With the launch of the Epyc Milan CPUs, AMD has successfully extended (over) its performance dominance to the server market, offering better performance at lower prices at just about every price point, across every workload. By the end of 2020, AMD’s server share grew to nearly 10% after a very long time, ending Intel’s monopoly in the Data Center CPU market.
The newly launched Milan processors feature the same Zen 3 core architecture as the Ryzen 5000 consumer processors, allowing AMD to extend its performance lead to single-threaded workloads. By the end of 2021, the company expects over 400 cloud providers including Dell, HPC, Tencent, Microsoft, Google, Lenovo, etc to adopt its 3rd Gen Epyc processors across more than 100 different platforms.
One of the most pressing issues in the semiconductor industry is the limited chip supply, rather than competition. In this regard, AMD has increased its 7nm orders at TSMC by as much as 80% YoY, surpassing Apple to become the foundry’s largest 7nm client. As Apple moves onto the 3nm process node from 5nm, the same will be the case with the 5nm capacity too. The Zen 4-based processors from AMD which are slated to launch in 2023 are going to leverage the 5nm EUV node, once again bringing an unprecedented increase in performance across every market segment.
Thanks to the node shrink, we’re also likely to see an increased core count in the server segment, with sources indicating an upper limit of 96 cores for the 4th Gen Epyc lineup (Genoa). It’s unclear whether the consumer CPU market will also have their core counts bumped up, but considering that Intel’s 12th-Gen Alder Lake hybrid CPUs will pack up to 16 heterogeneous cores, AMD may decide to offer up to 24 cores with the top-end Ryzen SKU.
