Specifications Compared
| Spec | A40 | H200 |
|---|---|---|
| TDP | 300W | 700W |
| VRAM | 48 GB | 141 GB |
| CUDA Cores | 10,752 | 16,896 |
| FP8 (dense) | Not published | 1,979 TFLOPS |
| Memory Type | GDDR6 | HBM3e |
| Architecture | Ampere | Hopper |
| FP16 (dense) | 149.7 TFLOPS | 989 TFLOPS |
| Form Factors | PCIe | SXM, NVL |
| INT8 (dense) | 299.3 TOPS | 1,979 TOPS |
| Interconnect | NVLink, PCIe 4.0 | NVLink, PCIe 5.0, InfiniBand |
| Tensor Cores | 336 | 528 |
| FP32 Performance | 37.4 TFLOPS | 67 TFLOPS |
| FP64 Performance | Not published | 34 TFLOPS |
| Memory Bandwidth | 696 GB/s | 4,800 GB/s |
| FP8 (with sparsity) | Not published | 3,958 TFLOPS |
| FP16 (with sparsity) | 299.4 TFLOPS | 1,979 TFLOPS |
| INT8 (with sparsity) | 598.6 TOPS | 3,958 TOPS |
Performance Analysis
The FP16 dense rating reaches 149.7 TFLOPS on the A40 and 989 TFLOPS on the H200. The FP16 figure with sparsity reaches 299.4 TFLOPS on the A40 and 1979 TFLOPS on the H200. These ratios indicate that the H200 can process larger batches during training or inference when memory bandwidth of 4800 GB/s is available compared with 696 GB/s on the A40. FP32 throughput of 67 TFLOPS on the H200 versus 37.4 TFLOPS on the A40 further widens the gap for precision sensitive workloads. The A40 INT8 dense rating of 299.3 TOPS remains far below the H200 INT8 dense rating of 1979 TOPS.
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.
A40
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| LeaderGPU | The Netherlands | 8 | $0.52 | $4.13 | Deploy |
| RunPod | global | 1 | $0.59 | — | Deploy |
2 providers in stock, 2 offers. All A40 offers, price history and alerts
H200
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| Vast.ai | Czechia, CZ | 1 | $4.19 | — | Deploy |
| RunPod | global | 1 | $5.29 | — | Deploy |
| VERDA | FIN-HEL | 2 | $5.60 | $11.20 | Deploy |
| Lyceum | Europe | 1 | $6.19 | — | Deploy |
| LeaderGPU | The Netherlands | 1 | $6.22 | — | Deploy |
5 providers in stock, 11 offers (cheapest per provider shown). All H200 offers, price history and alerts
Notify me when A40 drops below a price
One email when the cheapest in-stock on-demand price per GPU-hour falls below your threshold. Cheapest right now: $0.52/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 A40
The A40 fits workloads that remain within 48 GB VRAM and that operate under a 300 W power envelope. PCIe form factor deployment and lower TDP make the A40 suitable for dense rack configurations where sustained FP32 throughput of 37.4 TFLOPS meets requirements without exceeding thermal limits.
When to Choose the H200
The H200 suits large scale training and inference that require 141 GB HBM3e and 4800 GB/s bandwidth. FP16 dense performance of 989 TFLOPS together with FP8 dense performance of 1979 TFLOPS enable higher throughput on models that exceed the capacity of the A40.
Use Cases
The H200 supplies 989 TFLOPS FP16 dense and 141 GB memory which exceed the 149.7 TFLOPS and 48 GB of the A40.
The H200 delivers 1979 TFLOPS FP16 with sparsity and 4800 GB/s bandwidth that allow larger batch sizes than the A40 can sustain.
The H200 FP32 rating of 67 TFLOPS combined with 141 GB capacity supports fine tuning runs that fit poorly in 48 GB.
The A40 48 GB GDDR6 and 300 W TDP suffice for typical image generation workloads at lower power draw.
Both GPUs provide FP32 throughput yet the A40 meets many simulation needs at 37.4 TFLOPS within a 300 W limit.
Frequently Asked Questions
What is the memory difference between A40 and H200?▾
The A40 contains 48 GB GDDR6 while the H200 contains 141 GB HBM3e. Memory bandwidth stands at 696 GB/s for the A40 and 4800 GB/s for the H200.
How does FP16 performance compare?▾
FP16 dense performance measures 149.7 TFLOPS on the A40 and 989 TFLOPS on the H200. FP16 performance with sparsity measures 299.4 TFLOPS on the A40 and 1979 TFLOPS on the H200.
Which GPU uses less power?▾
The A40 TDP is listed at 300 W while the H200 TDP is listed at 700 W. The A40 therefore operates at lower power for equivalent rack space.
Does the A40 support NVLink?▾
The A40 lists NVLink as an interconnect option together with PCIe 4.0. The H200 also lists NVLink together with PCIe 5.0 and InfiniBand.
What architecture does each GPU use?▾
The A40 uses Ampere architecture released in 2020. The H200 uses Hopper architecture released in 2024.
Which is cheaper to rent, the A40 or the H200?▾
Cloud rental prices for both the A40 and H200 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 A40 have compared to the H200?▾
The A40 has 48 GB of GDDR6 memory. The H200 has 141 GB of HBM3e memory.
Can I find A40 and H200 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 A40 and the H200?▾
The A40 uses the Ampere architecture (2020) while the H200 uses Hopper (2024). The H200 delivers 6.6x the dense FP16 throughput (989 vs 149.7 TFLOPS, both without sparsity) and 6.9x the memory bandwidth of the A40.
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 A40 and the H200. 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 A40Every current offer by provider, daily price history and a price alert.
- Rent H200Every 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.