Specifications Compared
| Spec | A100 | L40 |
|---|---|---|
| TDP | 400W | 300W |
| VRAM | 40-80 GB | 48 GB |
| CUDA Cores | 6,912 | 18,176 |
| Memory Type | HBM2e | GDDR6 |
| Architecture | Ampere | Ada Lovelace |
| Form Factors | SXM4, PCIe | PCIe |
| Interconnect | NVLink, PCIe 4.0, InfiniBand | |
| Tensor Cores | 432 | 568 |
| FP16 Performance | 312 TFLOPS | 90.5 TFLOPS |
| FP32 Performance | 19.5 TFLOPS | 90.5 TFLOPS |
| FP64 Performance | 9.7 TFLOPS | |
| INT8 Performance | 624 TOPS | 724 TOPS |
| Memory Bandwidth | 2,039 GB/s | 864 GB/s |
Performance Analysis
The A100 SXM4 40GB dominates in FP16 performance at 312 TFLOPS, which accelerates matrix-heavy operations central to neural network training. This capability shortens training times for large language models compared to the L40's 90.5 TFLOPS in FP16. The A100's 2039 GB/s bandwidth further enables larger batch sizes, minimizing data transfer bottlenecks in memory-intensive workflows. In contrast, the L40's identical 90.5 TFLOPS across FP16 and FP32 supports balanced precision for inference tasks, where FP32 computations maintain accuracy without the A100's 19.5 TFLOPS limitation in that format. Lower bandwidth of 864 GB/s on the L40 may constrain batch sizes in high-throughput scenarios, though its Ada Lovelace tensor cores optimize sparse operations. Power efficiency favors the L40 at 300W TDP over 400W, reducing operational costs in prolonged inference deployments.
Live Cloud Pricing
Real-time prices from 25+ providers. Updated every 60 seconds.
A100 SXM4 40GB
| Provider | GPU Model | VRAM | Host Specs | Region | Price | Status | Action | |
|---|---|---|---|---|---|---|---|---|
 Vast.ai | NVIDIA A100 SXM4 80GB 80GB VRAM | 80GB | 256 vCPU 63GB RAM 2826GB Storage | Slovenia | $0.73/GPU/hr | Available | ||
Vast.ai | 2×NVIDIA A100 SXM4 80GB 80GB VRAM | 80GB | 256 vCPU 126GB RAM 794GB Storage | Slovenia | $0.73/GPU/hr $1.47/hr total (2×) | Available | ||
 LeaderGPU | 8×NVIDIA A100 PCIe 80GB 80GB VRAM | 80GB | 64 vCPU 384GB RAM 2000GB Storage | Netherlands | $0.90/GPU/hr $7.20/hr total (8×) | Available | ||
Vast.ai | NVIDIA A100 SXM4 80GB 80GB VRAM | 80GB | 64 vCPU 63GB RAM 646GB Storage | Czechia | $1.07/GPU/hr | Available | ||
 Denvr | 8×NVIDIA A100 SXM4 80GB 80GB VRAM | 80GB | 128 vCPU 1024GB RAM 15200GB Storage | Virginia | $1.15/GPU/hr $9.20/hr total (8×) |
L40
| Provider | GPU Model | VRAM | Host Specs | Region | Price | Status | Action | |
|---|---|---|---|---|---|---|---|---|
 TensorDock | NVIDIA L40S 48GB VRAM | 48GB | 0 vCPU 0GB RAM | Wolverhampton | $0.55/GPU/hr | Available | ||
 RunPod | NVIDIA L40 48GB VRAM | 48GB | 8 vCPU 94GB RAM | 🌍global | $0.82/GPU/hr | |||
RunPod | NVIDIA L40S 48GB VRAM | 48GB | 16 vCPU 94GB RAM | 🌍global | $0.86/GPU/hr | |||
 Massed Compute | NVIDIA L40 48GB VRAM | 48GB | 14 vCPU 72GB RAM 625GB Storage | Iowa | $0.86/GPU/hr | Available | ||
Massed Compute | 2×NVIDIA L40 48GB VRAM | 48GB | 26 vCPU 144GB RAM 1250GB Storage | Iowa | $0.86/GPU/hr $1.72/hr total (2×) | Available |
When to Choose the A100 SXM4 40GB
Select the A100 SXM4 40GB for deep learning training workloads demanding peak FP16 throughput. Its 312 TFLOPS and 2039 GB/s bandwidth excel in processing massive datasets with large batch sizes, such as LLM pre-training. High-speed HBM2e memory sustains performance in multi-GPU setups via NVLink interconnects.
