AMD CEO Teases Next Generation MCM Based Instinct MI200 GPU


AMD CEO and President Dr. Lisa Su announced on Twitter the company's intentions to grab even more market share in the HPC market with their newest Instinct MI200 MCM GPU and EPYC technology at the most recent AMD Accelerated event that is going to be held on November 8th.

NVIDIA Sinks As Weak Guidance Rings Warning Bells Of Chip Slowdown

It is expected that Dr. Su will be revealing the Zen3 EPYC server processors, aptly named "Milan-X," as well as the Instinct MI200, the company's first CDNA2-based accelerator series (and also the first GPU to be an MCM design). In fact, if you look closely at the image, we may already see a hint of the new lines.

The Instinct MI200 is a GPGPU designed to accelerate algorithms used in data centers. It will be the first Multi-Chip-Module GPU featuring two compute dies. The accelerator known as Instinct MI250X is to offer 110 Compute Units and up to 128 GB of HBM2e memory.

— VideoCardz

AMD's MI200 GPGPU series utilizes an OCP Accelerator Module (OAM) connectivity. Currently, the Instinct models are only manufactured with a PCIe configuration. The EPYC CPU that is in the above image is possibly our first look at Milan-X, the next-gen Zen3 based EPYC CPU that onboard utilizes 3D V-Cache. VideoCardz does point out that it does not appear to be the "Zen4 'Genoa" SP5 package," which supports eight channels of memory.

Both AMD series are intended to be offered for next-generational server environments which will use high data speeds and processing with the company's AMD Infinity Fabric interconnect system. AMD hopes to compete with NVIDIA's Hopper as well as Intel's Ponte Vecchio and Sapphire Rapids next-gen servers starting in 2022.

November 8th will see AMD Accelerated Data Center Premier, one day before NVIDIA has their GTC 2021 Fall keynote MC'd by CEO Jensen Huang.

Sources: @LisaSu, VideoCardz