Yesterday, AMD announced a number of new processors in their server lineup of EPYC CPUs. The 4th generation of EPYC processors will come in two segments – one codenamed ‘Genoa’ with up to 96 Zen 4 cores, and another codenamed ‘Bergamo’ with up to 128 Zen 4c cores. Along with this, AMD also announced first exascale-class GPU accelerators called AMD Instinct MI200 series accelerators.
Both the Genoa and Bergamo EPYC will be built on TSMC’s 5nm process node. They will also support upcoming technologies such as DDR5, PCIe 5.0, CXL 1.1, RAS, and AMD’s security suite.
Where they differ is the type of each core, the Genoa cores are Zen 4 cores that will be built for most applications, while Bergamo cores are Zen 4c cores that are higher density version purpose built for the cloud. Genoa is expected to arrive in 2022, while Bergamo is expected to arrive in late 2022/ early 2023.
During the same event, AMD also unveiled their Instinct MI200 series accelerators that are based on CDNA 2 architecture. These new accelerators are AMD’s fastest high-performance computing (HPC) and artificial intelligence (AI) accelerators, that AMD claims will offer up to 4.9 times better performance than competitive accelerators for double precision (FP64) HPC applications.
Along with CDNA 2 architecture that comes with 2nd Gen Matrix Cores, these accelerators also feature a multi-die GPU design with 2.5D Elevated Fanout Bridge technology, and their 3rd Gen AMD Infinity Fabric technology. AMD Instinct MI200 series accelerators are currently available from HPE in the HPE Cray EX Supercomputer, and later they are expected in systems from major OEM and ODM partners in enterprise markets in Q1 2022.