Ok im in a hurry and i will use more detail when i have the time.
Have iommu or vt-d enabled with above 4g decoding.
Install manjaro linux proprietary driver option
No need to install drivers mhwd will install everything out of the box in my case my nvidia settings will have a titan Black part and the tesla m40.
No need to change compute mode in linux teslas not locked for compute Mode but instead are unlocked.
Install the package nvidia prime reboot
If you change the steam shortcut or any app you want to use on the tesla or just write this on a terminal steam will run any game in the tesla m4
prime-run steam | grep “OpenGL renderer”
If you want to run with you normal gpu just use steam or the normal comand.
Games run fine and you dont need to run them on a Window they run fine on full screen.
For the cooling i removed the shroud exposed the sink forcing some fins open with a knife and placed 2 8 size fans blowing on the sink. Can manage 50 c on load.
Would you be able to test it for Ubuntu 20.04? Since im getting my hands on an NVIDIA A2 in a few weeks and curious to see if id be able to do something similar.