NVIDIA A100
The NVIDIA A100 is a data-center GPU accelerator built on the NVIDIA Ampere architecture (GA100 die). It comes in 40 GB and 80 GB memory variants and in two form factors: an SXM4 module (mounted on HGX baseboards) and a PCIe card. Memory uses HBM2e.
It features third-generation Tensor Cores supporting TF32, BFLOAT16, FP16 and INT8, with acceleration for sparse data (sparsity). Multi-Instance GPU (MIG) lets a single processor be partitioned into up to seven independent instances. Interconnect is provided by NVLink (600 GB/s) and PCIe Gen4 (64 GB/s).
Classification
GPU (discrete graphics processor) · serves as: AI compute, AI acceleration, AI Inference.
Which group NVIDIA A100 belongs to and how it is built
This subcategory groups integrated circuits designed to accelerate AI computation in data centers: NVIDIA H100/B200, Google TPU, AWS Trainium/Inferentia, Meta MTIA, Groq LPU, Cerebras WSE. Common properties: high-bandwidth HBM memory, low-precision modes (FP16/BF16/FP8/MX8/MX4/INT8/INT4), scalability to rack-scale (hundreds of chips in a single scale-up domain), liquid cooling, integration with ML frameworks (PyTorch, JAX, Triton, vLLM). This contrasts with hardwareSubcategory.ai-soc-edge-ai-soc where the criteria are the opposite (low power, no HBM, on-device inference).
A massively parallel graphics processing unit (GPU) used for both graphics rendering and general-purpose compute (GPGPU), including neural network training and inference.
Technical specification
312 TFLOPS with sparsity.
624 TFLOPS with sparsity.
624 TFLOPS with sparsity.
1248 TOPS with sparsity.
624 TFLOPS with sparsity.
624 TFLOPS with sparsity.
1248 TOPS with sparsity.