Nvidia unveils Blackwell (Perplexity AI summary)
NVIDIA has introduced the Blackwell platform, marking a significant advancement in computing technology. The Blackwell platform, featuring the new Blackwell GPU, is designed to power a new era of computing, enabling organizations to build and run AI models with up to trillion parameters. This innovation is expected to unlock breakthroughs in various fields, including data processing, engineering simulation, electronic design automation, computer-aided drug design, quantum computing, and generative AI.
The Blackwell GPU architecture incorporates six transformative technologies for accelerated computing. Among the notable advancements are new Tensor Cores and a TensorRT-LLM Compiler, which together reduce the operating cost and energy consumption of LLM inference by up to 25 times. This makes the Blackwell platform a highly efficient solution for running large-scale AI models.
The platform has garnered widespread support from major cloud providers, server makers, and leading AI companies, including Amazon Web Services, Dell Technologies, Google, Meta, Microsoft, OpenAI, Oracle, Tesla, and xAI. This broad adoption underscores the platform's potential to revolutionize computing across industries[1].
One of the key features of the Blackwell architecture is its ability to perform matrix math with floating point numbers just 4 bits wide, offering unprecedented precision and efficiency. The architecture also includes the fifth generation of NVIDIA's computer interconnect technology, NVLink, which delivers 1.8 terabytes per second bidirectionally between GPUs. This high-speed communication capability is crucial for building large-scale computers capable of processing trillion-parameter neural network models.
The DGX SuperPOD, powered by NVIDIA GB200 Grace Blackwell Superchips, represents the next generation of AI supercomputing. It is designed for generative AI supercomputing at a trillion-parameter scale, showcasing the immense computational power of the Blackwell platform.
NVIDIA's unveiling of the next-gen Blackwell GPUs highlights a significant reduction in costs and energy consumption for AI processing tasks. The GB200 Grace Blackwell Superchip, which consists of multiple chips in a single package, promises up to 30 times performance increase for LLM inference workloads compared to previous iterations. This advancement is set to herald a transformative era in computing, with potential applications extending beyond traditional fields to include gaming products in the future.
