German HPC education study reveals theory-practice gap at 102 universities
Only 23% of university HPC clusters documented for teaching – students miss hands-on skills
A new empirical study by Anna-Lena Roth and Jonas Posner systematically assessed High-Performance Computing (HPC) education across 102 German academic institutions. By reviewing module handbooks and course catalogs, the researchers identified 178 HPC-related courses and evaluated their competency coverage and curricular placement. They found that 67.6% of institutions offer at least one HPC course, but these are predominantly elective modules at the master's level, with bachelor's programs largely lacking HPC integration. Additionally, while 61.8% of institutions operate local HPC clusters, only 23.0% explicitly document these clusters as available for teaching, as most infrastructure is reserved for research.
Statistical analysis revealed a significant association between restricted teaching access to clusters and reduced emphasis on practical competencies such as resource management, cluster usage, parallel debugging, and performance analysis. The findings highlight a structural imbalance between theoretical instruction and practical HPC training in German higher education. The authors argue that without systematic integration of hands-on cluster access into curricula, students graduate without the performance-oriented skills required for modern computational science, high-performance AI workloads, and industrial HPC roles.
- 67.6% of German institutions offer HPC courses, but mostly as master's electives
- 61.8% have HPC clusters, yet only 23% document teaching availability
- Restricted cluster access correlates with weaker practical HPC competencies like debugging and performance analysis
Why It Matters
Hands-on HPC access is critical for training future engineers and scientists in AI, simulations, and data-intensive fields.