GPU compatibility - older hardware

We have been developing a computer vision application in the past year as a side project or experiment in our organization. We’ve bought a Jetson Xavier NX and were able to use the excellent (really accessible and easy to use) Jetson Python libraries to create a proof of concept quite fast.

Now we are in the proces of making the application more mature. Which means more easily deploy-able, configurable and most of all more performant. We do not necessarily need the edge-computing aspect of the Jetson platform, we just bought one because of the relatively low price and ease of development. But we WOULD like to keep our options open in that sense.

So, because of this, we are now looking into porting our application to DeepStream. Which is quite a big overhaul because of the different approaches (where DeepStream uses a GStreamer pipeline and eventloop kind of approach and Jetson libraries just use blocking functions). But before we invest heavily into this big re-write, and before we put a lot of money into a new GPU server with A100 GPU’s, I would like to be able to use our existing/older hardware for development in the meantime, and if it all works out we can invest in a few A100’s.

Question: can and will DeepStream run on RTX 2080 Ti GPU’s? The Quickstart Guide (Quickstart Guide — DeepStream 5.1 Release documentation) states that P4, T4, V100 and GA100 platforms are supported. Does this mean that Turing RTX is not? (T4 is also Turing architecture so I have some hope 2080’s will work as well?)

yes, it’s well supported.
All CUDA capable GPUs with SM > 53 can be supported.


This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.