Like rivals AMD and Intel, NVIDIA is also planning to transition to TSMC’s 5nm (N5) node with its next-gen GeForce RTX 40 series graphics cards sometime in 2022. The chipmaker switched to Samsung’s more mature (but cheaper) 8nm LPP node for its present generation Ampere graphics cards while AMD has been using TSMC’s 7nm node over the last 2-3 years. While this may have potentially prevented it from maximizing the performance of its RTX 30 series parts, it definitely helped improve the overall supply.
TSMC’s 5nm node is going to succeed its 7nm process, with all major customers set to take the leap sometime in 2022. While AMD and MediaTek get their share of wafers without paying any advance, NVIDIA and Qualcomm are required to make a hefty advanced payment to reserve their capacity.
TSMC’s 5nm node is 16% faster than the 7nm node at the same power or 14% more efficient at the same performance, with a density gain of around 70%. This will allow NVIDIA to offer the kind of performance uplift required to challenge AMD’s multi-chiplet RDNA 3 design.
Meanwhile, Intel is going to leverage TSMC’s 6nm node (7nm enhanced) for its 1st Gen Arc “Alchemist” graphics cards, and the 4nm (5nm enhanced) for its 2nd Gen Arc “Battlemage” lineup.
As per industry rumors, NVIDIA has paid TSMC $1.64 billion last quarter to reserve its share of the 5nm pie, with another $1.79 billion to be paid in the first quarter of 2022. Overall, the chipmaker will be spending $6.9 billion to secure its supply of 5nm wafers. These wafers will be used to fab AMD’s Ryzen 5000 CPUs, Radeon RX 7000 GPUs, MediaTek, and Qualcomm’s mobile SoCs in addition to NVIDIA’s RTX 40 series. (q2stadium.com) This might lead to a bottleneck in the supply chain at the very top, but TSMC has assured its partners that plenty of capacity has been ensured.