Nvidia patents chip logic that swaps memory rules for AI tasks
Nvidia describes a processor design that automatically chooses how to manage temporary memory, known as cache, while running artificial intelligence models. The system applies different memory management rules to different parts of the AI network to improve how data is stored and retrieved during processing.
In the filing’s own words
Apparatuses, systems, and techniques to select cache policies. In at least one embodiment, a system causes one or more cache policies of one or more caches to be selected based, at least in part, on one or more neural networks to use data stored in the one or more caches.
What exactly they claim (claim 1)
1. One or more processors, comprising: circuitry to; select a first one of two or more different cache eviction policies to process a first portion of one or more neural networks; select a second one of the two or more different cache eviction policies to process a second portion of the one or more neural networks; and cause the first one of the two or more different cache eviction policies to be used to process the first portion of the one or more neural networks and the second one of the two or more different cache eviction policies to be used to process the second portion of the one or more neural networks.
Facts
- Company
- Nvidia (NVIDIA Corporation)
- Official title
- Selectable cache policy
- Patent number
- US12792517B2
- Application number
- 17581801
- Filed
- Jan 21, 2022
- Granted
- Nov 10, 2026
- Inventors
- 2
- Claims
- 31
Classified under
G06F 12/0833 Physics & computingG06F 12/0871 Physics & computingG06F 12/0891 Physics & computingG06N 3/045 Physics & computingMore from Nvidia in Chips & hardware
- Nvidia stacks memory directly on chips to boost data speed Nov 10, 2026
- 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
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.