Specifications Compared
| Spec | L40 | L40S |
|---|---|---|
| TDP | 300W | 350W |
| VRAM | 48 GB | 48 GB |
| CUDA Cores | 18,176 | 18,176 |
| Memory Type | GDDR6 | GDDR6X |
| Architecture | Ada Lovelace | Ada Lovelace |
| Form Factors | PCIe | PCIe |
| Interconnect | PCIe 4.0 | |
| Tensor Cores | 568 | 568 |
| FP16 Performance | 90.5 TFLOPS | 362 TFLOPS |
| FP32 Performance | 90.5 TFLOPS | 91 TFLOPS |
| INT8 Performance | 724 TOPS | 724 TOPS |
| Memory Bandwidth | 864 GB/s | 864 GB/s |
Performance Analysis
The L40S outperforms the L40 significantly in half-precision computing: its 362 TFLOPS FP16 rate quadruples the L40's 90.5 TFLOPS, accelerating deep learning training and inference where models leverage mixed precision. FP32 performance remains comparable at 90.5 TFLOPS for the L40 and 91 TFLOPS for the L40S, suiting workloads like scientific simulations that demand single-precision accuracy. The L40S exclusive FP8 capability at 724 TFLOPS enables ultra-efficient quantized inference for large language models.
Identical 864 GB/s memory bandwidth on both GPUs means equivalent support for large batch sizes in training, preventing bottlenecks in data-heavy tasks. However, the L40S 350W TDP versus the L40 300W allows sustained higher performance under load, though it increases power and cooling demands. In real-world AI pipelines, the L40S handles 4x more FP16 operations per second, reducing training epochs for models like transformers.
For inference, the FP8 advantage on the L40S lowers latency in serving quantized models, while shared 48 GB VRAM ensures both manage billion-parameter LLMs without splitting.
Live Cloud Pricing
Real-time prices from 25+ providers. Updated every 60 seconds.
L40
| Provider | GPU Model | VRAM | Host Specs | Region | Price | Status | Action | |
|---|---|---|---|---|---|---|---|---|
 TensorDock | NVIDIA L40S 48GB VRAM | 48GB | 0 vCPU 0GB RAM | Wolverhampton | $0.55/GPU/hr | Available | ||
 RunPod | NVIDIA L40 48GB VRAM | 48GB | 8 vCPU 94GB RAM | 🌍global | $0.82/GPU/hr | |||
 Massed Compute | NVIDIA L40 48GB VRAM | 48GB | 14 vCPU 72GB RAM 625GB Storage | Iowa | $0.86/GPU/hr | Available | ||
RunPod | NVIDIA L40S 48GB VRAM | 48GB | 16 vCPU 94GB RAM | 🌍global | $0.86/GPU/hr | |||
Massed Compute | 2×NVIDIA L40 48GB VRAM | 48GB | 26 vCPU 144GB RAM 1250GB Storage | Iowa | $0.86/GPU/hr $1.72/hr total (2×) | Available |
L40S
| Provider | GPU Model | VRAM | Host Specs | Region | Price | Status | Action | |
|---|---|---|---|---|---|---|---|---|
TensorDock | NVIDIA L40S 48GB VRAM | 48GB | 0 vCPU 0GB RAM | Wolverhampton | $0.55/GPU/hr | Available | ||
RunPod | NVIDIA L40S 48GB VRAM | 48GB | 16 vCPU 94GB RAM | 🌍global | $0.86/GPU/hr | |||
Massed Compute | NVIDIA L40S 48GB VRAM | 48GB | 12 vCPU 72GB RAM 625GB Storage | Iowa | $0.88/GPU/hr | Available | ||
Massed Compute | 2×NVIDIA L40S 48GB VRAM | 48GB | 24 vCPU 144GB RAM 1250GB Storage | Iowa | $0.88/GPU/hr $1.76/hr total (2×) | Available | ||
Massed Compute | NVIDIA L40S 48GB VRAM | 48GB | 12 vCPU 72GB RAM 625GB Storage | Iowa | $0.88/GPU/hr | Available |
When to Choose the L40
The L40 suits cost-sensitive deployments where FP32 workloads dominate. Its 90.5 TFLOPS FP32 matches the L40S 91 TFLOPS closely, ideal for scientific computing or legacy graphics rendering. Lower 300W TDP reduces operational costs in power-constrained clouds, and pricing from $0.67 per hour appeals for long-running jobs across 11 offers.
