Specifications Compared
| Spec | H200 NVL | L40S |
|---|---|---|
| TDP | 700W | 350W |
| VRAM | 141 GB | 48 GB |
| CUDA Cores | 16,896 | 18,176 |
| FP8 (dense) | 1,979 TFLOPS | 733 TFLOPS |
| Memory Type | HBM3e | GDDR6 |
| Architecture | Hopper | Ada Lovelace |
| FP16 (dense) | 989 TFLOPS | 362.05 TFLOPS |
| Form Factors | NVL (PCIe card, NVLink bridged) | PCIe |
| INT8 (dense) | 1,979 TOPS | 733 TOPS |
| Interconnect | NVLink bridge, PCIe 5.0 | PCIe 4.0 |
| Tensor Cores | 528 | 568 |
| FP32 Performance | 67 TFLOPS | 91.6 TFLOPS |
| FP64 Performance | 34 TFLOPS | 1.4 TFLOPS |
| Memory Bandwidth | 4,800 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 H200 NVL delivers FP16 dense throughput of 989 TFLOPS compared to the L40S FP16 dense throughput of 362.05 TFLOPS while both share family memory bandwidth figures of 4800 GB/s and 864 GB/s respectively. FP32 performance stands at 67 TFLOPS for the H200 NVL family and 91.6 TFLOPS for the L40S family. The FP16 dense to FP32 ratio on the H200 NVL supports larger batch sizes during training and inference than the corresponding ratio on the L40S because the 4800 GB/s bandwidth sustains higher data movement than the 864 GB/s bandwidth. FP16 with sparsity reaches 1979 TFLOPS on the H200 NVL family versus 733 TFLOPS on the L40S family and the same pattern holds for FP8 dense figures of 1979 TFLOPS versus 733 TFLOPS.
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.
H200 NVL
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| RunPod | global | 1 | $5.29 | — | Deploy |
| Vast.ai | Czechia, CZ | 1 | $5.47 | — | Deploy |
| VERDA | FIN-HEL | 2 | $5.60 | $11.20 | Deploy |
| Lyceum | Europe | 4 | $6.19 | $24.76 | Deploy |
| LeaderGPU | The Netherlands | 1 | $6.22 | — | Deploy |
5 providers in stock, 9 offers (cheapest per provider shown). All H200 NVL offers, price history and alerts
L40S
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| Massed Compute | us-central-2 | 4 | $0.97 | $3.88 | Deploy |
| QuantaCloud | us-midwest-2 | 1 | $1.09 | — | Deploy |
| RunPod | global | 1 | $1.09 | — | Deploy |
| Ori | lille-4 | 1 | $1.55 | — | Deploy |
| VERDA | FIN-HEL | 1 | $1.75 | — | Deploy |
6 providers in stock, 17 offers (cheapest per provider shown). All L40S offers, price history and alerts
Notify me when H200 NVL drops below a price
One email when the cheapest in-stock on-demand price per GPU-hour falls below your threshold. Cheapest right now: $5.29/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 H200 NVL
The H200 NVL variant suits workloads that require the 141 GB HBM3e capacity and NVLink bridge for multi GPU scaling. Its 700 W TDP and 4800 GB/s bandwidth support training of models that exceed the 48 GB limit of the L40S variant.
When to Choose the L40S
The L40S variant fits inference or fine tuning tasks where the 48 GB GDDR6 capacity meets requirements and the 350 W TDP reduces power draw compared to 700 W. Its PCIe 4.0 interconnect suffices for single GPU deployments without NVLink needs.
Use Cases
The 141 GB HBM3e capacity and 989 TFLOPS FP16 dense throughput of the H200 NVL variant enable training of larger models than the 48 GB and 362.05 TFLOPS of the L40S variant.
The H200 NVL variant 4800 GB/s bandwidth sustains higher throughput for large batch inference than the 864 GB/s bandwidth of the L40S variant.
Both variants handle fine tuning when model size fits within 141 GB or 48 GB memory limits.
The L40S variant 350 W TDP and 48 GB capacity provide adequate performance for diffusion workloads at lower power than the 700 W TDP of the H200 NVL variant.
The H200 NVL variant 67 TFLOPS FP32 and NVLink interconnect support distributed scientific simulations beyond the PCIe 4.0 limits of the L40S variant.
Frequently Asked Questions
How does FP16 dense performance compare between these GPUs?▾
The H200 NVL family reports 989 TFLOPS FP16 dense while the L40S family reports 362.05 TFLOPS FP16 dense. The figures remain shared across variants within each family.
What interconnect does the H200 NVL variant use?▾
The H200 NVL variant uses an NVLink bridge together with PCIe 5.0. The L40S variant uses only PCIe 4.0.
What is the TDP of each GPU family?▾
The H200 family TDP is 700 W and the L40S family TDP is 350 W. These values apply to all variants including the NVL and PCIe models.
Which is cheaper to rent, the H200 NVL or the L40S?▾
Cloud rental prices for both the H200 NVL 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 H200 NVL have compared to the L40S?▾
The H200 NVL has 141 GB of HBM3e memory. The L40S has 48 GB of GDDR6 memory.
Can I find H200 NVL 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 H200 NVL and the L40S?▾
The H200 NVL uses the Hopper architecture (2024) while the L40S uses Ada Lovelace (2023). The H200 NVL delivers 2.7x the dense FP16 throughput (989 vs 362.05 TFLOPS, both without sparsity) and 5.6x 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 H200 NVL 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 H200Every 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.