NVIDIA unveils more details about the Blackwell architecture: capable of supporting models with up to 100 trillion parameters for AI training and real-time large language model inference

NVIDIA has released more details about the next-generation GPU architecture Blackwell, as well as the future product roadmap. NVIDIA Blackwell is the ultimate solution for general computing full-stack matrix, consisting of multiple NVIDIA chips, including Blackwell GPU, Grace CPU, BlueField data processing unit, ConnectX network interface card, NVLink switch, Spectrum Ethernet switch, and Quantum InfiniBand switch. NVIDIA claims that Blackwell has 6 revolutionary technologies that can support AI training for models with up to 10 trillion parameters and real-time large language model inference