GPU
With our GPU as a Service you get access to powerful GPU hardware for machine learning needs and other special computation tasks.
Product details
When creating a GPUaaS in your account, you select how many GPU cards you need. Per selected card, you get a defined slice of hardware (CPU cores, RAM, Storage), so that the overall computational power of your service scales with the number of GPU cards.
All this is dedicated hardware, meaning you don’t share the resources with other users. Because it’s dedicated hardware, it’s not possible to stop GPUaaS. Similar to our PaaS products, you pay for the GPU until you deleted it.
Using the GPU
The GPU service is accessed via SSH. Just like when creating a storage with a Linux distribution, you can either use an SSH key which is already stored in your account or add a new one.
A typical workflow would be
- create a GPU sized according to your needs.
- To do so, you need to create a project in a location with available GPU hardware. Have a look into the GPU section of your Cloud Panel for further info.
- connect via SSH to the GPU service’s public IP address
- push your data to the service
- run your computation
- pull your data
- delete the GPU
Portfolio
As the first type of GPU cards, we have AMD MI210 cards available in the location de/han1. If you don’t see this location in your location explorer, please ask your administrator to activate it for your account.