Recent Update Log

Updates:

June 25: Moved to new site
June 24: Added excellent new tables showing best GPUs and clouds by use case
June 24: Updated the H100 and A100 instant access availability and pricing table
June 23: Added Cirrascale to the list of additional cloud options
June 22: Updated the H100 and A100 instant access availability table
June 21: Added DGX info, VRAM and vCPU info, SXM/PCIe and NVLink info
June 20: Added a big list of alternative gpu clouds
June 20: Added tables for H100/A100 instant access (including refreshed price and capacity), GPU use cases, large H100 quantities, and alternative GPUs with price and availability, and chart comparing price of running Falcon-40B
June 19: Added Falcon-40B GPU cloud comparison chart
June 19: Added conversation with founder of Oblivus and founder of Runpod
June 16: Added conversation with founder of FluidStack
June 15: Added conversation with VP Product at Salad
June 14: Published initial post

Upcoming changes:

Possibly upcoming - email me to let me know which one would be helpful:

A flow chart or interactive tool for choosing the right GPU, whether to buy or rent, and which GPU cloud to use
A tool or spreadsheet for comparing the payback time of building vs renting
Uptime, reliability and ease of use comparisons
A tool for entering your usage and getting a list of which providers will be cheapest for your use case

I’m hiring freelancers. If you have ideas of things that could be good to add or improve, email me.

Thanks to HN and others for post corrections and improvements, keep them coming via comments or email.