NVIDIA L40 vs NVIDIA L40S, AI & Machine Learning Comparison
NVIDIA L40S wins 12 of 12 benchmarks, averaging 51.8% faster.
NVIDIA L40'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 LTX-Video (distilled), where NVIDIA L40S leads by 207% (2.61 vs 8.01 frames/s); the closest fight is Llama 3.1 8B (0% apart); VRAM decides part of this one: NVIDIA L40 runs 152 of our 12 AI workloads while the other card runs 11, models that don't fit score zero.
Benchmark results head-to-head
| Benchmark | NVIDIA L40 | NVIDIA L40S | Difference |
|---|---|---|---|
| Qwen3 4B tok/s | 211.33 | 212.05 | -0% |
| Llama 3.1 8B tok/s | 135.12 | 135.51 | -0% |
| Qwen2.5-Coder 14B tok/s | 74.27 | 74.6 | -0% |
| Qwen3 32B tok/s | 34.08 | 34.44 | -1% |
| Llama 3.3 70B tok/s | 0 | 16.49 | n/a |
| Stable Diffusion XL images/min | 11.2 | 17.02 | -34% |
| Z-Image Turbo images/min | 5.625 | 8.325 | -32% |
| FLUX.1 dev images/min | 2.464 | 3.836 | -36% |
| FLUX.1 Kontext dev images/min | 1.114 | 1.693 | -34% |
| Qwen-Image-Edit images/min | 0.48 | 1.06 | -55% |
| LTX-Video (distilled) frames/s | 2.61 | 8.01 | -67% |
| Wan 2.2 5B (720p) frames/s | 0.37 | 0.49 | -24% |
Whole-job comparison
How long each card takes to finish a complete pipeline, not just one model. NVIDIA L40 is faster on 0 of 8; NVIDIA L40S on 8.
| Workflow | NVIDIA L40 | NVIDIA L40S | Difference | Cost per run |
|---|---|---|---|---|
| 24-frame storyboard | 7 min | 4.1 min | NVIDIA L40S 1.73x faster | $0.080 vs $0.053 |
| 6-panel comic page | 9.1 min | 6.5 min | NVIDIA L40S 1.40x faster | $0.105 vs $0.086 |
| Character sheet, 12 poses | 12.1 min | 8.4 min | NVIDIA L40S 1.45x faster | $0.140 vs $0.110 |
| 60-second AI short film | 13 min | 4.9 min | NVIDIA L40S 2.62x faster | $0.149 vs $0.065 |
| Full codebase review | 13.5 min | 13.5 min | NVIDIA L40S 1.00x faster | $0.155 vs $0.177 |
| 10 short social clips | 28.2 min | 19.1 min | NVIDIA L40S 1.48x faster | $0.324 vs $0.251 |
| 40-product photo shoot | 41.1 min | 26.7 min | NVIDIA L40S 1.54x faster | $0.473 vs $0.352 |
| 100-photo restoration batch | 1 h 30 min | 59.6 min | NVIDIA L40S 1.51x faster | $1.038 vs $0.784 |
Renting by the hour, NVIDIA L40S finishes 7 of 8 cheaper. The quicker card is not automatically the cheaper way to get the work done.
Cost to rent
| Card | Per hour |
|---|---|
| NVIDIA L40 | $0.690 |
| NVIDIA L40S | $0.790 |
NVIDIA L40 is 1.14x cheaper per hour. Cheapest on-demand rate we see across RunPod and Vast.
Specifications compared
| NVIDIA L40 | NVIDIA L40S | |
|---|---|---|
| VRAM | 48GB | 48GB |
| Architecture | Ada Lovelace | Ada Lovelace |
| Memory bandwidth | 864 GB/s | 864 GB/s |
| Boost clock | 2,490 MHz | 2,520 MHz |
| TDP | 300 W | 350 W |
| Launch MSRP | $7,000 | $7,500 |
| Release | 2022-10-13 | 2023-08-08 |
FAQ
- Which is better for ai & machine learning: NVIDIA L40 or NVIDIA L40S?
- NVIDIA L40S performs better for ai & machine learning, winning 12 of 12 benchmarks in our suite with an average 51.8% advantage.
- What are the main hardware differences between NVIDIA L40 and NVIDIA L40S?
- NVIDIA L40 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 L40 and NVIDIA L40S?
- For AI, yes, NVIDIA L40 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 L40 and NVIDIA L40S?
- LTX-Video (distilled): NVIDIA L40S leads by roughly 207% (2.61 vs 8.01 frames/s) in our testing.
NVIDIA L40 full review · NVIDIA L40S full review · All AI & Machine Learning rankings