Specifications Compared
| Spec | GH200 | L40 |
|---|---|---|
| TDP | 900W | 300W |
| VRAM | 96 GB | 48 GB |
| CUDA Cores | 16,896 | 18,176 |
| FP8 (dense) | 1,979 TFLOPS | 362 TFLOPS |
| Memory Type | HBM3 | GDDR6 |
| Architecture | Hopper | Ada Lovelace |
| FP16 (dense) | 989 TFLOPS | 181.05 TFLOPS |
| Form Factors | SXM | PCIe |
| INT8 (dense) | 1,979 TOPS | 362 TOPS |
| Interconnect | NVLink-C2C, PCIe 5.0 | PCIe 4.0 |
| Tensor Cores | 528 | 568 |
| FP32 Performance | 67 TFLOPS | 90.5 TFLOPS |
| FP64 Performance | 34 TFLOPS | Not published |
| Memory Bandwidth | 4,000 GB/s | 864 GB/s |
| FP8 (with sparsity) | 3,958 TFLOPS | 724 TFLOPS |
| FP16 (with sparsity) | 1,979 TFLOPS | 362.1 TFLOPS |
| INT8 (with sparsity) | 3,958 TOPS | 724 TOPS |
Performance Analysis
The FP16 dense performance of 989 TFLOPS on the GH200 exceeds the 181.05 TFLOPS on the L40 by a factor of roughly five. This gap implies faster training iterations for large language models on the GH200 when using dense FP16 calculations. The FP32 performance of 90.5 TFLOPS on the L40 surpasses the 67 TFLOPS on the GH200 and this advantage may benefit certain scientific simulations that rely on single precision arithmetic. Memory bandwidth of 4000 GB/s on the GH200 compared to 864 GB/s on the L40 allows larger batch sizes during inference and training. The higher bandwidth reduces data movement bottlenecks in memory intensive operations. Sparse FP16 performance reaches 1979 TFLOPS on the GH200 versus 362.1 TFLOPS on the L40 and this ratio holds for workloads that exploit 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.
GH200
GH200 is not offered on-demand by any provider we track right now. See the GH200 rental page for last-seen listed prices and a price alert.
L40
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| ThunderCompute | USA | 1 | $0.79 | — | Deploy |
| Massed Compute | us-central-2 | 1 | $0.86 | — | Deploy |
| QuantaCloud | us-midwest-2 | 4 | $0.94 | $3.75 | Deploy |
| Hyperstack | CANADA-1 | 1 | $1.00 | — | Deploy |
4 providers in stock, 17 offers (cheapest per provider shown). All L40 offers, price history and alerts
Notify me when GH200 drops below a price
One email when the cheapest in-stock on-demand price per GPU-hour falls below your threshold.
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 GH200
The GH200 serves as the better choice for large scale model training. Its 96 GB HBM3 memory and 4000 GB/s bandwidth support extensive batch sizes with dense FP16 performance of 989 TFLOPS. The 900 W TDP accommodates sustained high throughput operations in data center environments that prioritize capacity over power draw.
When to Choose the L40
The L40 serves as the better choice for power constrained inference deployments. Its 300 W TDP and FP32 performance of 90.5 TFLOPS suit workloads that benefit from single precision arithmetic without requiring the 96 GB memory capacity of the GH200. The PCIe form factor further simplifies integration in standard server configurations.
Use Cases
The GH200 provides 989 TFLOPS dense FP16 performance and 96 GB memory that enable efficient handling of large models.
The GH200 supplies 4000 GB/s bandwidth and 1979 TFLOPS dense FP8 performance that support high throughput serving.
The GH200 offers 96 GB HBM3 memory and 989 TFLOPS dense FP16 performance that accommodate parameter updates on sizable datasets.
The L40 delivers 90.5 TFLOPS FP32 performance and 300 W TDP that align with graphics oriented tasks at lower power.
The L40 achieves 90.5 TFLOPS FP32 performance that exceeds the 67 TFLOPS of the GH200 for single precision workloads.
Frequently Asked Questions
What is the FP16 dense performance of each GPU?▾
The GH200 reaches 989 TFLOPS dense FP16 performance. The L40 reaches 181.05 TFLOPS dense FP16 performance. The ratio favors the GH200 for dense tensor operations.
Which GPU has higher memory bandwidth?▾
The GH200 has 4000 GB/s memory bandwidth. The L40 has 864 GB/s memory bandwidth. Higher bandwidth on the GH200 supports larger batch sizes in memory bound tasks.
How do the TDP values compare between the GH200 and L40?▾
The GH200 has a 900 W TDP. The L40 has a 300 W TDP. The lower TDP on the L40 reduces power requirements in dense deployments.
What FP32 performance does each GPU deliver?▾
The GH200 delivers 67 TFLOPS FP32 performance. The L40 delivers 90.5 TFLOPS FP32 performance. The L40 holds an advantage in single precision workloads.
Which interconnect does the GH200 support?▾
The GH200 supports NVLink-C2C and PCIe 5.0 interconnects. The L40 supports PCIe 4.0 interconnect. The GH200 interconnect options enable faster multi GPU communication.
Which is cheaper to rent, the GH200 or the L40?▾
Cloud rental prices for both the GH200 and L40 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 GH200 have compared to the L40?▾
The GH200 has 96 GB of HBM3 memory. The L40 has 48 GB of GDDR6 memory.
Can I find GH200 and L40 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 GH200 and the L40?▾
The GH200 uses the Hopper architecture (2023) while the L40 uses Ada Lovelace (2023). The GH200 delivers 5.5x the dense FP16 throughput (989 vs 181.05 TFLOPS, both without sparsity) and 4.6x the memory bandwidth of the L40.
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 GH200 and the L40. 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 GH200 Grace HopperEvery current offer by provider, daily price history and a price alert.
- Rent L40Every 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.