A2
nvidia · ampere · 16 GB GDDR6 · 60W TDP
VRAM
16 GB
BF16 TFLOPS
18
Bandwidth
200 GB/s
From
$0.15/hr
Spec Sheet
Pricing by Provider
| Provider | On-Demand | Reserved | Spot | Badge |
|---|---|---|---|---|
| retail | $0.15/hr | - | - | Cheapest |
Compatible Models (175)
Single GPU (175 models)
Training Capabilities
Estimated GPU count for full fine-tuning (AdamW, BF16) and QLoRA
| Model Size | Full Fine-Tune | QLoRA |
|---|---|---|
| 7B model | 9 GPUs | 1 GPU |
| 13B model | 16 GPUs | 1 GPU |
| 70B model | 83 GPUs | 3 GPUs |
Energy Efficiency
Estimated tokens/second per Watt for popular models
Similar GPUs
Methodology Note
Performance estimates for the A2are based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (16 GB GDDR6 available), KV-cache allocation, and activation memory. Throughput predictions use the A2's rated 200 GB/s memory bandwidth and 18 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 A2?
The A2 can run 175 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 1x A2.
What is the A2 inference throughput?
The A2 delivers 18 BF16 TFLOPS and 36 FP8 TFLOPS with 200 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 A2 cost per hour?
The A2 is available starting from $0.15/hour via retail. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.