Aetina adds certified and new 2U short-depth GPU servers
Aetina Corporation has expanded its SuperEdge portfolio with the NVIDIA Certified System (NCS) AEX-2UA1 and a new AEX-2UA2, both built on NVIDIA's MGX modular reference architecture in compact 2U short-depth form factors.
The AEX-2UA1 chassis measures 438 x 420 x 88 mm. It has achieved NVIDIA Certified System status when configured with 2x NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, and is separately qualified by NVIDIA for the L40S and H200 NVL GPUs. It is powered by a single-socket Intel Xeon 6700-series processor with P-cores and supports dual double-width GPUs up to 600W, with NVLink available for H200 NVL configurations for GPU-to-GPU interconnect. The system includes a dedicated PCIe Gen5 x16 slot for an NVIDIA BlueField-3 DPU or ConnectX-7 NIC, up to four hot-swappable PCIe Gen5 E1.S bays, and two PCIe Gen5 M.2 slots with NVMe RAID 0/1/5/10 support.
The AEX-2UA2 uses an alternative rear-access chassis design and pairs an Intel Xeon 6700P processor with what Aetina describes as a broader range of GPU configurations, intended to suit existing rack layouts, cabling and maintenance workflows.
Aetina positions the two systems across a range of on-premises generative and agentic AI workloads: it states that NVIDIA L40S suits cost-efficient semantic search and knowledge Q&A, L40S and RTX PRO 6000 Blackwell suit 7B-13B parameter models for AI agents, H200 NVL and RTX PRO 6000 Blackwell suit 70B+ parameter models for decision-support tasks, and dual H200 NVL GPUs with NVLink support on-premises fine-tuning on proprietary data. These workload mappings are Aetina's own characterisation rather than independently verified benchmarks.
"Enterprise AI is moving from experimentation toward production, where AI agents and generative models need reliable infrastructure that operates close to enterprise data," said Jame Su, Software Division Director of Aetina. "With the AEX-2UA1 achieving NVIDIA Certified System status and the addition of the AEX-2UA2, we're giving enterprises more flexibility to match GPU performance and deployment format to their AI workloads, helping them move from development to production with confidence."
