I have been struggling with understanding what the GPU-based licensing would mean for our setup. Here my questions:
- While we have only 2 users in a local network an individual Nucleus license is sufficient
- For larger projects > 2 users: we will need an enterprise license.
a. My assumption is this license works as follows: For each active GPU in the network, we will need a license.
b. Practical example: we have 3 pc’s with a 4090 for development + 1 pc with a 3060 which is hosting the nucleus. From the 3 pc’s only 2 are in use for the project => we need 3 licenses (1x for 3060, 2x for 4090)
i. Is this correct?
ii. This licensing model does not come automatically with Nvidias vGPU offering – correct? So if we would like to use vGPUs to max out the two 4090 machines this would require a separate license?
iii. If I would now connect the 3. PC with the 4090 to the network e.g. since one of the other 4090 pc’s is in use for a different project – would I need to buy a 4th license? So is the system based on the MAC address of the pc
iv. On a practical side this means from a cost-value perspective, having an A6000 is a much better value from the license, compared to having e.g. 2x 4090 in parallel?
Here the link to the license document: NVIDIA Omniverse Enterprise Pricing and Licensing Guide