Nvidia A100 80GB Tensor Core GPU
The Nvidia A100 80GB accelerator is built on the groundbreaking Ampere computing architecture, delivering unprecedented performance for demanding AI, data analytics, and high-performance computing (HPC) workloads.
Products Details
- Brand: Nvidia
- Part Number: A100
- CUDA: Core Technology
- Direct: Compute
- Open: CL
- Open: ACC
Processing Core Specifications
- Engine Architecture: Ampere
- CUDA® Cores: 6,912
- Tensor Cores: 432 (3rd Generation)
- Base GPU Clock: 1,065 MHz
- Boost Clock Speed: 1,410 MHz
Computational Throughput and Precision
- Double Precision (FP64): 9.7 TFLOPS / 19.5 TFLOPS (with Tensor Cores)
- Single Precision (FP32): 19.5 TFLOPS
- Tensor Float 32 (TF32): 156 TFLOPS
- Half Precision (FP16/BFLOAT16): 312 TFLOPS / 624 TFLOPS (with sparsity)
- Integer Precision (INT8/INT4): 624 TOPS / 1,248 TOPS & 1,248 TOPS / 2,496 TOPS respectively
Memory System Details
- Total Memory Capacity: 80 GB HBM2e
- Memory Interface Width: 5,120-bit
- Peak Memory Bandwidth: 1.94 TB/s
- Memory Clock Frequency: 1,512 MHz
- Error Correction Code (ECC): Supported and Enabled
Advanced System Interconnect Technology
High-Speed Data Transfers
- NVLink Transfer Speeds: 600 GB/s
Bus Support:
- PCI-Express 4.0
- Physical Bus Interface: PCIe 4.0 x16
Compatible Operating Systems
- Microsoft Windows 7, 8, 8.1, 10
- Microsoft Windows Server 2008 R2, 2016
- Linux (English US/UK)
