---
title: "Microsoft Connects PCs via USB4: RDMA Clusters for Local AI Model Execution"
description: "Microsoft has implemented RDMA over USB4 in Windows 11, allowing multiple PCs to form a cluster for running large AI models locally. In a test, three mini-PCs ran Qwen3-235B at speeds up to 22 tokens per second."
date: 2026-11-21T13:04:10.931Z
lang: en
url: https://xab.info/en/posts/microsoft-rdma-usb4-ai-cluster-pcs
tags: [rdma, usb4, microsoft, ai-cluster, windows-11, llm, open-source]
author: "Alexander Buran, Roman V."
publisher: "XAB.info"
---

# Microsoft Connects PCs via USB4: RDMA Clusters for Local AI Model Execution

![Two silver mini-PCs connected via USB4 cable to form an RDMA cluster for local AI execution](https://xab.info/media/2026/11/21/microsoft-rdma-usb4-ai-cluster-pk/microsoft-rdma-usb4-ai-cluster-pk-1.webp)

## Key Points

- Microsoft introduced RDMA over USB4 in Windows 11 to combine PCs into a computing cluster
- The technology provides 28–50 µs latency compared to 116–194 µs for Ethernet
- In a demonstration, three Ryzen AI Max+ 395 mini-PCs ran the Qwen3-235B-A22B model
- The NetworkDirect provider for USB4 and the llama.cpp test platform are open source

Microsoft has announced the implementation of Remote Direct Memory Access (RDMA) over the USB4 interface, enabling multiple personal computers to be combined into a single computing system. The new feature is primarily aimed at locally running large language models that do not fit into the RAM of a single device. Essentially, enthusiasts and developers are given a tool to build a home AI cluster without the need to purchase specialized server hardware or high-speed InfiniBand or RoCE network cards.

### Direct Memory Access Without the Network Stack

RDMA technology allows data to be transferred directly between the RAM areas of two systems, bypassing a significant portion of the traditional network stack. This ensures minimal latency and significantly reduces the load on the central processing unit, which in standard network exchanges must participate in every data copying cycle. This is why RDMA has long been an integral component of storage systems in data centers and distributed computing clusters. Now, this technology has found a new application in the consumer and workstation segment. According to tests, the latency when using RDMA over USB4 is 28–50 microseconds, whereas with a standard Ethernet connection it reaches 116–194 microseconds.

### Connection Architecture and Hardware Requirements

Implementing this solution requires a USB4 interface, which provides significantly higher bandwidth compared to previous USB generations. This means that only reasonably modern computers equipped with matching controllers can be linked this way. Technically, Microsoft utilizes the peer-to-peer device connection mechanism already present in Windows 11 over USB4: each system creates its own USB4 domain, after which the Windows connection manager establishes a peer-to-peer tunnel between the devices. A specialized RDMA software component links data transfer queues with the ring buffers of the USB4 controller's DMA, avoiding additional memory copying operations. Two pairs of ring buffers are allocated for each connection. When combining three computers, one of them must be equipped with two USB4 ports.

### Demonstration Test Results

To demonstrate functionality, Microsoft tested the technology on mini-PCs powered by the Ryzen AI Max+ 395 processor with 128 GB of RAM. Three connected computers successfully ran the Qwen3-235B-A22B model with 4-bit quantization. Microsoft reported a generation speed of around 18.5 tokens per second. Meanwhile, AMD, using its own hardware configuration, achieved 21.98 tokens per second. The discrepancy in figures is explained by differences in hardware configuration and test bench setups between the two companies.

### Open Source and Future Prospects

The NetworkDirect provider for USB4, as well as the llama.cpp-based test platform used for the demonstration, are open-source projects. This means the developer community can independently adapt the solution, refine drivers, and create custom cluster configurations. For local AI labs, small research groups, and enthusiasts working with large language models, this solution can become an affordable alternative to cloud GPU clusters by utilizing existing desktop hardware.

## Fact-Check & Sources

- [Теперь можно объединять несколько ПК по USB4 для работы с большими ИИ-моделями. Microsoft реализовала работу RDMA новым способом](https://www.ixbt.com/news/2026/11/21/teper-mozno-obieediniat-neskolko-pk-po-usb4-dlia-raboty-s-bolsimi-ii-modeliami-microsoft-realizovala-rabotu-rdma-novym-sposobom.html) - Primary source: technology overview, USB4 requirements, Ryzen AI Max+ 395 test results, Qwen3-235B execution, 18.5 and 21.98 tok/s metrics, open-source component status.
- [Теперь ПК на Windows 11 можно объединять в вычислительные кластеры](https://wtftime.ru/tech/170930/teper-pk-na-windows-11-mozhno-obedinjat-v-vychislitelnye-klastery/) - Additional context: 28–50 µs latency vs 116–194 µs for Ethernet, description of USB4 domain and peer-to-peer tunnel mechanism, DMA ring buffers.

## FAQ

### What is RDMA and why is it needed for AI?

RDMA (Remote Direct Memory Access) is a technology for direct data transfer between the memory of two computers without involving the network stack. For AI, this is critical because it pools the RAM of multiple PCs to run models that exceed a single device's capacity with minimal latency.

### What are the hardware requirements?

Computers with a USB4 interface are required. To connect three systems, one of them must feature two USB4 ports. Testing was conducted on mini-PCs with Ryzen AI Max+ 395 and 128 GB of RAM.

### How accessible is this solution for enthusiasts?

The NetworkDirect provider for USB4 and the llama.cpp-based testing platform are distributed as open-source projects, allowing the community to configure and modify clusters independently.