NVIDIA 900-21001-0020-100 A100 Tensor Core 80GB HBM2e PCI Express 4.0 x16 Dual Slot Full Height Accelerator Graphics Card

Massive Memory Capacity: With 80GB of high-bandwidth memory, the A100 PCIe GPU supports the most complex models and largest datasets, enabling faster results and improved throughput across AI and HPC workloads.

Advanced Compute Support: The A100 supports a wide range of math precisions including FP64 (double precision), FP32 (single precision), FP16 (half precision), and INT8 (integer), making it ideal for diverse workloads ranging from scientific computing to AI training and inference.

Unmatched Memory Bandwidth: Offering up to 1.94 TB/s of memory bandwidth, the A100 80GB PCIe sets a new standard for data transfer performance, significantly reducing bottlenecks in large-scale computing tasks.

High-Efficiency Cooling Design: The dual-slot, 10.5-inch PCIe Gen4 card features a passive heat sink, optimized for server airflow environments. It operates with a thermal design power (TDP) of 300W, ensuring peak performance in high-demand scenarios.

Multi-Instance GPU (MIG) Technology: The A100 can be partitioned into up to seven isolated GPU instances, allowing multiple users or tasks to run concurrently on a single card. This feature boosts GPU utilization and offers true elasticity for dynamic workloads in cloud-native and virtualized data centers.

Scalable Performance with NVLink: Using three NVIDIA® NVLink® bridges, multiple A100 80GB PCIe GPUs can be interconnected to achieve 600 GB/s total bandwidth, which is 10x faster than PCIe Gen4 alone. This enables massive scalability and accelerated performance for multi-GPU workloads.