Ask HN: How to compare costs of compute providers?

I've got some services which are mostly compute bound (image processing etc). They are packaged as Docker images, don't require much else (blob storage and a message queue), so it can run on almost any cloud provider. AWS, GCE, Azure, etc.. Naturally I'd like to use the cheapest. However it is really hard to compare different providers. The provided workers are all different performance (CPU/RAM/disk/network). There are steps and limits in the worker pricing. With autoscaling the timebase (per hour/minute/second) becomes significant..

Are there good cross-cloud comparison tools for pricing? How do you compare?