Our customer is a leading international technology company that develops sophisticated systems for optical capture and processing of visual information. Its business ranges from state-of-the-art camera systems and specialized sensor technology to complex software solutions that evaluate image and sensor data in real time. Applications can be found in areas such as industrial automation, research, security technology, and media production.
To ensure that these systems work accurately and reliably, enormous amounts of data must be captured, processed, and analyzed in the shortest possible time. Computing power, memory bandwidth, and fast data connectivity are crucial, especially when developing new solutions—such as for AI-supported object recognition, 3D visualization, or high-resolution image analysis.
Project period: Q3/Q4 2025
Project volume: High-end server project in the six-digit euro range.
Table of Contents
Project Description
The aim of the project was to provide a powerful GPU server platform that meets the high demands of modern visual computing workflows. The new server had to be capable of training complex AI models, analyzing large image and sensor data sets in real time, and rendering sophisticated visualizations with the highest level of detail. The hardware had to be selected and configured so that computing, storage, and network resources would work together optimally.
Key Requirements:
- Maximum GPU performance for parallel processing of large amounts of data
- High main memory capacity and bandwidth for memory-intensive workloads
- Fast and fail-safe mass storage for data analysis and AI training
- High-performance network interfaces for direct connection to data sources
- Availability and support for critical development environments
Project Implementation
The implementation was carried out with a Supermicro A+ GPU server 5126GS-TNRT2, optimized for up to ten high-end GPUs and extensive NVMe storage connectivity.
Main Components
- 2× AMD EPYC 9355
(32 cores / 64 threads, 3.55 GHz base clock, PCIe 5.0, 12-channel DDR5-6000) – offers high single and multi-thread performance and sufficient PCIe lanes for full GPU connectivity. - 10× NVIDIA RTX Pro 6000 Blackwell Server Edition (96 GB GDDR7 ECC)
– latest GPU generation with 24,064 CUDA, 752 Tensor, and 188 RT cores, ideal for AI training, rendering, and real-time analysis. - 1.5 TB DDR5-6400 ECC Registered DRAM
– high capacity and bandwidth for memory-intensive models and data pipelines. - 1× 15.36 TB Samsung PM9A3 U.2 NVMe SSD
– extremely fast and fail-safe main data storage with power loss protection and hardware encryption. - 1× 960 GB Micron 7450 PRO M.2 NVMe SSD
– dedicated to the operating system and system software with high IOPS values. - Dual 100 GbE AIOM/OCP3 network module
– enables integration into high-performance computing and data networks. - 3-year SLA 5/9/NBD on-site service
– guarantees minimal downtime and fast support.
Result
With this new server platform, the customer has a customized, future-proof infrastructure that can handle even the most demanding visual computing tasks. Large image and sensor data can be processed in real time, AI models can be trained faster, and complex visualizations achieve a new level of quality and speed. This accelerates the development of innovative products and strengthens competitiveness in the long term.