Imagine if folks used custom chips instead of standard gaming chips for deep learning. ~100X boost possible...https://twitter.com/mappingbabel/status/676853665500495872 …
@nachiketkapre @adapteva NV Maxwell has up to 336 GB/s global mem BW. Pascal is expected to have 1 TB/s. Hard to beat with a custom chip..
-
-
@oe1cxw@nachiketkapre Wrong premise, global memory is the past. Easy to get 100's of TB/s with truly distributed arch.#brain -
@adapteva@nachiketkapre But then you'd have to choose different training algos and ANN topologies than are used right now. 1/
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.