Configuration, assembly and integration of high-performance servers, GPU systems and HPC/AI clusters for research and industry
DELTA Computer Products GmbH
Glinde (DE)New on Amplifa
About DELTA Computer Products
DELTA Computer Products GmbH, based in Glinde near Hamburg, develops, configures and integrates high-performance IT infrastructures: servers (rack chassis 1U–4U and tower), GPU servers with up to eight GPU/MIC units per system, NVIDIA DGX systems, workstations as well as scaled cluster and SuperPOD solutions based on Intel Xeon 6, AMD EPYC 9005, NVIDIA Grace Blackwell, InfiniBand and Ethernet interconnects. The systems are assembled in Glinde from components from SuperMicro, Quanta, Tyan, Asus and Gigabyte; every system undergoes an HPL Linpack stress test (minimum 48, maximum 120 hours) as well as firmware updates and test installations. Customers include research and academia, industry, AI/deep learning environments and the public sector; references include DeepL SE, MTU Aero Engines, CERN, ETH Zurich and RWTH Aachen. DELTA is an NVIDIA Elite Partner and delivers across Europe and worldwide, including installation, data center connection and support with a response time of 1–3 hours.
- Site
- Glinde
- Processes
- Configuration and assembly of high-performance servers (rack servers 1U-4U, tower), GPU/MIC server assembly (up to 10 GPU adapters per chassis, up to 8x GPU/MIC, NVLink), Cluster systems / HPC cluster integration, Hardware installation & data center connection, Quality assurance: stress testing of systems (HPL Linpack, minimum 48 hrs, max. 120 hrs per 1GB, Artificial intelligence / deep learning / machine learning / HPC system solutions, System configuration/configurator, Quality assurance including stress test (HPL Linpack, 48–120h)
- Materials
- NVIDIA DGX B300 / B200 / H100 / GB300 Grace Blackwell Superchip, AMD EPYC 9005 processors, Intel Xeon 6 processors, NVIDIA Grace CPU Superchip, DDR5 RAM up to 6400 (up to 4096GB), NVMe/SATA/SAS HotSwap storage, NVIDIA DGX (B300/B200/H100/GB200-NVL72), NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
- Certificates
- ISO 9001, ISO 14001, ISO 27001, IATF 16949, EN 9100
- Founded
- 1985
- Industries
- High Performance Computing, Artificial Intelligence, Deep Learning, Machine Learning, Research & Academia, Industry
- 01
Inquiry & Consulting
- 02
System configuration (configurator)
- 03
Component selection & incoming goods inspection
- 04
Assembly/assembly in Glinde
- 05
Firmware update & Linux test installation
- 06
Stress test (HPL Linpack, 48–120 h)
- 07
Measurement of power consumption, temperature, performance
- 08
Shipping (UPS / Kühne & Nagel, worldwide)
- 09
Installation & data center connection
- 10
RMA/replacement service & support
Technologies & Components
NVIDIA DGX (B300/B200/H100/GB200-NVL72)
Unified AI platforms with 8× SXM GPUs over NVLink for deep learning training and inference.
NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
Superchip for water-cooled GPU workstations with 252 GB HBM3E.
Intel Xeon 6
Processors (up to 350/500 W TDP) for dual-Xeon GPU servers with up to 4096 GB DDR5-6400.
AMD EPYC 9005
Processors (up to 500 W TDP) for air-cooled rack servers up to 8× GPU/MIC.
NVIDIA Grace CPU Superchip
144-core CPU with 480 GB LPDDR5X for servers with up to 4× GPU.
DDR5 RAM up to 6400 (up to 4096 GB)
Main memory in 12–32 DIMM slots depending on chassis.
NVMe/SATA/SAS HotSwap storage
Gen4/Gen5 hot-swap storage media in 2.5"/3.5" bays and E1.S.
InfiniBand & Ethernet Interconnect
NDR/EDR InfiniBand and 400/800GbE switches for cluster networking.
