What makes NVIDIA's Blackwell B200 the most powerful AI chip ever built?
NVIDIA's Blackwell architecture packs 208 billion transistors into a single GPU, redefining what's possible for AI training.
The answer
Announced at GTC on March 18, 2024, NVIDIA's Blackwell B200 GPU is a dual-die processor containing 208 billion transistors connected by a 1 TB/s on-package interconnect. It delivers 9 petaFLOPS of FP4 AI compute, 192 GB of HBM3e memory with 8 TB/s bandwidth, and up to 30x the inference performance of the prior-generation H100. Its 5th-generation Tensor Cores introduced FP4 precision support, cutting memory usage by up to 3.5x while improving energy efficiency 25-50x. Despite consuming up to 1,000 watts, it offers 25x better energy efficiency per computation than Hopper.
Source: NVIDIA Newsroom
- NVIDIA
- Blackwell
- GPU
More from AI & Machine Learning
- What radical question did Alan Turing ask in 1950 that launched the entire field of artificial intelligence?
- Can a machine fool you into thinking it's human - and does that mean it's truly intelligent?
- What happened at a small Dartmouth workshop in 1956 that gave artificial intelligence its name?
- How did a 1958 invention called the Perceptron convince people that machines could learn like brains?
- Could a 1966 chatbot really make people believe they were talking to a therapist?
- What was the first robot that could 'reason' about its own actions?