Dedicated servers
Dedicated servers designed around real operation
We design each dedicated server configuration according to what it actually needs to do. We do not just solve a component list, but also cooling, power draw, PCIe layout, storage configuration, expandability, serviceability, management and long-term service operation. We prepare builds primarily for operation on our infrastructure, but also for deployment at your premises or in a datacenter - always tested and tuned in our test lab.
What the solution may include
Dedicated server configuration types
We work with different types of dedicated servers according to purpose, workload and operational requirements. We also have experience with very dense multi-GPU systems and non-standard PCIe configurations.
GPU servers
Servers for compute GPU workloads. We handle GPU selection, PCIe configuration, power delivery, cooling and practical testing under load.
Storage servers
Large-capacity storage with SAS, SATA or NVMe disks. RAID, JBOD, HBA designs, ZFS, high disk slot counts.
Virtualization nodes
Servers optimized for Proxmox VE and VM workloads. Focus on RAM, fast storage, network options and stable operation.
AI / ML servers
Servers for model training, inference and compute-intensive tasks. GPU, NVLink, high-bandwidth storage, performance cooling.
Render / compute servers
Multi-GPU systems for rendering, video processing and parallel compute. PCIe bifurcation, power delivery and thermal management.
Backup servers
Dedicated backup servers. Large capacity, reliable HBA, ZFS, optionally tape or object storage.
Proxmox nodes
Servers built specifically for Proxmox clusters. We optimize RAM, storage (Ceph, ZFS, LVM), network and IPMI access.
NVMe storage servers
High-performance NVMe systems for latency-sensitive workloads. Direct PCIe connection, bifurcation, performance storage for databases and VMs.
Multi-GPU and high-density PCIe
Non-standard systems with a high number of GPUs or PCIe devices. Practical testing of IOMMU groups, power delivery and thermal management.
Design
What we consider during design
Server design is not just about choosing CPU and RAM. Every decision affects operation, performance, power draw and future serviceability.
What the server does, how intensively, whether it needs GPU, large RAM or fast storage.
Slot count, bifurcation, IOMMU groups and compatibility with planned GPUs or cards.
SAS, SATA, NVMe, slot count, RAID or HBA, connector type and future expandability.
TDP, airflow, ambient temperature, EPS connectors, total power draw under load.
1U, 2U, 4U or tower form according to technical possibilities, operational background and project requirements.
Serviceability, spare parts availability, upgrade options, remote management via IPMI/BMC.
Example builds
What a configuration can look like
Three illustrative examples of typical builds. The final configuration always follows from the real workload and measurements in a pilot environment.
An AMD EPYC server platform or equivalent, ECC RAM, NVMe storage for virtual machines, high-speed network connectivity according to the configuration and remote IPMI/BMC management.
We size each Proxmox node for the actual virtual machine requirements, particularly CPU, RAM, storage performance, backups and required availability. The number of VMs is only an indicative parameter. We prepare the configuration and price for the specific workload.
A server with one or more NVIDIA GPUs selected for the specific workload and required VRAM, a server CPU, ECC RAM and NVMe storage. Support for GPU passthrough, CUDA workloads, AI/ML, video processing and other compute tasks.
GPU infrastructure pricing primarily depends on the GPU type and count, required VRAM, CPU and RAM, storage and workload characteristics. AI/ML inference, video processing and long-running compute jobs each require different configurations. We therefore prepare every configuration and price individually.
ZFS storage with HDDs or NVMe according to the required capacity and performance, appropriate redundancy, an HBA and high-speed network connectivity. It can provide standalone backups or replication.
Storage and backup pricing primarily depends on required usable capacity, drive type, redundancy, performance, network connectivity, retention policy and the backup or replication method. We design the configuration around the data volume and recovery requirements.
Hardware designed for a specific purpose
We design each dedicated configuration with real workload in mind, not paper specifications. We focus on stability under load, thermal management, PCIe compatibility, serviceability and the possibility of future expansion within the agreed service.
Need a custom dedicated server?
Describe the workload, performance requirements and constraints. We will propose a suitable dedicated configuration and an initial service scope.