You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Which older version of AMF should be paired with VCN firmware version 1.10? Or is it possible to upgrade the AMDGPU firmware which is loaded on an EC2 instance with GPU hardware?
The text was updated successfully, but these errors were encountered:
AMF runtime SO and some other related SOs should match VCN FW and should be installed all-together with driver package. I assume they are already installed on EC2. I am not sure what can be upgraded on EC2 but if so, it should be whole driver package.
AMF SDK doesn't have binaries, only header files and is fully backward compatible. You can use the latest SDK version but check for errors if your application would try to enable newer features.
The problem is, I am installing AMF library into a Docker image. I don't know any other way to do this, other than requiring the user to install their own AMF locally, then mount a path to the Docker container so it can access the host's copy of the AMF library.
Considering the problems, I guess it's better and more flexible to allow the user to mount their own /opt/amdgpu-pro from the host and supply their own compatible version. Latest works on Arch, but apparently not Ubuntu or Amazon Linux images.
I can only tell that installing FW into container has no effect as FW is loaded to HW during system boot time.
AMF relies on other SOs from driver package like libamdenc64.so and Vulkan driver.
You may start sharing AMF log to see the failure.
Which older version of AMF should be paired with VCN firmware version 1.10? Or is it possible to upgrade the AMDGPU firmware which is loaded on an EC2 instance with GPU hardware?
The text was updated successfully, but these errors were encountered: