Rent HPC Server
High-performance compute servers for AI, machine learning, scientific calculations and compute-intensive workloads. We exclusively use Dell PowerEdge R6715 / R7715 hardware with the current AMD EPYC 9005 'Turin' generation – up to 256 cores per system, 3 TB DDR5 ECC memory and NVMe storage. Operated in our data centers with 2N UPS redundancy in Frankfurt am Main.
- Dell PowerEdge R6715
- 1× AMD EPYC™ 9135 (16C / 3.65 GHz)
- (Expandable to 2× EPYC™ 9755)
- 384 GB DDR5 ECC
- (Expandable to 3 TB RAM)
- 4× 960 GB NVMe SSD
- (Expandable to 8× 3.84 TB NVMe)
- RAID controller: Dell PERC H965i
- iDRAC 10 Enterprise monitoring
- 5 TB traffic included
- Setup: EUR 1,095.00 · Term: 24 months
- SLA: 99.9%
- Dell PowerEdge R6715
- 1× AMD EPYC™ 9355 (32C / 3.55 GHz)
- (Expandable to 2× EPYC™ 9755)
- 512 GB DDR5 ECC
- (Expandable to 3 TB RAM)
- 4× 1.92 TB NVMe SSD
- (Expandable to 8× 3.84 TB NVMe)
- RAID controller: Dell PERC H965i
- iDRAC 10 Enterprise monitoring
- 5 TB traffic included
- Setup: EUR 1,245.00 · Term: 24 months
- SLA: 99.9%
- Dell PowerEdge R6715
- 1× AMD EPYC™ 9555 (64C / 3.2 GHz)
- (Expandable to 2× EPYC™ 9755)
- 768 GB DDR5 ECC
- (Expandable to 3 TB RAM)
- 8× 1.92 TB NVMe SSD
- (Expandable to 8× 3.84 TB NVMe)
- RAID controller: Dell PERC H965i
- iDRAC 10 Enterprise monitoring
- 5 TB traffic included
- Setup: EUR 1,395.00 · Term: 24 months
- SLA: 99.9%
- Dell PowerEdge R7715
- 2× AMD EPYC™ 9755 (128C / 2.7 GHz each)
- = 256 cores · 512 threads
- 1,536 GB DDR5 ECC
- (Expandable to 3 TB RAM)
- 12× 3.84 TB NVMe SSD
- (Expandable to 24× 7.68 TB NVMe)
- RAID controller: Dell PERC H965i
- iDRAC 10 Enterprise monitoring
- 5 TB traffic included
- Setup: EUR 1,795.00 · Term: 24 months
- SLA: 99.9%
* All prices net, plus 19% VAT. Configurations shown are for orientation. We will gladly configure your HPC server individually – with GPU acceleration (NVIDIA H100, L40S, RTX 6000 Ada), expanded RAM, NVMe storage expansion, alternative contract term or operating system of your choice.
What every HPC Server includes
Maximum compute performance on dedicated AMD EPYC hardware – with 12 memory channels per CPU and PCIe Gen5 for GPU acceleration.
Server hardware
- Dell PowerEdge R6715 (1U) or R7715 (2U)
- AMD EPYC™ 9005 'Turin' generation
- Up to 256 cores · 512 threads per system
- Dell PERC H965i RAID controller
- Enterprise NVMe / SSD storage
- Up to 3 TB DDR5 ECC memory
- PCIe Gen5 for GPU acceleration
GPU acceleration (optional)
- NVIDIA H100 (80 GB HBM3) for AI training
- NVIDIA L40S for inference & rendering
- NVIDIA RTX 6000 Ada for visualization
- Multi-GPU configurations available
Operating system
- Ubuntu LTS
- AlmaLinux · Rocky Linux
- Red Hat Enterprise Linux (RHEL)
- Windows Server
- Operating system of your choice
Data Center
All IMT-Systems data centers meet industry standard certifications. In Frankfurt am Main we use multiple high-performance data centers at Digital Realty (Interxion). Thanks to the extremely fast connection to Germany's largest internet exchange point, the DE-CIX, we offer optimal connectivity and the best possible latency within Germany and Europe.
