Specifications Compared
| Spec | H100 | L40S |
|---|---|---|
| TDP | 700W | 350W |
| VRAM | 80-94 GB | 48 GB |
| CUDA Cores | 16,896 | 18,176 |
| FP8 (dense) | 1,979 TFLOPS | 733 TFLOPS |
| Memory Type | HBM3 | GDDR6 |
| Architecture | Hopper | Ada Lovelace |
| FP16 (dense) | 989 TFLOPS | 362.05 TFLOPS |
| Form Factors | SXM5, PCIe, NVL | PCIe |
| INT8 (dense) | 1,979 TOPS | 733 TOPS |
| Interconnect | NVLink, PCIe 5.0, InfiniBand | PCIe 4.0 |
| Tensor Cores | 528 | 568 |
| FP32 Performance | 67 TFLOPS | 91.6 TFLOPS |
| FP64 Performance | 34 TFLOPS | 1.4 TFLOPS |
| Memory Bandwidth | 3,350 GB/s | 864 GB/s |
| FP8 (with sparsity) | 3,958 TFLOPS | 1,466 TFLOPS |
| FP16 (with sparsity) | 1,979 TFLOPS | 733 TFLOPS |
| INT8 (with sparsity) | 3,958 TOPS | 1,466 TOPS |
Performance Analysis
The FP16 dense rating of 989 TFLOPS on the H100 exceeds the 362.05 TFLOPS dense rating on the L40S by a factor derived directly from those two numbers. This gap affects training workloads where higher dense FP16 throughput supports larger batch sizes when paired with the 3350 GB/s bandwidth figure. The L40S FP32 rating of 91.6 TFLOPS surpasses the H100 FP32 rating of 67 TFLOPS and therefore favors certain inference paths that remain in single precision. Memory bandwidth of 3350 GB/s on the H100 versus 864 GB/s on the L40S determines feasible batch sizes for models that exceed 48 GB capacity. Sparse FP16 figures follow the same ratio pattern with the H100 at 1979 TFLOPS with sparsity against the L40S at 733 TFLOPS with sparsity.
Current On-Demand Offers
Cheapest secure on-demand offer per provider that is reported in stock right now, per GPU per hour; multi-GPU instances show the whole-instance price alongside. Deploy opens the DeployGPU console for providers on the platform, otherwise the provider's own site. Stock is rechecked every minute.
H100
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| QuantaCloud | us-midwest-2 | 2 | $2.59 | $5.18 | Deploy |
| Massed Compute | us-central-3 | 2 | $2.73 | $5.46 | Deploy |
| Ori | dallas-2 | 2 | $2.90 | $5.80 | Deploy |
| RunPod | global | 1 | $3.19 | — | Deploy |
| DigitalOcean | Toronto 1 | 1 | $4.41 | — | Deploy |
7 providers in stock, 18 offers (cheapest per provider shown). All H100 offers, price history and alerts
L40S
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| Massed Compute | us-central-2 | 1 | $0.97 | — | Deploy |
| RunPod | global | 1 | $1.09 | — | Deploy |
| QuantaCloud | us-midwest-1 | 1 | $1.09 | — | Deploy |
| Lyceum | Europe | 2 | $1.99 | $3.98 | Deploy |
4 providers in stock, 5 offers (cheapest per provider shown). All L40S offers, price history and alerts
Notify me when H100 drops below a price
One email when the cheapest in-stock on-demand price per GPU-hour falls below your threshold. Cheapest right now: $2.59/GPU-hr.
QuantaCloud
Comparing H-series providers? We broker across all of them.
Hopper stock changes by the hour and prices differ by provider. If you need 16+ GPUs reserved or a cluster in the next 90 days, we quote H-series or B300 inventory at partner rates: one quote, 24h turnaround.
When to Choose the H100
The H100 suits workloads that require the 80 to 94 GB HBM3 capacity together with 3350 GB/s bandwidth for training runs that exceed 48 GB model footprints. NVLink interconnect availability further supports multi GPU scaling on the H100 where the L40S remains limited to PCIe 4.0.
When to Choose the L40S
The L40S fits deployments that operate within 48 GB VRAM limits and benefit from its 91.6 TFLOPS FP32 rating at 350W TDP. PCIe form factor simplicity and lower power draw align with inference tasks that stay below the H100 memory threshold.
Use Cases
The H100 provides 989 TFLOPS dense FP16 and 80 to 94 GB memory that exceed the L40S values for large model training.
The H100 3350 GB/s bandwidth supports larger batch sizes than the L40S 864 GB/s bandwidth during inference.
The H100 1979 TFLOPS with sparsity in FP16 enables faster fine tuning iterations compared to the L40S 733 TFLOPS with sparsity.
The L40S 91.6 TFLOPS FP32 rating exceeds the H100 67 TFLOPS FP32 rating for diffusion workloads that remain in single precision.
The L40S 350W TDP and 91.6 TFLOPS FP32 rating suit sustained scientific workloads that value lower power over the H100 700W TDP.
Frequently Asked Questions
What is the memory difference between H100 and L40S?▾
The H100 offers 80 to 94 GB HBM3 while the L40S offers 48 GB GDDR6. This difference directly limits maximum model size on the L40S.
How does FP16 performance compare on H100 versus L40S?▾
The H100 lists 989 TFLOPS dense FP16 and 1979 TFLOPS with sparsity. The L40S lists 362.05 TFLOPS dense FP16 and 733 TFLOPS with sparsity.
Which GPU has higher FP32 throughput?▾
The L40S reaches 91.6 TFLOPS in FP32 while the H100 reaches 67 TFLOPS in FP32. Workloads that stay in FP32 therefore favor the L40S.
What interconnect options exist on each card?▾
The H100 supports NVLink and InfiniBand while the L40S supports only PCIe 4.0. Multi GPU scaling therefore differs between the two.
How do the TDP values affect deployment?▾
The H100 TDP is 700W and the L40S TDP is 350W. Power and cooling infrastructure must accommodate the higher figure on the H100.
Which architecture generation is newer?▾
The L40S uses Ada Lovelace architecture dated 2023 while the H100 uses Hopper architecture dated 2022.
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 GDDR6 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 2.7x the dense FP16 throughput (989 vs 362.05 TFLOPS, both without sparsity) and 3.9x the memory bandwidth of the L40S.
Rent these GPUs
Each GPU page lists every current on-demand offer by provider, per GPU-hour, updated every minute.
Related comparisons
How this page is made
- Specifications come from the NVIDIA, AMD and Intel datasheets for the H100 and the L40S. Dense and sparse throughput are listed separately.
- Prices and availability are live: every offer shown is an in-stock, on-demand listing from a provider we track, rechecked every minute.
- The written comparison (overview, performance notes, when to choose each card, verdict, use cases and FAQ) was drafted with an AI model from the specification table above and passed an automated check that rejects any figure not in that table. It was last generated on . It contains no prices; those are always read live.
- Read how we collect the data, or report an error on this page.
Next steps
- Rent H100Every current offer by provider, daily price history and a price alert.
- Rent L40SEvery current offer by provider, daily price history and a price alert.
- GPU price indexHow on-demand prices for the major GPUs have moved, updated daily.
- All GPUsEvery GPU we track, with current lows and provider counts.