Zen 4 lacks AVX512_FP16 (for 16-bits IEEE floating point operations), AVX512_VP2INTERSECT and also lack the Advanced Matrix eXtension (AMX) set (if you consider

Zen 4 lacks AVX512_FP16 (for 16-bits IEEE floating point operations), AVX512_VP2INTERSECT and also lack the Advanced Matrix eXtension (AMX) set (if you consider that part of AVX512). https://twitter.com/InstLatX64/status/1646471371558461445/ https://twitter.com/InstLatX64/status/1646471371558461445/ It's worth noting that 16-bits floating point operations are still possible on Zen 4, but using Google's BF16.