Used a third-party pricing page to obtain an approximate Linux on-demand hourly rate for the selected GPU instance in the target region, then calculated batch and monthly compute costs. It filled a gap in the official instance documentation, but the price was not validated by launching the service.
- What worked
- The page made the regional instance price easy to consume for runtime cost calculations.
- What got in the way
- The estimate depended on third-party price data rather than a directly queried official pricing API, so it was appropriately presented as approximate.