What I'm trying to say is that the Nvidia Driver plugin might be able to keep a certain version in the driver version option for a long time.
Allow me to describe the benefits of doing this:
The Nvidia Driver plugin currently keeps updating the host driver version in a rolling fashion, and each time I update the system version, the Nvidia Driver plugin forces me to install the new host driver. This creates a number of problems, and I've already seen some feedback about incompatibility with older Kepler cards, as well as problems with some people's Docker apps.
In my development workflow, I've long used a combination of LXC and Nvidia Driver to get both a VM-like experience and GPU compute power, but having to have the same version of the userspace driver installed in the LXC container as the Unraid host also forces me to update the driver one-by-one for each LXC container every time I update the Unraid system and the Nvidia driver. In the most recent update, my LXC development environment was completely destroyed. I'm sure this is a huge nuisance to those who have had similar experiences as me as well.
I understand that compiling a large number of driver versions for each unraid version is a difficult thing to do, which is the main reason why the nvidia driver plugin is in a rolling mode, but I'm sure it's not too difficult to maintain just one additional version, and if you're still not willing to support this request, perhaps you could explain the driver compilation process in a note that would allow me to do this myself?
Much appreciated!
What I'm trying to say is that the Nvidia Driver plugin might be able to keep a certain version in the driver version option for a long time.
Allow me to describe the benefits of doing this:
The Nvidia Driver plugin currently keeps updating the host driver version in a rolling fashion, and each time I update the system version, the Nvidia Driver plugin forces me to install the new host driver. This creates a number of problems, and I've already seen some feedback about incompatibility with older Kepler cards, as well as problems with some people's Docker apps.
In my development workflow, I've long used a combination of LXC and Nvidia Driver to get both a VM-like experience and GPU compute power, but having to have the same version of the userspace driver installed in the LXC container as the Unraid host also forces me to update the driver one-by-one for each LXC container every time I update the Unraid system and the Nvidia driver. In the most recent update, my LXC development environment was completely destroyed. I'm sure this is a huge nuisance to those who have had similar experiences as me as well.
I understand that compiling a large number of driver versions for each unraid version is a difficult thing to do, which is the main reason why the nvidia driver plugin is in a rolling mode, but I'm sure it's not too difficult to maintain just one additional version, and if you're still not willing to support this request, perhaps you could explain the driver compilation process in a note that would allow me to do this myself?
Much appreciated!