Consumercarts is reader-supported. Learn more
NVIDIA L40 vs NVIDIA L40S: Comparison
NVIDIA L40 and NVIDIA L40S separate most on tdp: NVIDIA L40 at 300 W, NVIDIA L40S at 350 W, 17% in favour of NVIDIA L40. Recommended PSU splits them by another 9%, again favouring NVIDIA L40. They match on 18 other specifications, including gpu brand, board partner and vram.
Specs
| Specification |  |  |
|---|---|---|
| GPU Brand | NVIDIA | NVIDIA |
| Series | RTX Ada Generation | RTX Ada Generation |
| Board Partner | NVIDIA Reference (data center/OEM only, passive cooling) | NVIDIA Reference (data center/OEM only, passive cooling) |
| VRAM | 48 GB | 48 GB |
| Tier | Flagship | Flagship |
| Memory Type | GDDR6 | GDDR6 |
| Boost Clock | 2490 MHz-1% | 2520 MHz |
| Ray Tracing Support | true | true |
| TDP | 300 W | 350 W-17% |
| Architecture | Ada Lovelace | Ada Lovelace |
| Process Node | TSMC 4N | TSMC 4N |
| Card Length | 267 mm | 267 mm |
| Slot Width | 2 | 2 |
| Memory Bus Width | 384 bit | 384 bit |
| Memory Bandwidth | 864 GB/s | 864 GB/s |
| Recommended PSU | 550 W | 600 W-9% |
| Power Connector | 1x 16-pin (12VHPWR) | 1x 16-pin (12VHPWR) |
| MSRP | 8000 | 8500 |
| Launch Date | 2022-10 | 2023-08 |
| Base / Game Clock | 735 MHz | 735 MHz |
| PCIe Interface | PCIe 4.0 x16 | PCIe 4.0 x16 |
| CUDA Cores | 18176 | 18176 |
| Display Outputs | 4x DisplayPort 1.4a (rarely used in server deployment) | 4x DisplayPort 1.4a (rarely used in server deployment) |
| Media Engine | 3x NVENC (8th gen), 3x NVDEC, AV1 encode/decode | 3x NVENC (8th gen), 3x NVDEC, AV1 encode/decode |
Verdict
NVIDIA L40
8.6 / 10 Expert Rating
A versatile data center GPU that genuinely splits the difference between graphics and AI workloads - since replaced by the L40S for AI-heavy deployments but still a strong all-rounder.
NVIDIA L40S
8.9 / 10 Expert Rating
One of the most widely deployed data center GPUs of its generation - a genuinely strong middle ground between pure graphics cards and dedicated AI accelerators.