Xinnor Collaborates with Megware to Enhance AI Supercomputer Helma at FAU Erlangen-Nürnberg
Xinnor has partnered with Megware, Celestica, and Phison to enhance the Helma AI supercomputer at Friedrich-Alexander-University Erlangen-Nürnberg, which now ranks third globally in the IO500 benchmark and first among Lustre-based solutions. The system features 192 nodes powered by AMD EPYC 'Genoa' CPUs and 768 Nvidia GPUs, along with Xinnor's NVMe-based RAID storage solution, ensuring high availability and fault tolerance for demanding AI applications.

Xinnor, in collaboration with Megware, Celestica, and Phison, has contributed to the Helma AI supercomputer at Friedrich-Alexander-University Erlangen-Nürnberg achieving third place in the global IO500 benchmark ranking and first among Lustre-based solutions. Helma consists of 192 nodes with AMD EPYC 'Genoa' CPUs and 768 Nvidia GPUs, ranking 51st in the Top-500 list.
The system utilizes Xinnor's NVMe-based RAID storage solution with a capacity of 4.9 PB, designed for demanding AI applications. The infrastructure supports high availability and fault tolerance using advanced clustering tools.
The implementation includes ten Celestica SC6100 storage bridges with AMD EPYC CPUs and Phison PCIe 5.0 drives. The architecture ensures no single point of failure, providing operational advantages while maintaining a small footprint and low energy requirements.




Comments