about the new product vComputeServer

I noticed that there is a new product called vComputeServer. I have checked it, I think it’s quite useful for building an internal AI laboratory in my company. But unfortunately, I couldn’t find the specific documents for introducing the features of that product or I don’t know where to find the documents.I have a lot questions to ask, for example:

  1. How many virtual GPU can I get with 1 physical V100 GPU?
  2. Should I buy the real physical server to use this product?
  3. Can I use cloud server instead? etc
    If all of these questions could be answered by checking the documents, can you send me a copy or the links? thanks

This document answers some of your questions:

https://images.nvidia.com/content/grid/pdf/nvidia-vcomputeserver-solution-overview.pdf

  1. You can get up to 8:1 subscription on a GPU. For a V100, the maximum subscription would be using the 4C profile (4 GB of memory per virtual GPU)
  2. You can use this product with server you have on your premises.
  3. You can use this product with servers in the cloud.