This generation of RTX 40 series graphics cards uses a new Ada Lovelace architecture GPU core, while the process upgrades to TSMC 4N 5nm process. Compared to the AMD Radeon graphics cards of the same period, NVIDIA has been a bit “uninspired” in the process over the past few years, first 12nm, then 8nm, and now only to 5nm, and not the more advantageous 4nm.
I didn’t expect NVIDIA CEO Jen-Hsun Huang to look more disappointed than the gamers.
Editor Dave Altavilla mentioned at the RTX 40 Series graphics card press conference that the upgrade from Samsung 8N to TSMC 4N was impressive, so what gains did the new process bring to the Ada architecture?
Lao Huang’s answer was somewhat surprising, from Samsung 8N to TSMC 4N, the microscopic level of improvement is about 15%, but very, unfortunately, the cost increase is more than 15%.
Huang Jen-Hsun also said that roughly from 7nm, Moore’s Law curve began to slow down, although not yet to a stop-start straight line, but has actually failed.
Huang went on to point out that as process technology approaches its physical limits, the complexity of production has increased dramatically, and the time to complete all the steps in the Fab (fab) has ballooned to as long as four months. Lao Huang admits that this is also not TSMC’s intention to delay or want to squeeze profits, but really needs so many steps, and the equipment is also more expensive.
