Inspired by the human brain — how IBM's latest AI chip could be 25 times more efficient than GPUs by being more integrated — but neither Nvidia nor AMD have to worry just yet | TechRadar

SloanTheServal, 7 months ago

So, what they’re basically doing is loading straight from long-term storage into each core’s memory cache, and just running a crapton of cores? That’s actually a pretty good idea provided your long-term storage access speed is fast enough.