1.7x
Faster training vs A100
GPU Compute · Ada Lovelace
NVIDIA L40S — instant access, maximum flexibility
A versatile GPU for generative AI, LLM inference, graphics and visualization — instantly deployable as on-demand 8-GPU bare-metal servers, managed through Gravity.
from $1.00/hr
Specifications
A versatile Ada Lovelace GPU
48GB of GDDR6 with fourth-generation Tensor Cores and FP8 precision for AI, graphics and media.
Architecture Ada Lovelace
CUDA cores 18,176
Tensor cores 568 (4th gen)
GPU memory 48GB GDDR6
Memory bandwidth 1.056 TB/s
FP32 / FP16 91.6 / 183.2 TFLOPS
Server configuration
NVIDIA L40S GPU Server
A fully integrated 8-GPU node — CPU, memory, storage and networking provisioned as one dedicated server.
- GPUs
- 8× NVIDIA L40S, 384 GB total GDDR6 (48 GB each)
- CPU
- 2× AMD EPYC 9654 — 192 cores total
- Memory
- 1.5 TB DDR5
- Local storage
- 8× 3.84 TB NVMe SSD (~30 TB)
- Networking
- 2× 25 GbE (SFP28) + 4× 10 GbE (10GBASE-T)
- Management
- IPMI/BMC
Performance
A meaningful step up from A100
Faster training and inference than the previous-generation NVIDIA A100, at on-demand pricing.
1.5x
Faster inference vs A100
$1.00
Per hour, on-demand
Use cases
Workloads that drive innovation
From multi-modal generative AI to LLM inference, rendering and ray tracing.
Generative AI
Multi-modal workflows across text, image and audio on a single versatile accelerator.
LLM training & inference
FP8 precision on fourth-generation Tensor Cores accelerates large language model workloads.
Graphics & visualization
Superior rendering and visualization for design, media and content-creation pipelines.
Ray tracing
State-of-the-art hardware ray tracing for design applications and real-time rendering.
Deployment
Deploy it your way
Instant L40S access with multiple deployment paths — service, purchase or leasing.
Bare metal
8-GPU on-demand servers from $1.00 per hour — single-tenant hardware, full control.
Networked & secure
Private networks, routers, floating IPs and security groups, configured per server through Gravity.
Platform, CLI or API
Flexible access — provision and manage from the portal, the command line, or programmatically.
Backed by a 20-year track record of 100% uptime and two decades of infrastructure management expertise.
Spin up an NVIDIA L40S today
Order an 8-GPU bare-metal server and deploy in minutes through Gravity. Talk to our team to get started.