Dedicated Instant Custom GPU Cloud VPS Unmetered Domains Pricing Data Centers
Marketplace Build your server
L40S · NVIDIA GPU servers

NVIDIA L40S dedicated servers

Deploy high-performance NVIDIA L40S GPU servers optimized for AI training, LLM inference, 3D rendering, and video production. Ada Lovelace with 48 GB of ECC GDDR6 and three NVENC encoders with AV1, fitted to the machine you configure.

Uptime SLA 48 GB ECC GDDR6 Global locations
Configure now

NVIDIA L40S GPU specifications

The NVIDIA L40S excels in AI training, graphics rendering, video transcoding, and virtualization with breakthrough Ada Lovelace architecture performance.

NVIDIA L40S

The L40S GPU achieves remarkable performance metrics: 1466 TFLOPS in Tensor operations, 212 TFLOPS in RT core capabilities, and 91.6 TFLOPS in Single-precision computing power.

Architecture

Ada Lovelace

Video memory

48GB GDDR6 with ECC

CUDA cores

18,176 pcs.

Max Bandwidth

864 GB/s

Max Power

350 W

Performance Metrics

Fourth-generation Tensor Cores with FP8 support deliver outstanding computational performance for AI training and inference workloads.

FP32

91.6 teraFLOPS

FP16 Tensor Core

733 teraFLOPS

FP8 Tensor Core

1,466 teraFLOPS

RT Core

212 teraFLOPS

GPU servers engineered for demanding workloads

NVIDIA L40S GPU bare metal servers powered by the Ada Lovelace Architecture, optimized for AI training, scientific computing, and high-performance visualization.

AI training performance

The L40S GPU boosts AI workload performance by 5X over its predecessor, enabling rapid generation of high-quality images and immersive content with advanced tensor processing.

LLM and generative AI

The L40S leverages fourth-generation Tensor Cores with FP8 support, delivering outstanding computational performance to accelerate AI and data science model training.

Ray tracing acceleration

L40S GPUs elevate rendering speeds in design and engineering tasks through advanced ray tracing capabilities, perfect for architectural visualization and product design.

3D visualization

NVIDIA L40S enhances 3D visualization, enabling faster rendering and real-time handling of complex designs for increased productivity and high-fidelity outputs.

Video production

NVIDIA L40S elevates streaming and video content tasks with three video encode and decode engines, including AV1 encoding for enhanced performance and reduced TCO.

Enterprise security

The L40S GPU meets data center standards, including NEBS Level 3 readiness, and offers secure boot with root of trust technology for enhanced security.

A100 vs L40S vs H100

The three data-centre cards we fit, compared as the PCIe add-in cards that go in our servers.

L40S A100 80 GB H100 80 GB
Architecture Ada Lovelace NVIDIA Ampere Hopper
Memory 48GB GDDR6 with ECC 80GB HBM2e 80GB HBM2e
Memory Bandwidth 864 GB/s 1,935 GB/s 2,039 GB/s
Power Up to 350W Up to 300W Up to 350W
Bus interface PCIe Gen 4 x16 PCIe Gen 4 x16 PCIe Gen 5 x16
Video encode 3 NVENC with AV1 None None
NVLink No Connector, bridge by quote Connector, bridge by quote
The card alone, per month Loading... Loading... Loading...

FAQ about NVIDIA L40S GPU bare metal servers

Common questions about deploying and managing NVIDIA L40S GPU-accelerated dedicated servers for AI, rendering, and professional visualization workloads.

What makes NVIDIA L40S ideal for mixed AI and graphics workloads?

NVIDIA L40S is built on Ada Lovelace architecture, uniquely combining AI acceleration with professional graphics capabilities. Featuring 18,176 CUDA cores, 48GB GDDR6 memory, and fourth-generation Tensor Cores with FP8 support, it excels at AI model training, LLM inference, 3D rendering, and video production. The L40S delivers 1,466 teraFLOPS of FP8 performance while maintaining advanced ray tracing and DLSS 3 support for visualization workflows.

How quickly can I deploy an L40S GPU server?

A machine already racked with an L40S is delivered within about 30 minutes of payment verification. When we fit a card we hold in stock, it takes 4 to 24 hours; a card we have to buy in takes five to ten working days, and you are told which before you pay. Every L40S server includes instant OS reload, so you can iterate without opening a support ticket.

What are the performance advantages of L40S for AI workflows?

L40S provides FP8 Tensor Core acceleration specifically optimized for efficient training and inference of large language models. The 48GB GDDR6 memory supports large model sizes, while advanced tensor operations deliver 5X performance improvement over previous generation GPUs. Combined with three video encode/decode engines featuring AV1 support, L40S handles multi-modal AI workloads that combine text, image, and video processing.

Which workloads benefit most from L40S GPU servers?

L40S GPU servers excel in environments requiring both AI compute and graphics acceleration. Optimal use cases include: large language model training and inference, AI-powered image and video generation, professional 3D rendering and CAD workflows, video transcoding with AV1 encoding, virtual workstation deployments (VDI), and hybrid workloads combining machine learning with real-time visualization.