Post by KostaAndreadis @ 04:21pm 15/05/20 | 2 Comments
But, the very first Ampere-powered tech unveiled by NVIDIA CEO Jensen Huang - the A100 GPU - is data-centre focused. That said, the company's Ampere GPU architecture will power all upcoming GeForce cards - we'll just have to wait and see how the next line of GeForce RTX cards makes use of its impressive technology.
And with the A100 utilising 7nm to great effect, it's been referred to as the greatest generational leap in the company's history. 7nm allows NVIDIA to get up to 20x the raw computing power over the previous NVIDIA Volta technology. It does so with specs like a staggering 54 billion transistor count, 3,456 FP64 CUDA Cores, 6,912 FP32 CUDA Cores, and 40GB of GPU Memory with 1.6 TB/s of bandwidth. On a huge 826mm² die size.
Now that may seem like a random selection of numbers, or NVIDIA going all out for data centre dominance - but it could mean the A100 is a blueprint for GeForce moving forward. Of course, in terms of size and what it's packing, the A100 is not a GeForce RTX 3080 by any stretch. But its unveiling kind of confirms that NVIDIA will be using 7nm and all the advances made with Turing (i.e. real-time ray-tracing and AI-based rendering) for the next range of RTX cards.
Rumours of massive performance gains for NVIDIA's next-generation of GeForce cards looks to be coming true - we're just hoping that whenever they do show up there's enough time for us install one before we install Cyberpunk 2077.
As an AI and data industry presentation, Jensen Huang and his formidable leather jacket (even in the quiet setting of a home kitchen) still managed to slip in one RTX graphics demo. A real-time ray-traced version of Marble Madness running on the company's very expensive NVIDIA Quadro RTX 8000.
Running in real-time this is a fully playable game demo created by a few NVIDIA engineers. Visually it looks amazing and showcases where the future of RTX and real-time ray-tracing is headed.