NVIDIA A100 40GB HBM2 PCIe Tensor Core GPU
699-21001-0200Available to order
NVIDIA A100 Tensor Core GPU accelerator, 40GB of HBM2 on a 5120-bit interface delivering up to 1,555 GB/s. Built on the Ampere GA100 processor with third-generation Tensor cores for TF32, BFLOAT16, FP16, INT8 and INT4, plus double-precision Tensor core acceleration. A 250W passively cooled full-height full-length 10.5-inch dual-slot PCIe 4.0 x16 card requiring system airflow, with Multi-Instance GPU partitioning into up to seven instances, SR-IOV across 16 virtual functions and ECC enabled by default.
- Architecture
- NVIDIA Ampere
- GPU SKU
- GA100-883AA-A1
- Base Clock
- 765 MHz
- Boost Clock
- 1,410 MHz
- GPU Boost
- Supported
- GPU Memory
- 40GB HBM2
- Memory Clock
- 1,215 MHz
- Memory Bus Width
- 5,120-bit
- Memory Bandwidth
- 1,555 GB/s
- ECC
- On by default
- Total Board Power
- 250 W
- Thermal Solution
- Passive
- Form Factor
- FHFL 10.5", dual-slot
- PCI Express
- PCIe 4.0 x16
- Power Connector
- 1× CPU 8-pin
- Board Weight
- 1,240 g
- Multi-Instance GPU
- Up to 7 instances
- SR-IOV
- 16 virtual functions
- vGPU Software
- vGPU 11.x or later
- NVLink
- Bridge supported
- CUDA Support
- CUDA 11.x or later
- Driver Branch
- R450.x or later
- NGC-Ready
- 2.x or later
- Precision
- FP64, FP32, FP16
- Interrupts
- MSI-X
- ARI Forwarding
- Supported
GPU
Memory
Board
Virtualization
Software
Priceavailable on request
Get a quote