After NVIDIA’s launch of turing architecture, their product line was actually two different architectures in parallel, with the Volta architecture-based GV100 being the main focus for high-performance computing, while the Turing architecture-based TU10X-based chip was a major role in the gaming graphics and professional graphics market. But that’s coming to an end soon, as nVIDIA’s CEO said at a press conference in front of GTC:
Ampere will eventually replace NVIDIA’s Turing and Volta chips, which will support NVIDIA’s GPU product line as a single platform.
Needless to say, this is the first time we’ve unified the acceleration load across the data center to a single platform.
Starting from Pascal architecture, different models of chips have diverged in the internal architecture, such as GP100 for high-performance computing and GP102 for game graphics, GP104, etc., in the internal architecture is quite different. This was followed by a generation of Volta architectures focused primarily on the computing market and a generation of Turing architectures focused on the graphics market.
But on Ampere, it seems that NVIDIA is about to re-enable the Pascal-era approach, reintegrating both computing and graphics into one architecture. Distinguish ingress by different core configurations. Although NVIDIA didn’t release any news that consumer-grade GPUs would use Ampere, its CEO, Huang Renxun, confirmed this, and when asked by a reporter how much the ampere architecture differs between enterprise and consumer levels, he replied that there would be a lot of architectural coincident, but not in configuration.
So, we’re now able to get a first look at the specifications of the Ampere game card that will be on the market in the future.