NVIDIA’s GeForce RTX 30 series graphics cards launched roughly a year ago, with most of their cycle plagued by semiconductor shortages. Even today, it’s impossible to find an Ampere part at or around the MSRP as the chip shortages extend into 2022. According to the latest rumors out of China, the supply of chips from Samsung’s 8nm foundries continues to be unstable.
While AMD decided to leverage TSMC’s 7nm (N7) node for both its 1st Gen and 2nd Gen RDNA graphics cards, NVIDIA went with the more mature but cheaper 8nm LPP process (from Samsung). We had heard earlier during Ampere’s product cycle that poor yields were contributing to GPU shortages, but that was several months back. If 8nm yields are still sub-par, then things are rather grim for Samsung’s foundries.
A while back, news broke out that NVIDIA was limiting the production of its Ampere graphics cards, most notably the higher-end SKUs to prevent a sudden drop in prices as supply stabilized. It’s unclear whether the two situations are related, but there’s a good chance that they’re interrelated, rather than being two isolated events.
NVIDIA’s next-gen Ada Lovelace lineup is expected to launch in the third quarter of 2022 alongside AMD’s Radeon RX 7000 series. Both the architectures will be fabbed on TSMC’s 5nm EUV (N5) process node, with core counts exceeding 15,000, and TDPs of more than 500W at the top-end. By that time, Ethereum 2.0 should also have been implemented, significantly reducing the revenue from GPU mining.