A remotely accessible GPU server for 3D/4D tomography visualization and analysis was co-funded by MAPEX
With the rise of high-resolution X-ray imaging, datasets often exceed 1 TB per experiment, far beyond standard computing capabilities. TOMCAT addresses this critical gap by offering centralized, high-performance computing for efficient processing and visualization of large volumetric and time-resolved (4D) data. Unlike isolated workstation setups, TOMCAT is a shared, remotely accessible server infrastructure designed for cross-faculty digital collaboration on imaging data across materials science, geoscience, engineering, and life sciences.
The system includes:
A Supermicro A+ Server 4125GS-TNRT equipped with 4× NVIDIA Ada Lovelace L40S GPUs (48 GB GDDR6 ECC VRAM each, 192 GB total), dual AMD EPYC 9754 processors (128 cores / 256 threads each, 256C/512T total), 1 TB DDR5-6400 RAM, and 15.36 TB high-speed NVMe SSD storage. The server runs Dragonfly 3D World for AI-driven 3D/4D image visualization, segmentation, and quantification, alongside open-source tools such as TomoPy, ParaView, Tomviz, and ImageJ/Fiji, as well as a pre-configured Python environment with scientific and ML libraries.
Key benefits:
TOMCAT is the first instrument at MAPEX explicitly designed for cross-faculty digital collaboration on imaging data. It enables processing of terabyte-scale datasets that exceed the memory and compute capacity of standard workstations.
Authorized users access the server remotely via web-based tools, including remote desktops for GUI applications with GPU-accelerated 3D rendering. The system supports both commercial (Dragonfly) and open-source frameworks, ensuring broad usability across disciplines.
Features:
4× NVIDIA L40S GPUs with 192 GB total VRAM and 18,176 CUDA cores each for GPU-accelerated volume rendering, deep learning model training, and CT reconstruction. 1 TB DDR5 system memory for handling large datasets in-memory. 15.36 TB NVMe SSD storage for fast data access. Dragonfly 3D World software supporting AI-powered segmentation, quantification, 3D surface reconstruction, porosity analysis, and publication-ready visualization from multi-modal imaging data including X-ray micro-CT, synchrotron tomography, FIB-SEM, and correlative microscopy. Open data formats (OME-TIFF, Zarr) to ensure FAIR, scalable access.
Applied by:
Prof. Dr. Oliver Plümper
(Mineralogy, Faculty of Geosciences/ MARUM)
