Blackwell is NVIDIA's high-performance GPU architecture designed specifically to accelerate trillion-parameter large language models, offering massive throughput improvements for AI training and inference workloads.
Directly governs the hardware efficiency and hardware-level token throughput when deploying ai supercomputing clusters, server gpu acceleration, and hardware scaling; optimizing Blackwell is a major factor in compute cost budgeting.
Blackwell is the codename for NVIDIA's next-generation graphics processing unit (GPU) architecture specifically designed for generative AI and trillion-parameter scale models. Named after mathematician David Blackwell, the architecture features dual-die GPUs, high-speed NVLink interconnects, and dedicated transformer engines. Blackwell delivers massive improvements in computational density, energy efficiency, and throughput for demanding LLM training and inference workloads.
Blackwell features dual-die architecture, high-bandwidth interconnects, and an advanced engine delivering up to 30x faster inference and 4x faster training compared to Hopper.
NVLink is NVIDIA's high-speed GPU-to-GPU interconnect technology, allowing high-density Blackwell clusters to communicate as a single massive GPU.
Every breakthrough AI model starts the same way: with a training run. The infrastructure running those training jobs shapes everything: how fast teams can...
AgentPerf from Artificial Analysis, the industry's first agentic AI benchmark, gives developers, enterprises and infrastructure providers a clear way to...