Register
Speed up graphics & simulation workflows picto

Speed up graphics & simulation workflows

Improve performance for heavy workloads, like complex 3D computer-aided design (CAD) or computer-aided engineering (CAE).

Handle massive datasets picto

Handle massive datasets

Workloads like data science and simulation require large memory. Ultra-fast GDDR6 memory allows you to handle such massive workloads and datasets easily.

Compute workloads not prices picto

Compute workloads not prices

​​Experience visual computing capabilities at a fair cost to performance ratio.

GPU3 based on NVIDIA A40

Our GPU3 is based on NVIDIA A40 graphic cards known for powerful visual computing capabilities. The Ampere GPU architecture is opening a new era in performance and multi-workload capabilities. By combining the latest Ampere RT Cores, Tensor Cores and CUDA Cores with 48 GB of graphics memory the A40 delivers a unique set for visual computing workloads.


A40 is designed to enable the most advanced visualization workloads. With 2-times the throughput compared to previous generation, this GPU offers yet another speedup to heavy workloads like ray-traced motion blur or movie content.

Start your GPU3

Stay on top of the latest visual developments

GPU3 is your starting point for many of the latest visual technical developments. Whether it is cave automatic virtual environments (CAVEs), broadcast-grade streaming, working with multiple video streams, or immersive AR and VR, you are well prepared with GPU3.

Discover the best cost to performance ratio

Our GPU3 comes at a fairly priced cost to performance ratio. Choose between four different options, from 1 to 8 GPUs coming with SSD storage from 100 GB up to 1.6 TB depending on the chosen instance type.

RAM CPU Cores GPU Cards Min Local Storage Max Local Storage Price / Hour ({{ currency | uppercase }})
Small 56 GB 12 Cores 1 GPU 100 GB 800 GB {{ prices.opencompute.gpu3.small[currency] | number:8 }}
Medium 120 GB 24 Cores 2 GPU 100 GB 1.2 TB {{ prices.opencompute.gpu3.medium[currency] | number:8 }}
Large 224 GB 48 Cores 4 GPU 100 GB 1.6 TB {{ prices.opencompute.gpu3.large[currency] | number:8 }}
Huge 448 GB 96 Cores 8 GPU 100 GB 1.6 TB {{ prices.opencompute.gpu3.huge[currency] | number:8 }}
  1. [1]

    Local Storage is not included in the displayed Instances price, and has a cost of {{ prices.opencompute.volume[currency] | number:8 }} {{ currency | uppercase }} / GiB hour.

  2. [2]

    Instance needs to be shutdown for hypervisor and platform updates as it cannot be live-migrated.

  3. [3]

    Available in the following Zones: DE-FRA-1

  4. [4]

    GPU3 Large and Huge Instances are only available on dedicated hypervisors.

  5. [5]

    Please note that GPU instances require account validation: access is provided with priority to established businesses, and is granted after a manual screening process.

Rendering made easy

Make photorealistic rendering, architectural design evaluations or virtual prototyping faster than ever. Rendering of ray-traced motion blur can be done in a significantly reduced timeframe. With the new generation of RT Cores used by the GPU3 the throughput is 2-times as high as with previous generations resulting in a massive performance increase.

Level up with NGC on Exoscale GPUs

Combine the simplicity and scalability of the Exoscale Cloud with the power of NVIDIA GPUs. With our Docker based template you can access the full potential of the NVIDIA GPU Cloud (NGC) and significantly reduce time to solution.


NVIDIA GPU Cloud (NGC) provides a selected set of GPU-optimized software for artificial intelligence applications, visualizations and HPC. The NGC Catalog includes containers, pre-trained models, Helm charts for Kubernetes deployments and specific AI toolkits with SDKs.


NGC catalog works with both Exoscale Compute and Exoscale SKS. The image template for SKS worker nodes embeds the optimized version of the container runtime for Nvidia cards making it quick work of starting any application from the catalog without fiddling with drivers and CUDA versions.

Learn more

GPU3 Features

Technical Specifications GPU3

Description Specifications
Graphic Card A40
Cuda Cores per card 10572
Tensor Cores per card 336
GPU memory per card 48 GB
GPU cards 1-8
CPU cores 12-96
RAM per instance 56-448
SSD local storage max. 1.6 TB (NVMe)
Zone DE-FRA-1
Works with Compute
SKS
Docker NVIDIA

Resources

Portal

Get started in our integrated environment with just a few clicks.

Why GPU3 on Exoscale

swiss knife picto
  • Shared or dedicated Hypervisors
  • Large SSD Storage
  • No resource sharing
network bandwidth picto
  • Cutting edge GPUs
  • NVIDIA cards
  • Pass-through access
cloud orchestration picto
  • Complete platform integration
  • Terraform support
  • API support
Choose from a Wide Selection of Officially Supported Templates
Multiple OS images support picto

Trusted by engineers across Europe.

When running mission critical production workloads in the cloud, a partner you can rely on makes all the difference. Our customer success engineers have helped hundreds of customers from all over Europe migrate, run and scale production workloads on Exoscale.

Contact us