Like rivals AMD and Intel, NVIDIA can also be planning to transition to TSMC’s 5nm (N5) node with its next-gen GeForce RTX 40 collection graphics playing cards someday in 2022. The chipmaker switched to Samsung’s extra mature (however cheaper) 8nm LPP node for its current era Ampere graphics playing cards whereas AMD has been utilizing TSMC’s 7nm node over the past 2-3 years. Whereas this will have probably prevented it from maximizing the efficiency of its RTX 30 collection elements, it undoubtedly helped enhance the general provide.
TSMC’s 5nm node goes to succeed its 7nm course of, with all main clients set to take the leap someday in 2022. Whereas AMD and MediaTek get their share of wafers with out paying any advance, NVIDIA and Qualcomm are required to make a hefty superior fee to order their capability.
TSMC’s 5nm node is 16% quicker than the 7nm node on the similar energy or 14% extra environment friendly on the similar efficiency, with a density achieve of round 70%. This can permit NVIDIA to supply the form of efficiency uplift required to problem AMD’s multi-chiplet RDNA 3 design.
In the meantime, Intel goes to leverage TSMC’s 6nm node (7nm enhanced) for its 1st Gen Arc “Alchemist” graphics playing cards, and the 4nm (5nm enhanced) for its 2nd Gen Arc “Battlemage” lineup.
As per business rumors, NVIDIA has paid TSMC $1.64 billion final quarter to order its share of the 5nm pie, with one other $1.79 billion to be paid within the first quarter of 2022. Total, the chipmaker will likely be spending $6.9 billion to safe its provide of 5nm wafers. These wafers will likely be used to fab AMD’s Ryzen 5000 CPUs, Radeon RX 7000 GPUs, MediaTek, and Qualcomm’s cellular SoCs along with NVIDIA’s RTX 40 collection. This may result in a bottleneck within the provide chain on the very high, however TSMC has assured its companions that loads of capability has been ensured.