Specifications Compared
| Spec | H100 | L40S |
|---|---|---|
| TDP | 700W | 350W |
| VRAM | 80-94 GB | 48 GB |
| CUDA Cores | 16,896 | 18,176 |
| Memory Type | HBM3 | GDDR6X |
| Architecture | Hopper | Ada Lovelace |
| Form Factors | SXM5, PCIe, NVL | PCIe |
| Interconnect | NVLink, PCIe 5.0, InfiniBand | PCIe 4.0 |
| Tensor Cores | 528 | 568 |
| FP8 Performance | 3,958 TFLOPS | 724 TFLOPS |
| FP16 Performance | 1,979 TFLOPS | 362 TFLOPS |
| FP32 Performance | 67 TFLOPS | 91 TFLOPS |
| FP64 Performance | 34 TFLOPS | 1.4 TFLOPS |
| INT8 Performance | 3,958 TOPS | 724 TOPS |
| Memory Bandwidth | 3,350 GB/s | 864 GB/s |
Performance Analysis
The H100's 1979 TFLOPS FP16 vastly outpaces L40S's 362 TFLOPS, accelerating deep learning training where half-precision dominates; this enables faster iterations on large datasets. Conversely, L40S edges FP32 at 91 TFLOPS over H100's 67 TFLOPS, suiting traditional simulations or rendering less optimized for low-precision. FP8 performance seals H100's lead at 3958 TFLOPS versus 724 TFLOPS, ideal for efficient large language model inference.
Memory specs dictate real-world scalability: H100's 80 to 94 GB HBM3 and 3350 GB/s bandwidth support enormous batch sizes in training, minimizing data movement bottlenecks, while L40S's 48 GB GDDR6X and 864 GB/s constrain it to smaller models or inference. H100's NVLink and PCIe 5.0 interconnects enhance multi-GPU scaling over L40S's PCIe 4.0.
Live Cloud Pricing
Real-time prices from 25+ providers. Updated every 60 seconds.
H100 NVL
| Provider | GPU Model | VRAM | Host Specs | Region | Price | Status | Action | |
|---|---|---|---|---|---|---|---|---|
 Hyperstack | 4×NVIDIA H100 PCIe 80GB VRAM | 80GB | 124 vCPU 720GB RAM 3300GB Storage | Canada | $1.90/GPU/hr $7.60/hr total (4×) | Available | ||
Hyperstack | 2×NVIDIA H100 PCIe 80GB VRAM | 80GB | 60 vCPU 360GB RAM 1600GB Storage | Canada | $1.90/GPU/hr $3.80/hr total (2×) | Available | ||
Hyperstack | 8×NVIDIA H100 PCIe 80GB VRAM | 80GB | 252 vCPU 1440GB RAM 6600GB Storage | Canada | $1.90/GPU/hr $15.20/hr total (8×) | Available | ||
Hyperstack | NVIDIA H100 PCIe 80GB VRAM | 80GB | 28 vCPU 180GB RAM 850GB Storage | Canada | $1.90/GPU/hr | Available | ||
Hyperstack | 8×NVIDIA H100 PCIe 80GB VRAM | 80GB | 252 vCPU 1440GB RAM 6600GB Storage | Canada | $1.95/GPU/hr $15.60/hr total (8×) | 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 H100 NVL
Opt for the H100 in LLM training or fine-tuning of models exceeding 48 GB VRAM, leveraging its 80 to 94 GB HBM3 and 3350 GB/s bandwidth for large batches. Its 1979 TFLOPS FP16 and 3958 TFLOPS FP8 deliver unmatched throughput for data center-scale AI workloads, despite 700W TDP and higher $1.40 to $2.89 per hour pricing.
When to Choose the L40S
Select the L40S for cost-sensitive inference, Stable Diffusion, or workloads fitting within 48 GB GDDR6X, where $0.40 per hour average $1.14 pricing and 350W TDP reduce expenses. Its 91 TFLOPS FP32 aids graphics or scientific tasks, and PCIe 4.0 suffices for single-node deployments.
Use Cases
H100's 1979 TFLOPS FP16 and 80 to 94 GB HBM3 enable training of massive models with large batches. L40S's 362 TFLOPS FP16 and 48 GB VRAM fall short for such scales.
H100's 3958 TFLOPS FP8 and 3350 GB/s bandwidth support high-throughput serving of large models. L40S manages smaller deployments at lower cost but lacks capacity.
H100 accommodates full model fine-tuning via 80 to 94 GB VRAM, unlike L40S's 48 GB limit. Its superior FP16 performance speeds convergence.
L40S's 91 TFLOPS FP32 and lower $0.40 per hour pricing suit image generation efficiently. H100 overkill for typical diffusion model sizes.
L40S's FP32 lead at 91 TFLOPS fits simulations; H100 excels in mixed-precision HPC with 3350 GB/s bandwidth. Choice depends on precision needs and budget.
Frequently Asked Questions
Which GPU has more VRAM?▾
The H100 offers 80 to 94 GB HBM3 VRAM, exceeding L40S's 48 GB GDDR6X. This advantage supports larger models in training and inference.
What are the cloud pricing differences?▾
H100 NVL starts at $1.40 per hour averaging $2.89 across 9 offers; L40S begins at $0.40 per hour averaging $1.14 from 22 offers. L40S provides better value for lighter workloads.
How do FP16 performances compare?▾
H100 achieves 1979 TFLOPS FP16, over five times L40S's 362 TFLOPS. This gap accelerates AI training significantly on H100.
What is the power consumption?▾
H100 draws 700W TDP; L40S uses 350W. Lower TDP makes L40S preferable for dense or power-constrained environments.
Which has higher memory bandwidth?▾
H100 delivers 3350 GB/s, nearly four times L40S's 864 GB/s. Higher bandwidth on H100 boosts large batch processing.
Do they support NVLink?▾
H100 includes NVLink for multi-GPU scaling; L40S relies on PCIe 4.0 only. NVLink enhances H100 in clustered training.
Which is cheaper to rent, the H100 or the L40S?▾
Cloud rental prices for both the H100 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 H100 have compared to the L40S?▾
The H100 has 80 to 94 GB of HBM3 memory. The L40S has 48 GB of GDDR6X memory.
Can I find H100 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 H100 and the L40S?▾
The H100 uses the Hopper architecture (2022) while the L40S uses Ada Lovelace (2023). The H100 delivers 5.5x the FP16 throughput and 3.9x the memory bandwidth of the L40S.