| Product SKU | A100-SXM4-80GB |
|---|---|
| GPU Memory | 80GB HBM2e |
| GPU Memory Bandwidth | 2,039 GB/s |
| Max Thermal Design Power (TDP) | 400W |
| Multi-Instance GPU | Up to 7 MIGs @ 10GB |
| Form Factor | SXM |
| Interconnect | NVLink: 600 GB/s PCIe Gen4: 64 GB/s |
| Server Options | NVIDIA HGX™ A100-Partner and NVIDIA-Certified Systems with 4,8, or 16 GPUs NVIDIA DGX™ A100 with 8 GPUs |
Data Center GPUs, GPU
NVIDIA A100 SXM4 80GB
The NVIDIA A100 SXM4 80GB is a data-center GPU, with 80GB HBM2e memory and 2,039 GB/s of memory bandwidth. It has a board power of 400 W in a SXM form factor.
Overview
The NVIDIA A100 SXM4 80GB is a data-center GPU on the NVIDIA Ampere architecture with 80 GB of HBM2e and 2,039 GB/s of memory bandwidth. Its SXM4 form factor and 600 GB/s NVLink interconnect are designed for four- and eight-GPU HGX A100 baseboards, where the GPUs communicate as one node for training and HPC.
At a glance
- Memory: 80 GB HBM2e at 2,039 GB/s
- Interconnect: NVLink 600 GB/s; PCIe Gen4 64 GB/s
- Form factor: SXM4; max TDP 400 W
- Multi-Instance GPU: up to 7 MIG instances of 10 GB
- Platforms: NVIDIA HGX A100 (4, 8 or 16 GPUs); DGX A100 (8 GPUs)
When to choose the A100 SXM4
The A100 SXM4 remains a workhorse for AI training, inference and HPC where the platform is an HGX A100 baseboard and the workload benefits from 600 GB/s NVLink and 80 GB of HBM2e. It is a cost-effective choice for established SXM deployments and workloads that do not require the newest generation.
Platform & compatibility
A100 SXM4 modules mount on NVIDIA HGX A100 partner systems with four, eight or sixteen GPUs and on DGX A100 systems with eight GPUs; they are not a standalone add-in card. Multi-Instance GPU supports up to seven 10 GB instances.
FAQ
Is it a PCIe card? No — it is an SXM4 module for HGX A100 / DGX A100 baseboards. For PCIe, see the A100 PCIe.
How much memory does it have? 80 GB of HBM2e at 2,039 GB/s.
Can it be shared? Yes — up to seven MIG instances of 10 GB.
Related
- All NVIDIA data-center GPUs
- Data-center GPU buying guide: A100 to B300
- Browse all GPUs and accelerators
Configuration, availability and lead time depend on the required quantity, server pairing and delivery destination. Send your requirement to receive a quotation.
