AI chip · NVIDIA
NVIDIA L40S
Ada Lovelace data-center GPU (48 GB), targeted at mid-range training + inference + generative AI serving. Positioned between H100 and A10 in the Nvidia stack.
Market position
L40S is the PCIe workhorse for shops that wanted H100-adjacent performance without HGX complexity: 48 GB GDDR6 and 350 W into a two-slot passive card. CoreWeave, Lambda and enterprise VDI/generative workloads standardised on it as the mainstream Ada card.
Efficiency and power
How much work you get per watt and per dollar, and how much power a full rack draws.
- Perf per watt
- 2.09 FP8 TFLOPS/W733 TFLOPS ÷ 350 W = 2.09
What fits in 48 GB
Which open-source LLMs run on one NVIDIA L40S, by precision. Weights only: add roughly 15% headroom for real serving. If a model does not fit at FP16, try INT8 or INT4 (smaller quality trade-off than most people expect).
| Model | Params | FP16 | INT8 | INT4 |
|---|---|---|---|---|
| Llama 3.1 8B dense | 8 B | 16 GB ✓ | 8 GB ✓ | 4 GB ✓ |
| Llama 3.1 70B dense | 70 B | 140 GB ✗ | 70 GB ✗ | 35 GB ✓ |
| Llama 3.1 405B dense | 405 B | 810 GB ✗ | 405 GB ✗ | 203 GB ✗ |
| Llama 3.3 70B dense | 70 B | 140 GB ✗ | 70 GB ✗ | 35 GB ✓ |
| DeepSeek V3 MoE (671B total, 37B active per token) | 671 B | 1342 GB ✗ | 671 GB ✗ | 336 GB ✗ |
| DeepSeek R1 MoE (671B total, 37B active per token) | 671 B | 1342 GB ✗ | 671 GB ✗ | 336 GB ✗ |
| Qwen 2.5 7B dense | 7 B | 14 GB ✓ | 7 GB ✓ | 4 GB ✓ |
| Qwen 2.5 72B dense | 72 B | 144 GB ✗ | 72 GB ✗ | 36 GB ✓ |
| Mixtral 8x7B MoE (46.7B total, 12.9B active per token) | 46.7 B | 93 GB ✗ | 47 GB ✓ | 23 GB ✓ |
| Mixtral 8x22B MoE (141B total, 39B active per token) | 141 B | 282 GB ✗ | 141 GB ✗ | 71 GB ✗ |
| Gemma 2 27B dense | 27 B | 54 GB ✗ | 27 GB ✓ | 14 GB ✓ |
| Command R+ dense | 104 B | 208 GB ✗ | 104 GB ✗ | 52 GB ✗ |
| Kimi K2 MoE (1T total, 32B active per token) | 1000 B | 2000 GB ✗ | 1000 GB ✗ | 500 GB ✗ |
Math: FP16 = params × 2 bytes; INT8 = params × 1 byte; INT4 = params × 0.5 bytes. MoE models sum every expert (full weights on disk), not the per-token active subset.
Common questions
How much does NVIDIA L40S cost?
NVIDIA L40S doesn't have a public launch list price. Vendors like this one usually price through direct sales rather than a public sheet. See list price per TFLOP for chips that do publish.
How much memory does NVIDIA L40S have?
48 GB of GDDR6, running at 864 GB/s. Straight from the vendor datasheet. See chips with the most memory for context.
How much power does one NVIDIA L40S draw?
350 W at the chip. A full server draws more once you add CPU, memory, networking and cooling: see the rack power number above. Compare to other chips on perf-per-watt.
Which open-source LLMs fit on one NVIDIA L40S?
Of 13 open-source LLMs we track, 2 fit at FP16, 4 at INT8, and 7 at INT4 (for example Llama 3.1 8B, Llama 3.1 70B and Llama 3.3 70B). Full table above with each model's memory need.
Key facts
- Class
- Compute SoC (AI accelerator)
- Designer
- NVIDIA
- Safety-critical?
- No (data-center inference / training)
- Record as of
- 2026-09-27
Specifications (15 fields, click to expand)
Straight from the vendor datasheet. Dense throughput shown first; sparse (2:4) numbers in parentheses where the vendor publishes them. Full datasheet linked below.
- Process node
- TSMC 4N
- Transistors
- 76.3 B
- CUDA cores
- 18,176
- Tensor cores
- 568 (4th gen (Ada Lovelace))
- RT cores
- 142 (3rd gen)
- TDP
- 350 W
- Memory
- 48 GB GDDR6
- Memory bandwidth
- 864 GB/s
- PCIe
- Gen 4 x16 (64 GB/s)
- FP32
- 91.6 TFLOPS
- FP16 (dense)
- 362 TFLOPS (733 sparse)
- FP8 (dense)
- 733 TFLOPS (1,466 sparse)
- INT8 (dense)
- 733 TOPS (1,466 sparse)
- Form factor
- PCIe
- Announced
- 2023-08-08
Source: vendor datasheet
Generation
Foundry & process
- Fabbed at
- TSMC (all chips TSMC makes →)
- Process node
- TSMC 4N
- Transistors
- 76.3 B
Compare with
Sources
- NVIDIA L40SNvidia
Adoption rows appear as we confirm each chip-in-robot pairing from a public source. See every chip we track for the full catalog or the NVIDIA page.