/Patent Tracker
NvidiaPatent applicationChips & hardwareInterest 6/10

Nvidia patents method to move math data into GPU memory faster

Nvidia describes a way for graphics processors to automatically move specific blocks of mathematical data, called tensors, into fast memory before they are needed. The system uses a special command to transform and store this data in the GPU's cache based on instructions from a software program.

Why it mattersThis could help computer chips handle complex calculations more smoothly by reducing the time spent waiting for data to arrive.

In the filing’s own words

Apparatuses, systems, and techniques to perform a tensor prefetch instruction to cause one or more tensors to be transformed and stored into one or more caches. In at least one embodiment, one or more circuits of a GPU are to perform a tensor prefetch instruction to cause one or more tensors to be transformed and stored into one or more GPU caches.

What exactly they claim (claim 1)

1. One or more processors, comprising: circuitry to, in response to a call to an application programming interface (API), cause prefetch of data of one or more tensors into one or more caches of an accelerator according to a tensor descriptor and based, at least in part, on an indication of at least a portion of a tensor in storage and an offset value.

Facts

Company
Nvidia (NVIDIA Corporation)
Official title
TENSOR DATA STORAGE
Publication number
US20260267647A1
Application number
19659113
Filed
Apr 27, 2026
Published
Sep 10, 2026
Inventors
11
Claims
20

Classified under

G06F 9/30047 Physics & computingG06F 12/0862 Physics & computingG06T 1/20 Physics & computingG06F 3/0625 Physics & computingG06F 17/16 Physics & computingG06F 2212/608 Physics & computing

More from Nvidia in Chips & hardware

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.