from hammering one GPU
If you're a token-hungry AI enthusiast, and if you or your family happen to have PCs with idle GPU cycles to spare in this economy, Nvidia wants to make it possible to harness those cycles so you can save cash on cloud tokens and keep your work private. At IFA 2026, the company is introducing a local distributed AI clustering tool called the Personal AI Router (PAIR) that dispatches agentic AI sub-tasks from your main PC to systems on your home network that have suitable GPU cycles to spare.
As Nvidia tells it, when a user runs a local AI agent and gives it a goal to complete, that central agent might then spawn several sub-tasks carved out of that larger goal. If those sub-tasks or sub-agents are all running on the same GPU, the contention they create might cause the task to finish more slowly than it could if each sub-agent had a dedicated compute node to work with.
PAIR is a tool that can make that distributed AI work happen on a home network. Assuming that a family or shared household is sufficiently flush with idle GPU resources, PAIR canYEa assign each participating system one of those sub-tasks to perform and return the results to the main node, potentially resulting in faster completion of the larger agentic task.
Of course, systems on your local network won't always be idle. Their owners will frequently use the GPUs in their systems for gaming, creative work, or AI tasks of their own. If a user needs their GPU back, PAIR purports to gracefully deal with those changing conditions. It doesn't reserve dedicated capacity from other PCs; it's elastic by design and will make the best of the resources available to it at any given moment.
This unpredictable availability of spare cycles does, of course, mean that quality of service is not assured from a PAIR cluster. But for long-running tasks that don't need to be done on a strict deadline, being able to put spare compute to work could still be more effective than running an agent swarm on a single node.
PAIR sounds relatively simple to set up. It creates a proxy for popular AI front-ends like LM Studio and Ollama to connect to. PAIR then orchestrates work across available nodes on the network and returns the results of that work to the originating application on the head node.
In turn, participating PAIR nodes also need to be running Ollama or LM Studio and have a PAIR installation of their own. Nvidia says that enrolling systems in a PAIR cluster is straightforward and relies on mDNS or an IP address fallback for discovery. PAIR will also help initiate model downloads on participating systems, but Nvidia says that nodes don’t need to have identical models or sets of models downloaded to participate.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
If more systems do have a given model available, though, it broadens the pool of potential nodes that can handle a request if the orchestrator agent needs a particular model’s capabilities.
PAIR will run on any DGX Spark (or other GB10) box, as well as GeForce RTX 20-series graphics cards or newer. It also supports Macs with M4-series processors or newer for inference. Accordingly, the PAIR client will be available for Windows, macOS, and Linux.
As the Senior Analyst, Graphics at Tom's Hardware, Jeff Kampman covers everything that has to do with graphics cards, gaming performance, and more. From integrated graphics processors to discrete graphics cards to the hyperscale installations powering our AI future, if it's got a GPU in it, Jeff is on it.
I vaguely recall distributed CUDA support over IB during Kepler times but back then the 100Gbit ConnectX-5 adapters I had available supported IB only with a matching switch ($$$), while Ethernet was free via direct connect cables, but was not supported for those distributed CUDA libraries... pretty sure that was a management decision by Mellanox, who wanted to sell switch gear to HPC customers.
Anyhow, this is agent routing, not splitting model layers, orders of magnitude less bandwidth to be workable, pretty sure you could emulate IP over IB, if that's what you happen to have.
I've experimented a bit with IP over Thunderbolt in the homelab, to get a lower latency cluster fabric: with bigger packet sizes it had both throughput and a touch of Connect-X direct connect latencies, albeit at only 10GBit/s.
Trouble is that while IP over TB shines on latencies, it can't handle topology events (plugging/unpluggin) with automatic recovery (for lack of a MAC?), so it's not very practical in a cluster.
Vendors finely balance an illusion of value and functionality vs. the need to sell extra hardware, pretty sure Nvidia is mostly holding a carrot here: it has a "Chat with RTX" kind of vibe for me.
Amortizing work size against link speed is entirely reasonable solution. Agent routing is breaking down a query into parts that can run independently? If so then yes, the seconds the GPU needs to do it's work will mask the milliseconds needed to transmit the query and response.
I got a bunch of ConnectX-2 cards and an IS5022 switch on ebay about 8 years ago, not bad price. 40Gb/s but it makes GigE look slow, and it can do RDMA for NFS and it used to do RDMA for P2P when I used RTX2080s. With 2080s splitting a CNN using 40Gb/s worked OK. Now with 5080s 40Gb/s is too slow for splitting CNNs nevermind LLMs.
But nV disabled DMA even over PCIe between cards in the same box in the 3000 series. Everything goes thru CPU. It was a kind of mean thing to do IMO.
