Specifications Compared
| Spec | H100 NVL | RTX A6000 |
|---|---|---|
| TDP | 700W | 300W |
| VRAM | 94 GB | 48 GB |
| CUDA Cores | 16,896 | 10,752 |
| FP8 (dense) | 1,979 TFLOPS | Not published |
| Memory Type | HBM3 | GDDR6 |
| Architecture | Hopper | Ampere |
| FP16 (dense) | 989 TFLOPS | 154.8 TFLOPS |
| Form Factors | NVL (PCIe card, NVLink bridged) | PCIe |
| INT8 (dense) | 1,979 TOPS | 309.7 TOPS |
| Interconnect | NVLink bridge, PCIe 5.0 | NVLink, PCIe 4.0 |
| Tensor Cores | 528 | 336 |
| FP32 Performance | 67 TFLOPS | 38.7 TFLOPS |
| FP64 Performance | 34 TFLOPS | Not published |
| Memory Bandwidth | 3,350 GB/s | 768 GB/s |
| FP8 (with sparsity) | 3,958 TFLOPS | Not published |
| FP16 (with sparsity) | 1,979 TFLOPS | 309.6 TFLOPS |
| INT8 (with sparsity) | 3,958 TOPS | 619.4 TOPS |
Performance Analysis
The H100 NVL variant delivers dense FP16 throughput of 989 TFLOPS compared with the dense FP16 throughput of 154.8 TFLOPS on the RTX A6000 variant. This sixfold difference in dense FP16 performance means the H100 NVL variant completes matrix operations in training workloads at a substantially higher rate than the RTX A6000 variant. When sparsity is enabled the H100 NVL variant reaches 1979 TFLOPS in FP16 while the RTX A6000 variant reaches 309.6 TFLOPS in FP16. The memory bandwidth shared by the H100 NVL family at 3350 GB/s versus 768 GB/s shared by the RTX A6000 family allows larger batch sizes during inference because data movement between memory and compute units occurs more than four times faster on the H100 NVL variant. FP32 throughput shared by the H100 NVL family stands at 67 TFLOPS against 38.7 TFLOPS shared by the RTX A6000 family.
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 NVL
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| Vast.ai | Czechia, CZ | 1 | $2.59 | — | Deploy |
| QuantaCloud | us-midwest-2 | 1 | $2.59 | — | Deploy |
| Massed Compute | us-central-3 | 1 | $2.73 | — | Deploy |
| Ori | dallas-2 | 2 | $2.90 | $5.80 | Deploy |
| Scaleway | Warsaw, Poland (WAW2) | 1 | $3.23 | — | Deploy |
11 providers in stock, 21 offers (cheapest per provider shown). All H100 NVL offers, price history and alerts
RTX A6000
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| LeaderGPU | The Netherlands | 8 | $0.44 | $3.54 | Deploy |
| QuantaCloud | us-midwest-2 | 4 | $0.48 | $1.92 | Deploy |
| Hyperstack | CANADA-1 | 2 | $0.50 | $1.00 | Deploy |
| Massed Compute | us-central-2 | 1 | $0.55 | — | Deploy |
| RunPod | global | 1 | $0.59 | — | Deploy |
7 providers in stock, 54 offers (cheapest per provider shown). All RTX A6000 offers, price history and alerts
Notify me when H100 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: $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 NVL
The H100 NVL variant suits workloads that require the 94 GB memory capacity and 3350 GB/s bandwidth shared by its family. Large scale training jobs benefit from the dense FP16 figure of 989 TFLOPS and the 700 W TDP shared by the H100 NVL family when model size exceeds the 48 GB limit of the RTX A6000 variant.
When to Choose the RTX A6000
The RTX A6000 variant fits deployments constrained by the 300 W TDP shared by its family and by the 48 GB GDDR6 capacity. Inference tasks that fit within 48 GB memory and require only the dense FP16 throughput of 154.8 TFLOPS operate efficiently under the lower power envelope of the RTX A6000 variant.
Use Cases
The H100 NVL variant supplies 94 GB memory and dense FP16 throughput of 989 TFLOPS that exceed the 48 GB and 154.8 TFLOPS of the RTX A6000 variant.
The H100 NVL variant supports larger batch sizes through 3350 GB/s bandwidth and 1979 TFLOPS sparse FP16 compared with 768 GB/s and 309.6 TFLOPS on the RTX A6000 variant.
The H100 NVL variant provides 94 GB capacity and 989 TFLOPS dense FP16 that accommodate fine-tuning runs beyond the 48 GB limit of the RTX A6000 variant.
The RTX A6000 variant meets requirements within its 48 GB GDDR6 memory at 300 W TDP shared by its family for diffusion workloads that do not need the higher figures of the H100 NVL variant.
Both variants deliver FP32 throughput of 67 TFLOPS on the H100 NVL family and 38.7 TFLOPS on the RTX A6000 family for compute tasks that fit their respective memory sizes.
Frequently Asked Questions
What is the memory difference between the H100 NVL variant and the RTX A6000 variant?▾
The H100 NVL variant contains 94 GB HBM3 while the RTX A6000 variant contains 48 GB GDDR6. Memory bandwidth shared by the H100 NVL family reaches 3350 GB/s compared with 768 GB/s shared by the RTX A6000 family.
How does FP16 performance compare between these two variants?▾
Dense FP16 throughput on the H100 NVL variant is 989 TFLOPS while the RTX A6000 variant provides 154.8 TFLOPS. With sparsity the H100 NVL variant reaches 1979 TFLOPS and the RTX A6000 variant reaches 309.6 TFLOPS.
What are the TDP values for the H100 NVL variant and the RTX A6000 variant?▾
The H100 NVL variant shares the 700 W TDP of its family and the RTX A6000 variant shares the 300 W TDP of its family. These figures apply across all variants within each family.
Can the RTX A6000 variant handle models that require more than 48 GB?▾
No. The RTX A6000 variant is limited to its 48 GB GDDR6 capacity while the H100 NVL variant provides 94 GB HBM3 for larger models. Bandwidth of 3350 GB/s shared by the H100 NVL family further supports those larger workloads.
Which is cheaper to rent, the H100 NVL or the RTX A6000?▾
Cloud rental prices for both the H100 NVL and RTX A6000 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 NVL have compared to the RTX A6000?▾
The H100 NVL has 94 GB of HBM3 memory. The RTX A6000 has 48 GB of GDDR6 memory.
Can I find H100 NVL and RTX A6000 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 NVL and the RTX A6000?▾
The H100 NVL uses the Hopper architecture (2022) while the RTX A6000 uses Ampere (2020). The H100 NVL delivers 6.4x the dense FP16 throughput (989 vs 154.8 TFLOPS, both without sparsity) and 4.4x the memory bandwidth of the RTX A6000.
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 NVL and the RTX A6000. 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 RTX A6000Every 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.