GPU server with 16x Tesla V100s

Built with NVSwitch to attain full NVLink connectivity (300 GB/s) between all 16 Tesla V100 GPUs. Save up to 90% by moving your deep learning workloads from cloud to on-premise.

Customize now

1000+ research groups trust Lambda

Easy system administration

Our GPU servers comes with Lambda Stack, which includes frameworks like TensorFlow, PyTorch, and Keras. Lambda Stack makes upgrading these frameworks easy.

Ubuntu 20.04
Keras
TensorFlow
PyTorch
Caffe
Theano
NVIDIA CUDA
NVIDIA cuDNN

Bestsellers

Our most popular options

Our bestsellers are designed to avoid CPU, memory, and storage bottlenecks for deep learning workloads.

Basic
Hyperplane Basic, a GPU Server with 4x Tesla V100

16x Tesla V100 Server

16-Way NVLink
2x Xeon Platinum 8268 (24 Cores)
768 GB of System Memory
1.92 TB NVMe OS Drive
Customizable Data Drives
-
-
$ 224,632
Customize
Premium
Hyperplane Premium, a GPU Server with 8x Tesla V100

16x Tesla V100 Server

16-Way NVLink
2x Xeon Platinum 8268 (24 Cores)
1.5 TB of System Memory
1.92 TB NVMe OS Drive
Customizable Data Drives
Dual-Port 100 Gb/sec Ethernet/IB for Storage
8x 100 Gb/sec IB for Multi-Node Training
$ 241,392
Customize
Max
Hyperplane Max, a GPU Server with 8x Tesla V100

16x Tesla V100 Server

16-Way NVLink
2x Xeon Platinum 8280M (28 Cores)
3 TB of System Memory
1.92 TB NVMe OS Drive
Customizable Data Drives
Dual-Port 100 Gb/sec Ethernet/IB for Storage
8x 100 Gb/sec IB for Multi-Node Training
$ 266,376
Customize
custom

4x or 8x Tesla V100

NVLink
Any Processor
Up to 768 GB RAM
Fully Customizable Storage
100 Gbps InfiniBand
+1 (866) 711-2025
Live Chat

Not seeing what you're looking for?

GPUs, processors, storage, networking, and memory are fully customizable. Get in touch and we'll design a system that matches your specifications.

Explore our research

Lambda's research papers have been accepted into the top machine learning and graphics conferences, including ICCV, SIGGRAPH Asia, NeurIPS, and ACM Transactions on Graphics (TOG).

HoloGAN: Unsupervised Learning of 3D Representations from Natural Images
We propose a novel generative adversarial network (GAN) for the task of unsupervised learning of 3D representations from natural images. Learn More
RenderNet: A Deep Conv. Network for Differentiable Rendering from 3D Shapes
Traditional computer graphics rendering pipelines are designed for procedurally generating 2D images from 3D shapes with high performance. Learn More
Adversarial Monte Carlo Denoising with Conditioned Aux. Feature Modulation
Denoising Monte Carlo rendering with a very low sample rate remains a major challenge in the photo-realistic rendering research. Learn More