When renting GPUs for ML workloads, how do you actually choose between providers? There are now so many GPU cloud / GPU sharing platforms, and many of them seem to offer similar GPU options....
So, if the GPU model is the same and providing similar functionalities, do you mostly choose the cheapest provider? Or do reliability, availability, networking/storage, and setup environment matter more for you?
Trying to understand what the real pain point is and make right decision for me when I am choosing the provider.
Also curious: would you rather manually compare providers yourself, or use a service that recommends the right GPU/provider based on your workload?
Comments URL: [https://news.ycombinator.com/item?id=48537835](https://news.ycombinator.com/item?id=48537835)
Points: 1
source & further reading
news.ycombinator.com — original article
What are you looking for when reviewing LLM generated code?
Ask HN: Developers, are you being forced into prompt-only engineering?
Ask HN: Did you try Claude's "Fable 5" model before it was pulled?