Nvidia patents direct GPU-to-GPU messaging system
Nvidia describes a method where graphics processing units (GPUs) send messages directly to one another without going through the main computer processor. The system uses temporary software tasks called "gadgets" that start, handle specific messages between GPUs, and shut down once the work is complete.
In the filing’s own words
Approaches presented herein relate to active messaging for parallel processing architectures, such as utilizing active messages (AMs) for direct graphics processing unit (GPU)-to-GPU communications. Gadgets may be launched across multiple GPUs for specific types of message handling. When a gadget is launched, the gadget may run across multiple GPUs to send messages between the different GPUs handling a given task. In operation, a gadget may include multiple handlers, and may be executed for specific tasks through completion prior to shutting down. In this manner, multiple gadgets may be launched sequentially to execute different selected operations across the GPUs.
What exactly they claim (claim 1)
1. A processor, comprising: one or more circuits to: launch, based on a signal from a host device, a gadget across a plurality of graphics processing units (GPUs), the gadget including at least a kernel entry point and a handler; write an active message to a buffer; send, to a receiver using the handler, the active message; determine all messages in the buffer have been handled; and cause the gadget to shut down.
Facts
- Company
- Nvidia (Nvidia Corporation)
- Official title
- ACTIVE MESSAGING FOR PARALLEL PROCESSING ARCHITECTURES
- Publication number
- US20260310529A1
- Application number
- 19418676
- Filed
- Dec 12, 2025
- Published
- Oct 8, 2026
- Inventors
- 2
- Claims
- 20
Classified under
G06F 9/546 Physics & computingG06F 9/544 Physics & computingMore from Nvidia in Chips & hardware
- Nvidia patents method to skip zero data in neural network math Oct 13, 2026
- Nvidia patents hardware to speed up motion blur in ray tracing Oct 13, 2026
- Nvidia patents method to sync data across split processor chips Sep 17, 2026
- Nvidia patents chip math trick to speed up AI calculations Oct 8, 2026
- Nvidia patents data center cooling that uses cryogenic fluids Oct 15, 2026
Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.