NVIDIA A40 vs NVIDIA L40S, AI & Machine Learning Comparison
NVIDIA L40S wins 9 of 9 benchmarks, averaging 64.3% faster.
NVIDIA A40's numbers are anchored estimates calibrated against our measured cards, pending first-party measurement. Treat small gaps as ties.
What the numbers say
The gap is widest in Z-Image Turbo, where NVIDIA L40S leads by 158% (3.225 vs 8.325 images/min); the closest fight is Llama 3.1 8B (28% apart); VRAM decides part of this one: NVIDIA A40 runs 152 of our 12 AI workloads while the other card runs 8, models that don't fit score zero.
Benchmark results head-to-head
| Benchmark | NVIDIA A40 | NVIDIA L40S | Difference |
|---|---|---|---|
| Qwen3 4B tok/s | 160.79 | 212.05 | -24% |
| Llama 3.1 8B tok/s | 106.18 | 135.51 | -22% |
| Qwen2.5-Coder 14B tok/s | 57.85 | 74.6 | -22% |
| Qwen3 32B tok/s | 26.89 | 34.44 | -22% |
| Llama 3.3 70B tok/s | 0 | 16.49 | n/a |
| Stable Diffusion XL images/min | 9.2 | 17.02 | -46% |
| Z-Image Turbo images/min | 3.225 | 8.325 | -61% |
| FLUX.1 dev images/min | 1.95 | 3.836 | -49% |
| Wan 2.2 5B (720p) frames/s | 0.31 | 0.49 | -37% |
Whole-job comparison
How long each card takes to finish a complete pipeline, not just one model. NVIDIA A40 is faster on 0 of 3; NVIDIA L40S on 3.
| Workflow | NVIDIA A40 | NVIDIA L40S | Difference | Cost per run |
|---|---|---|---|---|
| 24-frame storyboard | 10.2 min | 4.1 min | NVIDIA L40S 2.52x faster | $0.060 vs $0.053 |
| Full codebase review | 17.3 min | 13.5 min | NVIDIA L40S 1.28x faster | $0.101 vs $0.177 |
| 10 short social clips | 33.4 min | 19.1 min | NVIDIA L40S 1.75x faster | $0.195 vs $0.251 |
Renting by the hour, NVIDIA A40 finishes 2 of 3 cheaper. The quicker card is not automatically the cheaper way to get the work done.
Cost to rent
| Card | Per hour |
|---|---|
| NVIDIA A40 | $0.350 |
| NVIDIA L40S | $0.790 |
NVIDIA A40 is 2.26x cheaper per hour. Cheapest on-demand rate we see across RunPod and Vast.
Specifications compared
| NVIDIA A40 | NVIDIA L40S | |
|---|---|---|
| VRAM | 48GB | 48GB |
| Architecture | Ampere | Ada Lovelace |
| Memory bandwidth | 696 GB/s | 864 GB/s |
| Boost clock | 1,740 MHz | 2,520 MHz |
| TDP | 300 W | 350 W |
| Launch MSRP | $5,000 | $7,500 |
| Release | 2020-10-05 | 2023-08-08 |
FAQ
- Which is better for ai & machine learning: NVIDIA A40 or NVIDIA L40S?
- NVIDIA L40S performs better for ai & machine learning, winning 9 of 9 benchmarks in our suite with an average 64.3% advantage.
- What are the main hardware differences between NVIDIA A40 and NVIDIA L40S?
- NVIDIA A40 has 48GB VRAM and a 300W TDP, while NVIDIA L40S has 48GB VRAM and a 350W TDP.
- Does VRAM matter more than speed between NVIDIA A40 and NVIDIA L40S?
- For AI, yes, NVIDIA A40 fits 8 more of our 12 workloads than NVIDIA L40S. A model that exceeds VRAM doesn't run slower, it doesn't run at all, so the card that fits the model wins that workload outright.
- Where is the biggest performance difference between NVIDIA A40 and NVIDIA L40S?
- Z-Image Turbo: NVIDIA L40S leads by roughly 158% (3.225 vs 8.325 images/min) in our testing.
NVIDIA A40 full review · NVIDIA L40S full review · All AI & Machine Learning rankings