Hello,
We are considering creating a cloud of GPU enabled remote desktop VMs. Instead of RDP, we are using an "ultra-low latency solution". I'm quite keen to plan this based on the current AMD hardware. However, I'm not quite sure even where to start with planning the hardware. With Nvidia, the information is a bit more available, but with AMD, there is a ton of questions, we are not able to find answers to. Is there someone who can advise us here. At prima vista here are some of the actual topics:
- Which GPUs are appropriate?
- Because of the streaming we will heavily rely on the "hardware video encoding" of those GPUs. Which GPUs offer sich capabilities.
- As we want to virtualize several users on a single GPU, how many fps in total can the GPU handle (e.g. 5 users per GPU * 60 fps = total of 300 fps)
- How much a single server with specific specs will costs
- How much a single GPU costs.
- Can we use a large number of retail-grade GPUs, and will it be at any value at all in comparison with datacenter grade GPUs
- Are there any licensing issues with using retail GPUs in a datacenter?
- Are there any additional license costs associated with the GPU virtualizations (Nvidia licenses are outraging here)
- Which hypervisors are supported
- Can we white label AMD Link, if we found it viable?
- How is the AMD partner program works?
This is what just appears at a "prima vista". Any idea who can help with obtaining this information?
Thank you in advance! :)
