It would seem that NVIDIA’s next-gen RTX 40 series graphics cards will be fabbed on TSMC’s 5nm process node with EUV lithography. Furthermore, the Ampere lineup is expected to be succeeded by Ada Lovelace which like its predecessor is also a monolithic design while AMD’s competing RDNA 3 parts are slated to use an MCM design at the higher-end. Certain enthusiasts believe that the latter will have the upper hand this time around due to a sheer compute advantage, but considering how NVIDIA managed to squeeze over 10 thousand cores onto Samsung’s 8nm LPP node, I don’t see how it can’t fit twice the amount on TSMC’s much more advanced 5nm process node.
From what we’ve heard, Lovelace should easily offer twice as much performance as the contemporary Ampere parts, with an FP32 core count of up to 18,432. The AD102 flagship is expected to feature a total SM count of 144 distributed across 12 GPCs. That’s a 71% gain in raw compute performance over the GA102. (hitechgazette) Pair that with architectural and process refinements, and you’re easily looking at a gain of 2x.
There’s also some info on NVIDIA’s Hopper Data Center graphics cards which are also said to be based on an MCM design. According to @kopit7kimi, every GPC (Graphics Processing Cluster) in the GH100 will consist of three CPCs which in turn will feature three TPCs (Texture Processing Cluster), further sub-divided into two SMs. It’s worth noting that we are expecting a separate architecture for the gaming and data center markets, with Ada Lovelace coming to the former as a monolithic design and Hopper with its MCM design headed to the next-gen Tensor core accelerators.
The same source has stated that the volume production of Zen 3D processors (Zen 3 with 3D V-Cache) is expected to begin shortly, while the risk production/package validation of the 5nm Zen 4 chips has already started in China. It is worth noting that the volume production of Zen 3D was originally planned for Q4 2021, with Zen 4 expected to begin mass production sometime in Q2 2022.
That’s just a few quarters before NVIDIA plans on launching its RTX 40 series graphics cards based on the Ada Lovelace architecture. The Hopper GPUs with an MCM design aimed at data centers are expected to be announced earlier, most likely at GTC 2022, as we’ve seen with Volta and Ampere.