> The counterargument would be that they've deployed a ton of network capacity and GPU resources for cloud computing anyway, and so Stadia is just riding on those coattails... But it's certainly not a promising track record.
Exactly, in fact, Stadia is really just a way for Google to subsidize their cloud GPU buildout. Do you think those expensive GPUs are sitting idle in the meantime? No way! They're crunching numbers or spot instances for AI/ML workloads.
That's...probably not true at this moment. They're using AMD GPUs, which generally support the sorts of workloads folks need, and AFAIK they haven't opened up support for AMD GPU spot instances to users.
The OpenCL ML ecosystem is really lacking at the moment.
(Disclaimer -- I'm one of the devs. Even though we were bought by Intel a year or so ago, we've been pretty good about keeping the AMD / OpenCL support intact.)
Exactly, in fact, Stadia is really just a way for Google to subsidize their cloud GPU buildout. Do you think those expensive GPUs are sitting idle in the meantime? No way! They're crunching numbers or spot instances for AI/ML workloads.