NVIDIA GeForce RTX 5090 vs NVIDIA L40S, AI & Machine Learning Comparison
NVIDIA GeForce RTX 5090 wins 6 of 9 benchmarks, averaging 55.4% faster.
Both cards were measured first-party on our bench, same suite, same test rig.
What the numbers say
The gap is widest in Qwen3 32B, where NVIDIA GeForce RTX 5090 leads by 107% (71.15 vs 34.44 tok/s); the closest fight is Stable Diffusion XL (24% apart); VRAM decides part of this one: NVIDIA GeForce RTX 5090 runs 152 of our 12 AI workloads while the other card runs 7, models that don't fit score zero.
Benchmark results head-to-head
| Benchmark | NVIDIA GeForce RTX 5090 | NVIDIA L40S | Difference |
|---|---|---|---|
| Qwen3 4B tok/s | 375.6 | 212.05 | +77% |
| Llama 3.1 8B tok/s | 268.14 | 135.51 | +98% |
| Qwen2.5-Coder 14B tok/s | 149.75 | 74.6 | +101% |
| Qwen3 32B tok/s | 71.15 | 34.44 | +107% |
| Llama 3.3 70B tok/s | 0 | 16.49 | n/a |
| Stable Diffusion XL images/min | 21.08 | 17.02 | +24% |
| Z-Image Turbo images/min | 10.725 | 8.325 | +29% |
| FLUX.1 dev images/min | 2.036 | 3.836 | -47% |
| Qwen-Image-Edit images/min | 0 | 1.06 | n/a |
Whole-job comparison
How long each card takes to finish a complete pipeline, not just one model. NVIDIA GeForce RTX 5090 is faster on 1 of 2; NVIDIA L40S on 1.
| Workflow | NVIDIA GeForce RTX 5090 | NVIDIA L40S | Difference | Cost per run |
|---|---|---|---|---|
| Full codebase review | 6.7 min | 13.5 min | NVIDIA GeForce RTX 5090 2.02x faster | $0.036 vs $0.177 |
| 24-frame storyboard | 12.5 min | 4.1 min | NVIDIA L40S 3.07x faster | $0.067 vs $0.053 |
Renting by the hour, NVIDIA L40S finishes 1 of 2 cheaper. The quicker card is not automatically the cheaper way to get the work done.
Cost to rent
| Card | Per hour |
|---|---|
| NVIDIA GeForce RTX 5090 | $0.321 |
| NVIDIA L40S | $0.790 |
NVIDIA GeForce RTX 5090 is 2.46x cheaper per hour. Cheapest on-demand rate we see across RunPod and Vast.
Specifications compared
| NVIDIA GeForce RTX 5090 | NVIDIA L40S | |
|---|---|---|
| VRAM | 32GB | 48GB |
| Architecture | Blackwell (GB202) | Ada Lovelace |
| Memory bandwidth | 1792 GB/s | 864 GB/s |
| Boost clock | 2,407 MHz | 2,520 MHz |
| TDP | 575 W | 350 W |
| Launch MSRP | $1,999 | $7,500 |
| Release | 2025-01-30 | 2023-08-08 |
FAQ
- Which is better for ai & machine learning: NVIDIA GeForce RTX 5090 or NVIDIA L40S?
- NVIDIA GeForce RTX 5090 performs better for ai & machine learning, winning 6 of 9 benchmarks in our suite with an average 55.4% advantage.
- What are the main hardware differences between NVIDIA GeForce RTX 5090 and NVIDIA L40S?
- NVIDIA GeForce RTX 5090 has 32GB VRAM and a 575W TDP, while NVIDIA L40S has 48GB VRAM and a 350W TDP.
- Does VRAM matter more than speed between NVIDIA GeForce RTX 5090 and NVIDIA L40S?
- For AI, yes, NVIDIA GeForce RTX 5090 fits 6 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 GeForce RTX 5090 and NVIDIA L40S?
- Qwen3 32B: NVIDIA GeForce RTX 5090 leads by roughly 107% (71.15 vs 34.44 tok/s) in our testing.
NVIDIA GeForce RTX 5090 full review · NVIDIA L40S full review · All AI & Machine Learning rankings