GPU Dedicated Server

A GPU server is a dedicated server equipped with one or more powerful graphics cards (GPUs). These servers are optimized for computationally intensive tasks such as artificial intelligence (AI), 3D rendering, scientific simulations, and video encoding.

GPU servers are often used for the following tasks:

• Artificial intelligence (AI) / machine learning
• Deep learning training & inference
• 3D rendering (e.g., Blender, OctaneRender)
• Video encoding & streaming (e.g. with NVENC)
• Scientific calculations & simulations
• Blockchain & mining (permitted depending on provider)
• CAD & GIS applications
• Graphics accelerators for appointment / RDP services

Our GPU-based server systems exclusively use server GPUs. These are graphics cards that are suitable for continuous operation in data centers and can be installed in appropriate rack enclosures. Various NVIDIA graphics cards are available, such as the NVIDIA Quadro RTX 5000 ADA for AI applications.

Normal servers mainly use CPUs. GPU servers also have one or more graphics cards that are optimized for parallel computing. This enables massively higher computing power for certain tasks – especially in the field of AI and graphics processing.

In general, virtualization can be performed on any of our servers. However, with NVIDIA, you may need to purchase an additional license to use certain GPU features in order to be able to use the GPU features in a virtual environment.

Ada is the codename for the architecture of the latest generation of NVIDIA GPUs, succeeding the previous generation codenamed Ampere. These are enterprise graphics cards that are more focused on AI, real-time rendering, and hybrid workloads.

Our graphics card servers are operated exclusively in highly certified data centers in Frankfurt am Main.

• Location: Germany (GDPR-compliant)
• Certifications: e.g., ISO 27001, EN 50600, Tier III or higher
• Redundant power and network connections
• 24/7 access control, video surveillance, fire protection systems

Yes – especially in the AI environment, it may be necessary to implement a strict WAN/LAN separation. Particularly if your own AI is only to be used for internal business purposes and no data from it is to be transferred to the public internet, the use of a firewall solution would be recommended here. We offer affordable PaaS solutions for protecting environments with critical data.

In general, we do not monitor our customers with regard to the services they use. As a rule, we are not aware of the purposes for which the servers are used, so we do not rule out mining at this point.

For a small AI environment, e.g., for local AI experiments, model inference (not training), and possibly fine-tuning smaller models, VRAM (graphics memory) and computing power (especially tensor performance) are crucial. The RTX 2000 ADA graphics card is a suitable entry-level option. It is suitable for inference of medium-sized models such as LLaMA 2 7B.