Infrastructure
NVIDIA A100
NVIDIA's 2020 datacentre GPU — 40/80GB HBM2e, 312 TFLOPS FP16, the workhorse of pre-2024 AI.
Definition
The A100 was NVIDIA's flagship from 2020 to 2023 and trained nearly every major LLM up to that point. Its 80GB HBM2e variant remains widely deployed and is the baseline against which H100 and B200 performance are reported.
Common use cases
- Training
- Inference
- Research clusters