Originally published at: https://developer.nvidia.com/blog/advanced-api-performance-async-copy/
Using async copy on your NVIDIA GPU effectively can significantly increase your performance. These quick recommendations give guidelines on how to arrange your work queues to maximize your gains.
1 Like
I hope this article was helpful, I also learned a lot when consulting my fellow Nvidians while writing it. While Async Copy and Async Compute may seem like almost the same thing at first glance, they do have key differences to master when putting into practice. If you have any questions or need any clarifications I’ll be more than happy to help!
Hi Katherine,
Thank you for your introduction regarding async copy. I would like to ask about the parallelization of graphics and compute tasks. If our GPU is used for both graphics and LLM (Large Language Model) tasks, the LLM will occupy the GPU for a significant amount of time fetching data from memory, which could squeeze the GPU resources allocated for graphics tasks. Can we use async copy to fetch the data required by the LLM during the graphics processing?
Thank you.