Changes

: Current GPU [https://en.wikipedia.org/wiki/Processor_register registers] and Vector-[https://en.wikipedia.org/wiki/Arithmetic_logic_unit ALUs] are 32 bit wide and have to emulate 64 bit integer operations.<ref>[https://en.wikichip.org/w/images/a/a1/vega-whitepaper.pdf AMD Vega White Paper]</ref> <ref>[https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/technologies/turing-architecture/NVIDIA-Turing-Architecture-Whitepaper.pdf Nvidia Turing White Paper]</ref> 

: Some architectures like AMD [https://en.wikipedia.org/wiki/AMD_RX_Vega_series Vega] or Intel [https://en.wikipedia.org/wiki/Intel_Xe Xe] offer higher throughput with lower precision. They double the [https://en.wikipedia.org/wiki/FP16 FP16] and quadruple the [https://en.wikipedia.org/wiki/Integer_(computer_science)#Common_integral_data_types INT8] or octuple the INT4 throughput.<ref>[https://en.wikipedia.org/wiki/Graphics_Core_Next#fifth Vega (GCN 5th generation) from Wikipedia]</ref><ref>[https://www.servethehome.com/intel-xe-sg1-hp-and-dg1-at-architecture-day-2020/intel-architecture-day-2020-xe-lp-int8-increase/ xe-lp-int8 from servethehome]</ref>