After upgrading its consumer processor lineup just recently, Intel again came on the stage and announced new Xeon Scalable CPUs, codenamed Cooper Lake.
Intel says its next-gen Xeon Scalable CPUs have higher core count and will arrive with up to 56 cores per socket. In comparison to the older Xeon 8200 series, the new server chips will provide higher memory bandwidth, higher AI inference, and training performance.
Furthermore, Intel has added support for bfloat16 format to Intel Deep Learning Boost, making Cooper Lake the first x86 CPUs with built-in AI training acceleration.
All in all, it means that new Xeon chips would be even better at handling AI workloads such as image classification, objection recognition, speech recognition, etc.
This builds on top of the existing Xeon Platinum 9200 series (Cascade Lake-AP) released earlier this year. One of the major highlights of Cooper Lake is their socketed format which helps them fit into more diverse deployment scenarios. This gives it an edge over the existing 9200 series which offers BGA-mounted configuration.
Intel also said that Cooper Lake CPUs will be compatible with the upcoming 10nm Ice Lake CPUs. The new Xeon Scalable CPUs with up to 56 cores are scheduled to arrive sometime in the first half of 2020.