Some of this stuff is above my current knowledge of AI hardware so maybe I missed it, but this didn’t really explain why x86 is better suited for this application? Is it just because chips already exist that do this very specific thing that they need them to do?
Some of this stuff is above my current knowledge of AI hardware so maybe I missed it, but this didn’t really explain why x86 is better suited for this application? Is it just because chips already exist that do this very specific thing that they need them to do?
You need x86 (or arm) to run the general purpose OS and systems management tasks that feed the GPUs.