Sites
Glinde
DEHeadquarters with assembly, quality assurance and test center
- Lot size
- Configuration from one-off pieces to cluster scale
- Materials
- Intel Xeon 6, AMD EPYC 9005, NVIDIA DGX/Grace Blackwell, DDR5, NVMe
- System configurator
- Stress test environment (HPL Linpack)
- Test center
- Incoming goods inspection with anti-static/climate treatment
Systems & Products
Inside the production
Industries
Processes & technical limits
| Processes | Tolerance | Max. part | Weight | Lot size | Materials |
|---|---|---|---|---|---|
| Configuration and assembly of high-performance servers (rack servers 1U-4U, tower)Assembly | – | – | – | – | Server components (SuperMicro, Quanta, Tyan, Asus, |
| GPU/MIC server assembly (up to 10 GPU adapters per chassis, up to 8x GPU/MIC, NVLink)Assembly | – | – | – | – | NVIDIA GPUs, Intel Xeon, AMD EPYC |
| Cluster systems / HPC cluster integrationAssembly | – | – | – | – | NVIDIA DGX/GB200-NVL72, AMD dual EPYC, InfiniBand NDR/EDR |
| Hardware installation & data center connectionAssembly | – | – | – | – | – |
| Quality assurance: stress testing of systems (HPL Linpack, minimum 48 hrs, max. 120 hrs per 1GBTesting | – | – | – | – | – |
| Artificial intelligence / deep learning / machine learning / HPC system solutionsEngineering | – | – | – | – | – |
5 Anlagen & Maschinen
| Machine | Manufacturer / model | Quantity | Working area | Year built | Site |
|---|---|---|---|---|---|
| High-performance server (2U rack chassis, air-cooled) | DELTA (based on SuperMicro, Quanta, Tyan, Asus, Gigabyte) D22x-M2-SP-4xGPU | – | 2x Intel Xeon 6 (max. 350W TDP), 32 DIMM (max. 4096GB DDR5-6400), up to 4x GPU/MIC | – | Glinde |
| High-performance server (4U rack chassis, air-cooled) | DELTA D22x-M4-SP-8xGPU / D23z-M4-ZT-8xGPU | – | 2x Intel Xeon 6 or 2x AMD EPYC 9005, up to 8x GPU/MIC (max. 600W TDP) | – | Glinde |
| GPU workstation (tower, water-cooled) | DELTA / NVIDIA D15n-WK-GB-GB300-EDU | – | NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, 252GB HBM3E | – | Glinde |
| AI platform (NVIDIA DGX) | NVIDIA DGX B300 2304GB | – | 8x NVIDIA SXM B300 288GB GPUs, 2x Intel Xeon 6776P, ~15kW max | – | Glinde |
| Ethernet/InfiniBand switches | NVIDIA, SuperMicro, Edge-corE, QuantaMesh, Netgear e.g. NVIDIA MSN5610 (64x 800GbE OSFP), MSN5400, MSN4700 | – | up to 64x 800GbE OSFP | – | Glinde |
Additional services
Reference customers
Quality & Certifications
Regardless of form factor and value, every system undergoes an HPL Linpack stress test (one hour per 1 GB of RAM, minimum 48, maximum 120 hours) including firmware updates, test installation and measurement of power consumption, temperatures and performance values in idle and load scenarios. Configurations are archived on request, so that reorder deliveries provide identical machines.
| Evidence | Glinde |
|---|---|
| ISO 9001 | |
| ISO 14001 | |
| ISO 27001 | |
| IATF 16949 | |
| EN 9100 |
Company History
- 1985
Founded as a provider of high-performance servers and special solutions for professional IT infrastructures.
- 1990
Establishment as a specialist for SMP servers and cluster systems (1990s).
- 2000
Focus on numerical simulation, flow research and HPC workloads in Linux and Windows environments (2000s).
- 2010
Integration of NVIDIA GPU-accelerated high-performance computers; Europe-wide delivery of HPC and deep learning infrastructures.
- 2015
First delivery of NVIDIA DGX systems.
- 2018
NVIDIA Elite Partner; multiple Star Performer Central Europe awards.
- 2019
ML/AI training clusters in use across Europe; Philip Hellmann joins the company.
- 2023
Delivery of 100× NVIDIA DGX H100 (800× H100) to DeepL SE.
- 2024
HPC cluster with 192× AMD dual EPYC (12,288 cores) for MTU Aero Engines.
- 2025
40th anniversary, 8× NVIDIA DGX GB200-NVL72 for DeepL SE, opening of the new company building in Glinde.