NVIDIA A100 specifications (web)
Citation
NVIDIA. “NVIDIA A100 Tensor Core GPU.” Product web page with specifications, date not stated. https://www.nvidia.com/en-us/data-center/a100/. Read 2026-11-09.
Bottom line
- The A100 is NVIDIA’s Ampere-generation data center GPU. The 80GB SXM version has 80GB of HBM2e memory, 2,039 GB/s of memory bandwidth and a 400W maximum power.
- One A100 can be split into up to seven Multi-Instance GPU (MIG) instances for different users.
Background
NVIDIA makes GPUs for AI and high-performance computing (HPC). This is its product page for the A100, with performance claims and a specifications table for the 80GB PCIe and 80GB SXM versions. The page gives no price.
Key claims
- “Powered by the NVIDIA Ampere Architecture, A100 is the engine of the NVIDIA data center platform.” (Accelerating the Most Important Work of Our Time)
- Specifications, PCIe then SXM: ”| GPU Memory | 80GB HBM2e | 80GB HBM2e |”, ”| GPU Memory Bandwidth | 1,935 GB/s | 2,039 GB/s |”, ”| Max Thermal Design Power (TDP) | 300W | 400W *** |” (Specifications).
- Specifications: ”| FP64 | 9.7 TFLOPS |”, ”| FP16 Tensor Core | 312 TFLOPS | 624 TFLOPS* |”, where ”* With sparsity” (Specifications).
- “With MIG, an A100 GPU can be partitioned into as many as seven independent instances, giving multiple users access to GPU acceleration.” (Enterprise-Ready Utilization)
- “Server Options … NVIDIA DGX™ A100 with 8 GPUs” (Specifications, SXM column).
Evidence and method
A vendor’s product page. The performance claims are NVIDIA’s own benchmarks, with test settings in footnotes. The table gives rated figures, not measured ones.
Limits
- No price or cost of ownership.
- The A100 is an older generation. See NVIDIA H100 specifications (web) and NVIDIA H200 specifications (web).