Let’s put that into perspective, shall we? The current-generation GeForce RTX 3090 is the fastest consumer card NVIDIA has ever launched. It is built around a GA102 Ampere GPU with 72 streaming multiprocessors and 10,496 CUDA cores, along with 328 texture mapping units, 112 render output units, 328 Tensor cores, and 82 dedicated ray tracing cores.
- GeForce RTX 3090: 72 SMs, 10,496 CUDA cores
- GeForce RTX 3080: 68 SMs, 8,704 CUDA cores
- GeForce RTX 3070: 46 SMs, 5,888 CUDA cores
- GeForce RTX 3060 Ti: 38 SMs, 4,864 CUDA cores
The same leaker has now suggested that Lovelace could wield a 18,432 CUDA cores. Have a look…
It sounds more like speculation than inside information, but given the leaker’s track record, it could speculation based on inside information. Unfortunately @kopite7kimi does not elaborate on the presumed 12*6 structure, but it is interesting all the same.
What this amounts to is a bump in graphics processing clusters (GPCs) to 12, versus a full-fat GA102 Ampere GPU having 7 GPCs. If that is accurate, Lovelace could potentially accommodate 72 texture processing clusters (TPCs) and 144 streaming multiprocessors, lending way to a whopping 18,432 CUDA cores.
That would be an absolute monster, if the information is accurate. And who really knows at this early stage. Bear in mind that Turing first launched in September 2018, while Ampere came out two years later. So this is an architecture that we might not see until the latter part of 2022.