HPC: peak performance without a cluster
Compute-intensive workloads from numerics to genomics on dedicated hardware and GPU instances – HPC performance, booked rather than built. High performance computing has evolved from pure simulation technology into an infrastructure for simulation and machine learning: virtual HPC servers for AI provide compute-intensive workloads flexibly and as needed.
- HPC-optimised servers for maximum computing power – dedicated servers with AMD EPYC / Intel Xeon and NVMe for exclusive, constant performance.
- GPU-accelerated computing – for AI, simulations and data-intensive workloads with NVIDIA cloud GPUs.
- Flexible and without your own hardware – computing power, network and storage scale by adding nodes, billed by the hour.
- Cluster-capable sovereign cloud – made in Germany with BSI C5 attestation, predictable costs, clear architecture and a personal contact.
The cluster you don’t need to buy
In-house HPC clusters tie up capital and staff – rented high-performance computing only pays off when calculations are being carried out.
Dedicated compute nodes
Dedicated servers with AMD EPYC / Intel Xeon and NVMe deliver exclusive, consistent performance for your jobs.
GPU nodes for this purpose
For GPU-accelerated code, RTX instances can be added to the cluster – from €92.59 per month (billed by the hour).
Nodes linked
VPC networks connect computing nodes with low latency – MPI jobs run across multiple machines.
Computing time rather than investment
No procurement requests, no maintenance, no floor space – just computing time, whenever you need it.
Concrete areas of application for HPC
Companies need a cloud for HPC if their calculations take hours instead of minutes, if they train AI models, simulate complex systems or if their servers regularly hit the load limit.
Science
High performance computing enables complex scientific simulations and numerical calculations in fields such as physics, chemistry or biotechnology. Large data volumes can be processed in parallel and evaluated much faster – more on cloud for science.
Product design and optimisation
HPC accelerates simulations, prototyping and flow analyses in the product development process. Engineers test and optimise models iteratively – without physical prototypes; in detail on CFD simulation and CAD in the cloud.
Machine learning and data analysis
GPU-accelerated high performance computing considerably shortens training times for AI models and also speeds up the subsequent LLM inference. Extensive data analyses and deep learning workloads can also be scaled efficiently.
Climate modelling
Climate models require the processing of enormous data volumes and complex systems of equations. HPC clusters enable more precise forecasts through parallel calculations and high computing power.
HPC environments tailored to your requirements
Individual large-scale jobs run on a dedicated server; recurring workloads are assigned to a small cluster with a scheduler (Slurm) spread across several nodes within the VPC. Input data and results are managed by S3, whilst Kubernetes orchestrates containerised jobs via Kubernetes. Maximum performance comes from NVMe storage for extremely fast I/O, a high-speed network with high throughput and low latency for cluster computing, GPU acceleration for more parallelisation and predictably allocated vCPU/RAM/storage without hardware investment – growing datasets go into the AI data lake.
- Exclusive knots – Dedicated server with NVMe
- Scheduler-ready – Slurm & container jobs
- GPU expansion – RTX nodes as required
- Data handling – S3 for input and output
The right centron products
Customers typically implement this use case using these building blocks – which can be combined and expanded at any time.
- AMD & Intel CPUs
- NVMe as the standard
- Contractually agreed SLA
- RTX A4000 from €92.59 per month
- Dedicated, not shared
- No minimum term
- AutoScaler included
- Traffic at a fixed price
- CI/CD-ready
How much do HPC servers cost at Centron?
On-demand HPC servers: High-Performance Computing for simulation and research – billed by the hour rather than investing in your own clusters. The core component is a dedicated server on request – billed by the hour, with no minimum contract term. This is supplemented, as required, by Cloud GPU and Kubernetes. Hosting is provided in compliance with the GDPR in centron’s own certified to ISO 27001 on the basis of IT-Grundschutz data centres, with a BSI C5:2020 Type 1 attestation. New accounts receive a €200 starting credit.
| Building block | Price |
|---|---|
| Dedicated Server | on request |
| Cloud GPU | from €92.59 per month |
| Kubernetes | from €29.99 per month |
Frequently asked questions
What is high performance computing (HPC)?
Is an HPC cloud suitable for your company?
How does HPC differ from classic servers?
Can I run MPI jobs across several nodes?
What role do GPUs play in high performance computing?
How does an HPC cloud support your AI models?
How much does it cost to get started?
Is high performance computing possible with data sovereignty?
What advantages does a German HPC cloud offer?
Can this be implemented in a way that complies with the GDPR?
Are there personal contacts at centron for your HPC project?
More on AI & Machine Learning
Get expert advice
Tell us about your project. Our sales team will get back to you within one business day.
The consultation is free of charge and without obligation.