A100 40GB SXM
nvidia · ampere · 40 GB HBM2e · 400W TDP
VRAM
40 GB
BF16 TFLOPS
312
Bandwidth
1555 GB/s
From
$0.85/hr
Spec Sheet
Pricing by Provider
| Provider | On-Demand | Reserved | Spot | Badge |
|---|---|---|---|---|
| tensordock | $1.19/hr | - | $0.85/hr | Cheapest |
| vast_ai | $1.30/hr | - | $0.89/hr | |
| runpod | $1.64/hr | - | $1.19/hr | |
| lambda | $1.29/hr | - | - | |
| aws | $3.06/hr | $1.96/hr | - | |
| gcp | $2.93/hr | $1.98/hr | - |
Compatible Models (265)
Single GPU (190 models)
Multi-GPU (75 models)
Training Capabilities
Estimated GPU count for full fine-tuning (AdamW, BF16) and QLoRA
| Model Size | Full Fine-Tune | QLoRA |
|---|---|---|
| 7B model | 4 GPUs | 1 GPU |
| 13B model | 7 GPUs | 1 GPU |
| 70B model | 33 GPUs | 2 GPUs |
Energy Efficiency
Estimated tokens/second per Watt for popular models
Similar GPUs
| GPU | VRAM | BF16 TFLOPS | BW (GB/s) | From |
|---|---|---|---|---|
| A100 40GB PCIe | 40 GB | 312 | 1555 | $0.69/hr |
| RTX A6000 | 48 GB | 38.7 | 768 | $0.49/hr |
| A40 | 48 GB | 37.4 | 696 | $0.42/hr |
| A10G | 24 GB | 35 | 600 | $0.30/hr |
| A30 | 24 GB | 165 | 933 | $0.35/hr |
Methodology Note
Performance estimates for the A100 40GB SXMare based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (40 GB HBM2e available), KV-cache allocation, and activation memory. Throughput predictions use the A100 40GB SXM's rated 1555 GB/s memory bandwidth and 312 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (ampere). See our full methodology.
Frequently Asked Questions
How many AI models can run on A100 40GB SXM?
The A100 40GB SXM can run 265 AI models from our database within a single node. Compatible models range across various parameter sizes depending on the quantization precision (BF16, FP8, INT4). Smaller models fit on a single GPU while larger models may require multi-GPU setups up to 8x A100 40GB SXM.
What is the A100 40GB SXM inference throughput?
The A100 40GB SXM delivers 312 BF16 TFLOPS and 312 FP8 TFLOPS with 1555 GB/s memory bandwidth. Actual inference throughput (tokens/sec) depends on the model size, precision, and batch size. Use our calculator for model-specific throughput estimates.
How much does A100 40GB SXM cost per hour?
The A100 40GB SXM is available starting from $0.85/hour via tensordock. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.