Choose the L40 for balanced performance without FP16 premiums, such as in environments prioritizing efficiency over peak AI throughput.
When to Choose the L40S
The L40S excels in modern AI tasks leveraging FP16 and FP8 precision. Its 362 TFLOPS FP16 delivers four times the L40's rate, speeding LLM training and inference, while 724 TFLOPS FP8 optimizes quantized serving. Despite higher average $1.10 per hour pricing across 18 offers, minimum $0.40 per hour options make it viable.
Select the L40S for high-throughput workloads like fine-tuning large models, where 350W TDP sustains superior compute.
Use Cases
L40S 362 TFLOPS FP16 quadruples L40 90.5 TFLOPS, reducing training time for large models. Shared 48 GB VRAM supports massive batches.
L40S FP8 at 724 TFLOPS enables quantized serving with low latency. FP16 advantage accelerates real-time queries over L40.
Higher 362 TFLOPS FP16 on L40S speeds parameter updates versus L40 90.5 TFLOPS. Identical bandwidth aids large datasets.
L40S FP16 performance generates images 4x faster than L40. 48 GB VRAM handles high-resolution diffusion models.
L40 FP32 at 90.5 TFLOPS nearly matches L40S 91 TFLOPS for simulations. Lower 300W TDP cuts costs in FP32-heavy tasks.
Frequently Asked Questions
What is the main performance difference between L40 and L40S?▾
The L40S offers 362 TFLOPS FP16 versus L40 90.5 TFLOPS, a 4x boost for AI tasks. It adds 724 TFLOPS FP8 absent on L40, with FP32 at 91 TFLOPS versus 90.5 TFLOPS.
Which has better pricing on gpuperhour.com?▾
L40S starts at $0.40 per hour across 18 offers, cheaper minimum than L40 $0.67 per hour over 11 offers. L40 averages lower at $0.86 versus L40S $1.10.
Do L40 and L40S have the same VRAM?▾
Both provide 48 GB VRAM with 864 GB/s bandwidth. L40 uses GDDR6, L40S GDDR6X for potential efficiency gains.
What is the TDP difference?▾
L40 consumes 300W, L40S 350W. Higher TDP on L40S supports sustained peak performance in demanding workloads.
Are they the same architecture?▾
Both use Ada Lovelace from 2023 in PCIe form factors. L40S specifies PCIe 4.0 interconnect.
Is L40S better for inference?▾
Yes, L40S FP8 724 TFLOPS excels in quantized LLM inference. FP16 362 TFLOPS also outperforms L40 90.5 TFLOPS.
Which is cheaper to rent, the L40 or the L40S?▾
Cloud rental prices for both the L40 and L40S vary by provider, configuration, and availability. This page shows live pricing from 25+ providers updated every 60 seconds. Scroll to the Live Cloud Pricing section to compare current rates.
How much VRAM does the L40 have compared to the L40S?▾
The L40 has 48 GB of GDDR6 memory. The L40S has 48 GB of GDDR6X memory.
Can I find L40 and L40S GPUs available to rent right now?▾
Yes. This page shows real-time availability across 25+ cloud GPU providers. The Live Cloud Pricing section displays only in-stock offers with current pricing.
What is the main difference between the L40 and the L40S?▾
The L40 uses the Ada Lovelace architecture (2023) while the L40S uses Ada Lovelace (2023). The L40S delivers 4.0x the FP16 throughput and 1.0x the memory bandwidth of the L40.