Back to all GPUs

NVIDIA A100 80GB SXM Cloud GPU Specs

HexGrid Cloud rents the NVIDIA A100 80GB SXM at $1.49 per GPU hour, billed per minute — about 12% below the $1.70/hr average across 2 other clouds. The A100 SXM carries 80GB of HBM2e memory on the Ampere architecture.

80GB VRAMAmpereDeploys in ~1 minSecure Cloud

HexGrid Cloud price

$1.49/hr

Billed per minute

VRAM

80GB

HBM2e

FP32

19.5 TFLOPS

Compute performance

Market average

$1.70/hr

You save 12%

Running the A100 SXM on HexGrid Cloud

What you get when you deploy with us, beyond the hourly rate.

Live in under a minute

Pick a region, hit deploy, and SSH in. No quota requests, no sales call.

Per-minute billing

You pay for the minutes you use. Stop the instance and billing stops with it.

Ready for training day one

CUDA, PyTorch and the usual drivers are preinstalled, or bring your own image.

No egress charges

Move checkpoints and datasets out without a surprise bandwidth bill.

A100 SXM pricing across the market

Published on-demand rates from other clouds, so you can see where our A100 SXM price sits.

Market average $1.70/hr across 2 clouds

12% below market average
  • HexGrid CloudYou're here$1.49/hr
  • RunPod$1.59/hr
  • Lambda AI$1.80/hr

Other providers' rates are their published on-demand list prices. They exclude committed-use discounts and can change at any time. Shown for reference only.

About the NVIDIA A100 80GB SXM

What the A100 SXM is built for, and where it falls short.

The NVIDIA A100 80GB SXM is the high-bandwidth SXM4 version of NVIDIA's Ampere datacenter accelerator. It combines 80GB HBM2e memory, NVLink, MIG, Tensor Cores, and strong FP64 performance for large-scale training, HPC, and multi-GPU workloads.

Best suited for

The SXM version of the A100 offers higher memory bandwidth and is designed for tightly interconnected multi-GPU servers.

  • Distributed AI training
  • LLM fine-tuning
  • LLM inference
  • HPC
  • Scientific computing
  • Data analytics

Strengths

  • 80GB HBM2e
  • 2,039 GB/s memory bandwidth
  • 600 GB/s NVLink
  • MIG support
  • Strong multi-GPU scaling

Limitations

  • No FP8 acceleration
  • Older than Hopper and Blackwell
  • Requires SXM server infrastructure

NVIDIA A100 80GB SXM specifications

Full technical specifications for the NVIDIA A100 80GB SXM.

VendorNVIDIAArchitectureAmpereChipGA100CategoryDatacenter GPUReleased2020Form factorSXM4InterfaceSXM4

Memory

Memory
80 GB
Type
HBM2e
Bandwidth
2,039 GB/s
Bus width
5120-bit
ECC
Yes

Compute

CUDA cores
6,912
Tensor cores
432
Tensor generation
Gen 3
Boost clock
1,410 MHz

AI & compute performance

FP64
9.7 TFLOPS
FP32
19.5 TFLOPS
TF32 Tensor
156 TFLOPS / 312 TFLOPS sparse
BF16 Tensor
312 TFLOPS / 624 TFLOPS sparse
FP16 Tensor
312 TFLOPS / 624 TFLOPS sparse
INT8 Tensor
624 TOPS / 1,248 TOPS sparse

Power & physical

TDP
400 W
Form factor
SXM4
Cooling
Server-cooled

Interconnect

Interface
SXM4
NVLink
Yes
NVLink bandwidth
600 GB/s
Infinity Fabric
No

Virtualization

MIG
Yes
vGPU
Yes

Media engines

NVENC engines
0
NVDEC engines
5
AV1 encode
No
AV1 decode
Yes

Software

CUDA
Yes
ROCm
No

Precision support

FP64
Yes
FP32
Yes
TF32
Yes
BF16
Yes
FP16
Yes
FP8
No
INT8
Yes

A100 SXM frequently asked questions

Common questions about NVIDIA A100 80GB SXM pricing, specifications and workloads on HexGrid Cloud.

What is the difference between A100 PCIe and A100 SXM?

The SXM version provides higher memory bandwidth, a higher power envelope, and is designed for high-bandwidth NVLink-connected server platforms.

How much VRAM does A100 SXM have?

This configuration has 80GB of HBM2e memory.

Does A100 SXM support NVLink?

Yes. NVIDIA specifies up to 600 GB/s of NVLink bandwidth.

Does A100 SXM support MIG?

Yes. It supports up to seven MIG instances.

How much does A100 SXM cost per hour?

Pricing varies by cloud provider and multi-GPU configuration. HexGrid compares available A100 SXM rates.

Other GPUs on HexGrid Cloud

If the A100 SXM isn't the right fit, these are also available to deploy.

Browse every GPU we rent

Start training on the A100 SXM today

$1.49 per GPU hour, billed per minute, running in about a minute.

Specification sources