Just yesterday, we learned that NVIDIA’s next-gen Ampere graphics cards would be based on TSMC’s 7nm node, rather than Samsung’s 8nm process. While this is a big blow for Samsung, sources were optimistic that the Hopper lineup succeeding Ampere might return to Samsung foundries, utilizing the upcoming 5nm EUV process. However, as per a Digitimes report, NVIDIA will continue to work with TSMC even with Hopper, leveraging the foundry’s 5nm node. This means that for the first time in a very long time, both NVIDIA and AMD’s competing graphics cards would be based on the same process node.
AMD’s Navi 2x GPUs powered by the RDNA 2 architecture is expected to launch in the coming months. The new Radeon graphics cards are going to use TSMC’s 7nm process and bring hardware-level ray-tracing and Variable Rate Shading to AMD hardware. These GPUs will go up against NVIDIA’s Ampere graphics cards, also based on the 7nm process. It’s not clear whether both lineups will use the same process node or whether AMD will be leveraging a more mature version.
Regardless, towards 2022, both NVIDIA and AMD will be shift to TSMC’s 5nm process, essentially bringing node parity to the graphics card market. By that time, Intel’s Xe graphics cards will also most likely be available in the consumer market, though I’m not sure if they’ll be worth the attention.
This basically means that Samsung’s foundries are in for a hard time. It’s possible that NVIDIA or AMD might leverage Samsung’s 5nm EUV process for certain lower-end parts, but it looks like the bulk of orders will go to TSMC.