Nvidia patents method to keep graphics data on specific chips in multi-chip GPUs
Nvidia describes a way to organize graphics data so that each part of a multi-chip GPU stores the image details for specific areas of the screen in its own local memory. By assigning pixels to specific chips based on their screen location, the system avoids sending data back and forth between chips, which can speed up rendering.
In the filing’s own words
Methods and systems are described for localizing render targets in a multi-chip graphics processing unit (GPU) thereby reducing the inter-chip traffic and improving the latency of certain rendering. The localization techniques provide for storing surfaces of different bit densities on the same locally connected memory, and enabling access to the different surfaces in a similar manner. The described localizing may be implemented over an interleaving of the virtual memory on the locally connected memories of each die in the multi-chip GPU.
What exactly they claim (claim 1)
1. In a system including a first processor locally connected to a first memory and a second processor locally connected to a second memory, a method comprising: selecting one of the first processor or the second processor to execute one or more threads associated with one or more pixels, wherein the selecting is based on screen location coordinates of the one or more pixels and a mapping, the mapping being configured to map pixels in a first plurality of areas of a screen to the locally connected memory of the one of the first processor or the second processor and to map pixels in a second plurality of areas of the screen to the locally connected memory of the other of the first processor or the second processor; and storing data of the one or more pixels in a first surface of a render target in the locally connected memory of the one of the first processor or the second processor.
Facts
- Company
- Nvidia (NVIDIA Corporation)
- Official title
- RENDER TARGET LOCALIZATION IN A MULTI-CHIP GPU
- Publication number
- US20260289721A1
- Application number
- 19087064
- Filed
- Mar 21, 2025
- Published
- Sep 24, 2026
- Inventors
- 8
- Claims
- 35
Classified under
G06T 1/60 Physics & computingG06T 1/20 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.