HPC Servers

The distinction lies in the component density and architecture. A standard server might run a single 8-core CPU and 32GB of RAM. An HPC server, by contrast, frequently utilises processors with 64, 96, or 128 cores per socket (such as the AMD EPYC Genoa or Bergamo series). Crucially, the supporting hardware must match this power. It requires ECC memory to prevent data corruption during long calculations, and NVMe storage that bypasses the SATA controller to communicate directly with the CPU via PCIe lanes. If the network or storage cannot keep up with the processor, it is not true High-Performance Computing.

Yes. We specialize in renting out these exact HPC configurations on a monthly basis. You do not need to invest the capital expenditure (CapEx) to buy hardware like NVIDIA H100s or AMD EPYC chips. You simply rent the server from us, and we handle the power, cooling, network, and hardware maintenance. You get full root access to the machine, allowing you to install your preferred Linux distribution, hypervisor, or clustering software (like Kubernetes or Slurm) to manage your jobs.

High Performance Computing refers to the practice of aggregating computing power to solve problems that are too large for standard computers. In a business context, this usually means clustering multiple servers together to work as a single supercomputer, or using a single "fat node" with massive resources. It is used for tasks that require massive parallel processing, such as weather forecasting, rendering 3D animation, training AI models, or analyzing big data sets in real-time.

Configuration depends entirely on your specific bottleneck. If your workload is CPU-bound (like compiling code or physics simulations), you should maximise core count—we offer dual-socket systems providing 128 to 256 physical cores. If your workload is Memory-bound (like in-memory databases or large simulations), you should prioritise RAM capacity and speed; we support configurations from 512GB up to 2TB of DDR4 & DDR5 memory. For AI and Rendering, the CPU is secondary to the GPU; you would select a chassis that supports 4x or 8x GPUs. If you are unsure, our infrastructure team can review your workload requirements and recommend a balanced build.

Yes, this is possible. Because high-end HPC hardware (especially GPU servers) is in high demand and carries significant operational costs, we require a first introduction talk to understand your project and legitimacy. If you have a sufficient reason to use the trial (e.g. you would like to test our connectivity beforehand), please speak to our sales team to arrange a free trial.

The price varies significantly based on the components. A basic entry-level compute node might start around €100-€200 per month. However, a fully loaded HPC node with dual AMD EPYC processors, terabytes of RAM, and enterprise NVMe storage can range into the thousands of euros/dollars per month. Since we own our hardware and operate our own network, we can often offer more competitive pricing than public cloud providers and other bare metal providers. You can check our webshop for live pricing or ask for a custom quote for clustered setups.