L40S
nvidia · ada · 48 GB GDDR6 · 350W TDP
VRAM
48 GB
BF16 TFLOPS
362
Bandwidth
864 GB/s
From
$0.85/hr
Spec Sheet
Pricing by Provider
| Provider | On-Demand | Reserved | Spot | Badge |
|---|---|---|---|---|
| fluidstack | $1.09/hr | - | $0.85/hr | Cheapest |
| tensordock | $1.19/hr | - | $0.89/hr | |
| vast_ai | $1.29/hr | - | $0.95/hr | |
| lambda | $1.59/hr | $1.19/hr | - | |
| coreweave | $1.84/hr | $1.34/hr | - | |
| runpod | $1.90/hr | - | $1.49/hr | |
| gcp | $2.45/hr | $1.62/hr | - | |
| aws | $2.56/hr | $1.69/hr | - |
Pricing History
Compatible Models (267)
Single GPU (198 models)
Multi-GPU (69 models)
Training Capabilities
Estimated GPU count for full fine-tuning (AdamW, BF16) and QLoRA
| Model Size | Full Fine-Tune | QLoRA |
|---|---|---|
| 7B model | 3 GPUs | 1 GPU |
| 13B model | 6 GPUs | 1 GPU |
| 70B model | 28 GPUs | 1 GPU |
Energy Efficiency
Estimated tokens/second per Watt for popular models
Similar GPUs
| GPU | VRAM | BF16 TFLOPS | BW (GB/s) | From |
|---|---|---|---|---|
| L40 | 48 GB | 362 | 864 | $0.75/hr |
| RTX 6000 Ada | 48 GB | 91.1 | 960 | $0.59/hr |
| L20 | 48 GB | 239 | 864 | $0.80/hr |
| L4 | 24 GB | 121 | 300 | $0.29/hr |
| RTX 4090 | 24 GB | 165 | 1008 | $0.39/hr |
Methodology Note
Performance estimates for the L40Sare based on InferenceBench's roofline performance model with CUDA kernel-level optimization including FlashAttention v2 and PagedAttention. Memory calculations account for model weights (48 GB GDDR6 available), KV-cache allocation, and activation memory. Throughput predictions use the L40S's rated 864 GB/s memory bandwidth and 362 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 L40S?
The L40S can run 267 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 L40S.
What is the L40S inference throughput?
The L40S delivers 362 BF16 TFLOPS and 733 FP8 TFLOPS with 864 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 L40S cost per hour?
The L40S is available starting from $0.85/hour via fluidstack. Prices vary by provider and pricing tier (on-demand, reserved, spot). Compare pricing across all providers in the table above.