Location & Redundancy
- 700+ carriers, ISPs and CDNs available on site
- Direct access to the DE-CIX – Europe's largest internet exchange
- Direct peering to all major cloud platforms
- 2N redundant UPS systems with N+1 diesel backup (60 hours)
- Multi-stage security: biometric access, CCTV, 24/7 security personnel
- Argon-based fire suppression with VESDA early detection
- 100% electricity from renewable sources
- 24/7 support and ECC command center on site
- GDPR-compliant data residency in Germany
Why HPC servers from IMT-Systems
High Performance Computing requires more than just many CPU cores. It needs the right combination of compute density, memory bandwidth, storage performance and network connectivity – and hardware that runs stably 24/7 under full load. That’s exactly what our HPC servers are built for.
AMD EPYC 9005 – the current generation
We use the latest AMD EPYC 9005 ‘Turin’ generation based on the Zen 5 architecture. Compared to Intel Xeon, AMD EPYC offers:
- Up to 128 cores per socket (256 cores in a dual-socket system)
- 12 memory channels per CPU – 50 % more memory bandwidth than Intel
- PCIe Gen5 with up to 128 lanes per socket for GPU and NVMe connectivity
- Higher efficiency through advanced manufacturing process
For most HPC and AI workloads, AMD EPYC is today the economically superior platform.
Real bare-metal performance
Unlike public-cloud providers, you don’t share resources with other customers. Each HPC server is physically dedicated – you get the full compute power, full memory throughput and full storage IOPS. No noisy-neighbor problem, no hidden vCPU overcommit.
Typical use cases
Artificial intelligence & machine learning: Training and inference of neural networks, LLM finetuning, computer vision, optionally with GPU acceleration (NVIDIA H100, L40S).
Scientific computing: Computational fluid dynamics (CFD), molecular dynamics simulations, climate models, genome analysis, Monte Carlo simulations.
Financial modeling: Risk calculations, algorithmic trading, portfolio backtesting, pricing models.
Rendering & 3D visualization: Architectural visualization, animation, VFX rendering, CAD/CAM calculations.
Database and analytics workloads: OLAP cubes, data warehouse queries, in-memory databases (SAP HANA style).
Optional: GPU acceleration
On request we expand your HPC server with one or more NVIDIA GPUs:
- NVIDIA H100 (80 GB HBM3): Gold standard for LLM training and large-scale AI workloads
- NVIDIA L40S: Efficient solution for inference, rendering and visual AI
- NVIDIA RTX 6000 Ada: Professional rendering and visualization
Multi-GPU configurations with NVLink connectivity are also possible.
Memory bandwidth for in-memory workloads
With 12 DDR5 channels per CPU and up to 3 TB ECC memory per server, our HPC platforms are ideal for in-memory workloads such as real-time analytics, large cache layers and memory-intensive databases. Memory bandwidth often determines the actual performance of HPC workloads more than raw core count.
Cluster-ready
You need more power than a single server provides? We also build dedicated HPC clusters for you with InfiniBand or 100 GbE interconnect, Slurm or Kubernetes orchestration and shared storage (Lustre, BeeGFS, Ceph). Get in touch for an individual cluster configuration.
GDPR-compliant data residency
All HPC servers operate exclusively in our data centers in Frankfurt am Main – Digital Realty (Interxion) FRA8 and FRA15. This means full GDPR compliance for your training data, research data and models – a critical aspect for sensitive AI projects involving personal data or proprietary research.
Personal support
You always reach engineering staff with HPC experience – no first-level tickets, no scripts. We help with performance tuning, NUMA pinning, kernel configuration and similar optimizations.
Get in touch or call directly: +49 (0)6196-2021089-0