RTX 4070 Super
nvidia · ada · 12 GB GDDR6X · 220W TDP
VRAM
12 GB
BF16 TFLOPS
55
Bandwidth
504 GB/s
From
$0.22/hr
Spec Sheet
Pricing by Provider
| Provider | On-Demand | Reserved | Spot | Badge |
|---|---|---|---|---|
| tensordock | $0.35/hr | - | $0.22/hr | Cheapest |
| vast_ai | $0.40/hr | - | $0.25/hr |
Compatible Models (226)
Single GPU (129 models)
Multi-GPU (97 models)
Training Capabilities
Estimated GPU count for full fine-tuning (AdamW, BF16) and QLoRA
| Model Size | Full Fine-Tune | QLoRA |
|---|---|---|
| 7B model | 11 GPUs | 1 GPU |
| 13B model | 21 GPUs | 1 GPU |
| 70B model | 110 GPUs | 4 GPUs |
Energy Efficiency
Estimated tokens/second per Watt for popular models
Similar GPUs
| GPU | VRAM | BF16 TFLOPS | BW (GB/s) | From |
|---|---|---|---|---|
| RTX 4070 Ti | 12 GB | 93 | 504 | $0.25/hr |
| RTX 4080 | 16 GB | 97 | 717 | $0.32/hr |
| RTX 4060 Ti 16GB | 16 GB | 44 | 288 | $0.30/hr |
| RTX 4060 | 8 GB | 30 | 272 | $0.22/hr |
| L4 | 24 GB | 121 | 300 | $0.29/hr |
Methodology Note
Performance estimates for the RTX 4070 Superare based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (12 GB GDDR6X available), KV-cache allocation, and activation memory. Throughput predictions use the RTX 4070 Super's rated 504 GB/s memory bandwidth and 55 BF16 TFLOPS compute capacity as roofline ceilings, with empirical correction factors per GPU architecture (ada). See our full methodology.
Frequently Asked Questions
How many AI models can run on RTX 4070 Super?
The RTX 4070 Super can run 226 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 4x RTX 4070 Super.
What is the RTX 4070 Super inference throughput?
The RTX 4070 Super delivers 55 BF16 TFLOPS and 110 FP8 TFLOPS with 504 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 RTX 4070 Super cost per hour?
The RTX 4070 Super is available starting from $0.22/hour via tensordock. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.