Short Description
This GPU cluster at the University of Graz (Department of Digital Humanities) provides central computational infrastructure for compute-intensive research in Digital Humanities (DH) within the DHInfra.at project. The infrastructure currently comprises four nodes with a total of 12× NVIDIA H200 (141 GB HBM3e each), 4× NVIDIA RTX PRO 6000 (96 GB each), and 4× NVIDIA L40S (48 GB each), interconnected via a 400 Gb/s InfiniBand network (NDR), plus a dedicated login, storage, and service node. An additional compute node (4× RTX PRO 6000) is planned.
The facility supports a broad spectrum of research applications: In addition to training and inference for natural language processing and computer vision (particularly automatic text recognition of historical documents and analysis of cultural artifacts), the cluster provides LLM inference APIs (OpenAI-compatible), dedicated services, and interactive Jupyter environments as managed services. This enables researchers with varying technical backgrounds – from interactive notebooks to multi-GPU batch training – to work efficiently.
The central service node enables federated authentication via SAML/Shibboleth, providing standardized access for researchers and students from the CLARIAH-AT consortium, associated institutions, and external partners. Resource allocation, usage monitoring, and project management are handled through a centralized management tool with a self-service portal. The cluster supports container-based workflows as well as traditional batch scheduling for scalable research projects.
The cluster is currently in a closed testing phase and will open to the broader DH community by June 2026 at the latest. The system specifications are expandable at both node and cluster level. The cluster is connected via a high-speed link through ACOnet, among others, to the Krems site, which enables distributed storage (Ceph-based) and coordinated use of additional GPU resources in the medium term. The system is increasingly integrated into European research infrastructures CLARIN and DARIAH via CLARIAH-AT and operates with professional support.
More information is available at https://www.dhinfra.at
Contact Person
Florian Atzenhofer-Baumgartner
Research Services
GPU computing for ML model training and inference
LLM inference APIs as managed services
Interactive Jupyter environments with GPU access
Container-based environments and batch scheduling
Federated access via SAML/Shibboleth for CLARIAH-AT partners and associated institutions
Centralized project and resource management with self-service portal
Methods & Expertise for Research Infrastructure
The Department of Digital Humanities develops and applies machine learning methods for humanities research. Focus areas include automatic text recognition of historical documents (OCR/HTR), natural language processing for text corpora analysis, and computer vision for cultural artifacts. The department has expertise in GPU cluster operations, infrastructure and networking, and container orchestration. The provision of services is supported by the IT Services of the University of Graz.
University of Salzburg
University of Innsbruck
University for Continuing Education Krems
University of Applied Arts Vienna
University of Vienna
Austrian National Library
Austrian Academy of Sciences