Deepstream 6.2 installation on ubuntu 20.04

Hello everyone,

I’m trying to install Deepstream 6.2 on a fresh ubuntu 6.2 installation. When I install the nvidia driver from the Ubuntu drivers manager or from the official nvidia website, I’m getting deiver versions more recent that the version listed in the Deepstream installation guide. And these recent versions are coming with more recent CUDA version (now it’s brining 12.1).
Is this ok ? Or I need to search for old driver that comes with older cuda version?
UPDATE: the Driver listed in the documentation is the NVIDIA driver 525.85.12 comes already with the CUDA 12 based on the changelog. How can we require only CUDA 11.8 in the documentation ?

I have an other question please, does the latest Cudnn and TensorRT works for Deepstream 6.2 or I need to have the specific versions listed in the installation guide.

Is there any easy process to install all in one without pain ?

Thank you for your assistance and help.

• Hardware Platform (Jetson / GPU): RTX 2070 Max Q
• DeepStream Version 6.2
• TensorRT Version N/A
• NVIDIA GPU Driver Version (valid for GPU only) N/A
• Issue Type( questions, new requirements, bugs) Questions

If you have installed NVidia driver via driver manager, you can uninstall it and re-install from the file listed in the installation guide.
If you don’t want to uninstall&re-install, you can consider using Docker container:

Thank you for your response… but I think there is a misunderstood… the driver listed in the documentation is the one dedicated for TESLA and A1000 cards and go on… it embeds CUDA 12… while the Deepstream 6.2 is requiring CUDA 11.8… How this is can be done ? I there any way to downgrade the CUDA after installation ?

The two proposed solutions needs the environment to be installed to be able to run DS inside Docker with nvidia container toolkit.

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

Deepstream related development and testing are based on its own documentation. So we suggest you install the CUDA11.8 by yourself.

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