News for hackers. Stuff that matters.
Re:Next-Generation IBM POWER10 Processor by gok on Monday August 17, 2020 @11:20AM (#24188104)
There's also the 512-bit matrix units which would be more like 3840 FLOP/cycle (single precision) for workloads that can be expressed that way
Re:Next-Generation IBM POWER10 Processor by dragontamer on Monday August 17, 2020 @11:50AM (#24188358)
I'm not sure if that plays to the advantage of POWER10.

NVidia 2070 Super also has FP16 matrix-multiplication units, achieving 57 FP16-matrix TFLOPs. These are the NVidia "Tensor Cores".

Ampere (rumored to be released within a few weeks...) even has sparse matrix-multiplication (!!) units being added in.