There have been multiple rumors surrounding the next-gen Instinct MI200 accelerator from AMD. Sources claim that the MI200 will essentially be an MCM approach to the existing MI100, featuring two dies instead of one, with the same CDNA architecture. However, it’s important to note that the MI200 has been specially designed for the Cray Supercomputer, named Frontier which is supposed to be the first Exascale Supercomputer.
Although the CPU is codenamed Trento, it’s based on the same Zen 3 core architecture as Milan, with some modifications allowing AMD to pair each chip with four Instinct MI200 GPUs using the Infinity Fabric 3.0 interconnect. As for the MI200 itself, it’s reportedly going to be a step up from the MI100, but it’s unclear whether that’s only going to be in terms of the sheer compute horsepower or in the form of an update microarchitecture as well.
The Frontier SC will be based on a unique design with each Epyc Trento CPU paired with four MI200 accelerators using the IF 3.0 interconnect, with each GPU directly connected to the CPU and every other GPU. This mesh design is what really makes the Trento-MI200 combo unique, as each chip has access to the data stored in the associated memory in a coherent manner, completely eliminating the need for direct management of memory copies on the program side.
Although Trento is being designed for Cray, it will also be available to other OEMs/ODMs. The main advantage of this platform is scalability, bandwidth, latency, and of course, the relatively simpler programmability thanks to the use of the Infinity Fabric 3.0 interconnect and a unified memory pool across the CPU and GPU.