We should have a price for "GPU"

Ideas
,
1

For the GPU related-queries. There is no current way to add those to a service plan or create a price set/price based on GPU cores.

The ‘Software/Service’ mapping I did was an attempt to overcome that limitation at the service plan level. However, I can fully appreciate that it’s not an accurate workaround (visually maybe, but the quantity wouldn’t be correct because it’s counting the software/service quantity and not the actual number of GPU cores). It would be nice to have some type of ‘GPU’ component type.

2

A few quick questions to understand:

  • Do you subdivide your GPUs cuda cores?

    • If so, is it predefined per VM or can a user type in any number?

  • Are your clusters always the same GPUs?

3
  • Do you subdivide your GPUs cuda cores? —> Current. We predefined for per VM.
    • If so, is it predefined per VM or can a user type in any number? —> Future. I think Morpheus should have a function for user type in any number
  • Are your clusters always the same GPUs? —> No. There will be a difference.