NVIDIA L40
Expert rating8.6 / 10
- GPU Brand:
- NVIDIA
- VRAM:
- 48GB GDDR6
- Memory Type:
- GDDR6
- TDP:
- 300W
Consumercarts is reader-supported. Learn more
Disclaimer: The price and specification of the given product may differ from the actual product. Please confirm on the retailer's website before buying.
Manufacturer: NVIDIA Last updated: 2026-09-12
NVIDIA L40 Review
If you're comparing the NVIDIA L40 to NVIDIA's other data center GPUs, it's an Ada Lovelace-based card designed for graphics, rendering, and AI workloads in the data center, rather than a pure compute accelerator like the A100 or H100.
It has 18,176 CUDA cores and 48GB of GDDR6 in a passively-cooled, dual-slot design meant for server chassis airflow, giving data centers a GPU that's genuinely good at both 3D rendering/virtual workstation duty and generative AI inference in the same card.
Performance & Use Case
It handles cloud-based virtual workstation graphics, 3D rendering farms, and generative AI inference (image and mid-size LLM) workloads well within its 48GB frame buffer. It's built for data centers running virtual workstations, rendering-as-a-service, and generative AI inference at scale, not for desk-side workstation use.
Comparison to Alternatives
Against the NVIDIA L40S (nvidia-l40s), its direct successor released about a year later, the L40 has an identical CUDA core count and VRAM but runs at a lower 300W TDP and slightly reduced sustained clocks, since the L40S was specifically tuned for higher AI training throughput. Against the NVIDIA RTX 6000 Ada Generation (nvidia-rtx-6000-ada-generation), a desk-side workstation card sharing the same AD102 die and CUDA core count, the L40 trades active cooling and display convenience for passive, rack-server-friendly cooling.
Specifications
Platform
| GPU Brand | NVIDIA |
|---|---|
| Board Partner | NVIDIA Reference (data center/OEM only, passive cooling) |
Identity and type
| Series | RTX Ada Generation |
|---|---|
| Tier | Flagship |
Performance
| VRAM | 48 GB |
|---|---|
| Memory Type | GDDR6 |
| Boost Clock | 2490 MHz |
| Ray Tracing Support | Yes |
Physical & Power
| TDP | 300 W |
|---|---|
| Card Length | 267 mm |
| Slot Width | 2 |
Silicon
| Architecture | Ada Lovelace |
|---|---|
| Process Node | TSMC 4N |
| Shader Units | 18176 |
Memory
| Memory Bus Width | 384 bit |
|---|---|
| Memory Bandwidth | 864 GB/s |
Power
| Recommended PSU | 550 W |
|---|---|
| Power Connector | 1x 16-pin (12VHPWR) |
Host interface
| PCIe Generation | PCIe 4.0 x16 |
|---|
Market and workstation
| MSRP | 8000 USD |
|---|---|
| Launch Date | 2022-10 |
| Base / Game Clock | 735 MHz |
|---|---|
| Display Outputs | 4x DisplayPort 1.4a (rarely used in server deployment) |
| Media Engine | 3x NVENC (8th gen), 3x NVDEC, AV1 encode/decode |
Expert Review
Reviewed by Burooj Alam. Rated by us, not by users.
Expert rating8.6 / 10
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.
Best for
- cloud/virtual workstation infrastructure
- rendering-as-a-service data centers
- generative AI inference at data center scale
Not for
- desktop workstation builds (passive cooling, server-only)
- gaming
- pure AI training workloads (the L40S is tuned for this instead)
FAQs
What are the key specifications of the NVIDIA L40?
The NVIDIA L40 is a graphics card from NVIDIA. Key specs: GPU Brand NVIDIA, Tier Flagship, Board Partner NVIDIA Reference (data center/OEM only, passive cooling), VRAM 48GB GDDR6, Memory Type GDDR6.
Who is the NVIDIA L40 best suited to?
It suits cloud/virtual workstation infrastructure, rendering-as-a-service data centers, generative ai inference at data center scale.
What are the drawbacks of the NVIDIA L40?
It is a poor fit for desktop workstation builds (passive cooling, server-only), gaming, pure ai training workloads (the l40s is tuned for this instead).