Nvidia CEO talks with TSMC about 3nm capacity

NVIDIA’s new generation of GeForce RTX 40 series GPUs based on the Ada Lovelace architecture uses a 4nm process customized by TSMC, which has a higher energy consumption ratio than the previous generation. While the GeForce RTX 40-series graphics cards have only just been released, Nvidia seems to be preparing for the next step.

RTX 4090 Founders Edition

Image: Nvidia

According to Wccftech, Nvidia founder and CEO Mr. Jensen Huang recently visited Taiwan to discuss with TSMC about 3nm production capacity to ensure that next-generation chips use new process nodes. It is expected to wait until 2024 to see related products. Jensen Huang also visited the partners of the board, mainly for the inventory problem of GeForce RTX 30 series graphics cards, to help them clear the inventory.

Nvidia has also recently launched a number of GeForce RTX 30 series products, including GeForce RTX 3070 Ti with GA102, GeForce RTX 3060 Ti (GA104) with upgraded GDDR6X memory, and GeForce RTX 3060 (GA106) with 8GB of memory. For board manufacturers, it can speed up inventory clearance so that they can focus on the new GeForce RTX 40 series products as soon as possible. There have been recent reports that Nvidia is taking some of the marketing costs for some brands by canceling the GeForce RTX 4080 12GB that was released but not yet sold.

Rival AMD has similar plans. It has been reported before that AMD CEO Dr. Lisa Su will lead a group of executives and plan to visit relevant partners from the end of September to the beginning of November this year, mainly involving chip manufacturing, packaging, and PC manufacturers. During this period, Lisa Su will meet with Wei, CEO of TSMC. The main topic of discussion between the two parties is the future cooperation plan of TSMC’s 2nm/3nm process node, including various available processes. It is rumored that AMD intends to use the N3P process.