Over on the Open Compute Project Global Summit 2019 Intel’s Jason Waxman introduced a collaborative effort with social media big Facebook to develop the upcoming Cooper Lake server CPUs. Available with as much as 112 cores throughout a four-socket design and the bfloat 16 format, these Xeon chips are constructed to speed up machine studying and AI smarts.
Built on the 14nm course of node, Xeon ‘Cooper Lake’ server-grade processors will function bfloat 16, a 16-bit floating level format notably helpful for deep studying implementations. This numerical format condenses a spread equal to that of a 32-bit floating worth with some intelligent bit administration suited to AI schooling, and its notably conducive to bettering picture classification, speech recognition, suggestion engines, and machine translation. Intel additionally has plans to roll out bfloat 16 throughout its total Xeon, FPGA, and AI processor lineup.
Facebook’s newest tech enterprise has seen it dipping its toes into the AI and machine studying silicon game with seasoned vets Intel. The social media big’s temporary is much larger than only a place to submit canine photos these days, and the final twelve months have been notably newsworthy for the tech big following congressional grillings and the Cambridge Analytica scandal.
Yikes. Is this an organization we wish to have entry to higher, extra highly effective synthetic intelligence? I’m not so positive. But it nonetheless stays to be seen to what extent Facebook collaborated with Intel on Cooper Lake.
Consumer silicon: These are the best CPUs for gaming
Yet it’s GPUs, not CPUs, which have been touted as the perfect tech for the burgeoning synthetic intelligence scene. Intel’s been working on its own discrete graphics product to deal with the data-centric workloads of in the present day, appointing ex-Radeon boss Raja Koduri to steer the cost into the graphics card game in opposition to AMD and Nvidia.
That’s all anticipated to come back to a head in 2020 if Intel stays on monitor, close to sufficient the identical time Intel’s long-awaited 10nm course of will begin transport en masse.
It’s not totally clear skies forward for Intel’s Cooper Lake CPUs, nevertheless. Intel’s server market share has come beneath hearth from AMD’s EPYC Rome server CPUs, packing extra cores than ever due to the 7nm course of node and Zen 2 chiplet design. While the crimson workforce nonetheless hasn’t managed to get wherever near toppling Intel’s datacenter dominance, it’s not going to be a simple experience for Intel’s newest server methods.
Source
