TSMC has announced that it's using AMD EPYC processors (opens in new tab) to power workloads for its manufacturing business. Yup, TSMC is using chips it has produced to make the chips it produces. To be fair, this is a recent upgrade to its data centers, so it's not like it had to partake in time travel to achieve this seemingly paradoxical situation.
In case it wasn't obvious, TSMC is not a small company—over 50,000 people work for the manufacturing giant, across sites in Taiwan, Europe, Japan, China, and the United States. TSMC is using HPE's DL324 G10 platform, which runs 2nd-gen AMD EPYC 7702 (opens in new tab) CPUs with 64 cores, a base clock of 2GHz, and a boost of 3.35GHz.
While TSMC and AMD already have a working relationship, apparently it wasn't a given that it would use AMD's hardware for its own servers, and had to show benefits over the competition. Something that it apparently managed due to the excellent power efficiency AMD EPYC offers thanks to TSMC's 7nm production process.
TSMC has clearly been impressed with its experiences with EPYC so far and plans to implement two more data centers in 2021. It's also looking at the new EPYC 7F72 CPU for its manufacturing teams. These CPUs have 24 cores and a base clock of 3.2GHz, with the higher clock speeds being the clincher here.
A closer working relationship between AMD and TSMC can only mean good things for AMD going forward, although we doubt this would directly lead to more preferential treatment. Although anything to get more silicon out of TSMC would be welcome—particularly GPUs, which have been almost impossible to track down since the original launch.