When to Choose the L40
Opt for the L40 in inference-heavy or mixed-precision applications. Balanced 90.5 TFLOPS in both FP16 and FP32, paired with 48 GB GDDR6 VRAM, handles real-time serving efficiently at lower 300W TDP. Cost savings from $0.67 per hour pricing make it ideal for scalable cloud deployments.
Use Cases
A100's 312 TFLOPS FP16 and 2039 GB/s bandwidth handle large-scale training batches far better than L40's 90.5 TFLOPS and 864 GB/s.
L40's balanced 90.5 TFLOPS FP16 and FP32 with 48 GB VRAM supports efficient real-time inference at lower 300W TDP and $0.89 per hour average cost.
A100's superior 312 TFLOPS FP16 accelerates fine-tuning iterations on 40 GB HBM2e, outperforming L40 in memory-bound adjustments.
L40's Ada Lovelace architecture and 90.5 TFLOPS FP32 optimize generative tasks like Stable Diffusion more effectively than A100's 19.5 TFLOPS FP32.
A100 suits FP16-heavy simulations with 312 TFLOPS, while L40's 90.5 TFLOPS FP32 and lower power fit precision-focused HPC at reduced cost.
Frequently Asked Questions
Which has more VRAM: A100 SXM4 40GB or L40?▾
The L40 provides 48 GB GDDR6 VRAM, exceeding the A100 SXM4 40GB's 40 GB HBM2e. This edge aids L40 in handling slightly larger models during inference. Bandwidth differs markedly: A100 at 2039 GB/s versus L40's 864 GB/s.
Is the A100 faster than L40 for training?▾
Yes, A100's 312 TFLOPS FP16 significantly outpaces L40's 90.5 TFLOPS for training workloads. High 2039 GB/s bandwidth supports bigger batches on A100. L40 excels more in FP32 at 90.5 TFLOPS.
What are the power and pricing differences?▾
A100 SXM4 40GB consumes 400W TDP and averages $2.80 per hour, while L40 uses 300W and averages $0.89 per hour. Lower L40 costs suit budget-conscious inference. A100 justifies expense for high-throughput training.
Which GPU is newer?▾
L40 employs 2023 Ada Lovelace architecture, succeeding A100's 2020 Ampere design. Ada improvements yield balanced 90.5 TFLOPS FP16 and FP32 on L40. A100 retains FP16 lead at 312 TFLOPS.
Best for inference: A100 or L40?▾
L40 performs better for inference with 90.5 TFLOPS FP32 matching FP16, plus 48 GB VRAM. A100's 19.5 TFLOPS FP32 lags despite 312 TFLOPS FP16 strength. Pricing favors L40 at $0.67 per hour minimum.
Do they support the same form factors?▾
L40 uses PCIe exclusively, while A100 SXM4 40GB supports SXM4 and PCIe with NVLink interconnects. This enables denser A100 clusters. Both fit cloud PCIe instances.
Which is cheaper to rent, the A100 or the L40?▾
Cloud rental prices for both the A100 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 A100 have compared to the L40?▾
The A100 has 40 to 80 GB of HBM2e memory. The L40 has 48 GB of GDDR6 memory.
Can I find A100 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 A100 and the L40?▾
The A100 uses the Ampere architecture (2020) while the L40 uses Ada Lovelace (2023). The A100 delivers 3.4x the FP16 throughput and 2.4x the memory bandwidth of the L40.