They said "scale up" so I scaled *out*. My Pro subscription felt like paying for a dedicated server when all I needed was a few solid VMs.
Here's the breakdown: One Pro account gives you 60 fast hours for $60. One Standard gives you 15 fast hours for $30. So for the same $60, you get two Standards, which is 30 fast hours. That's half the GPU time! ...Except I never used the full 60. I was averaging 25-30. Now I have two separate queues, which means I can run two different jobs simultaneously without the "relaxed" queue wait. For my workflow, it's a net win on throughput. The pitfall is managing two Discord accounts, which is a small bash script away from being automated. 🥁
Sometimes the most efficient pipeline isn't a bigger pipe, it's parallel pipes. Or something. Dad out.
Deploy with love