MiTAC Unveils Cutting-Edge AI Server at Computex 2025
Taipei, Monday, 19 May 2025.
MiTAC introduces the NVIDIA MGX G4527G6, a scalable AI server, promising enhanced AI application performance with up to 800 Gb/s networking and improved computational efficiency at Computex 2025.
Advanced Hardware Specifications
The G4527G6 represents a significant leap in server technology, featuring support for up to eight NVIDIA GPUs and an impressive 8TB of DDR5-6400 memory capacity [1][2]. The system’s storage capabilities are enhanced with sixteen hot-swappable E1.s drives, while connectivity is managed through an NVIDIA BlueField-3 DPU and four NVIDIA ConnectX-8 SuperNICs, delivering maximum network throughput of 800 Gb/s for both InfiniBand and Ethernet connections [3].
Performance Benchmarks
The integration of NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs delivers remarkable performance improvements, achieving up to 1.8 times faster LLM inference and 1.3 times higher HPC performance compared to previous generations [2]. The platform includes over 100 customizable settings specifically designed to optimize enterprise AI workloads [3], positioning it as a versatile solution for diverse computing needs.
Enterprise AI Applications
MiTAC Computing has developed this server platform in collaboration with NVIDIA to create enterprise AI factories with comprehensive on-premises capabilities [3]. The G4527G6 is engineered to support a wide spectrum of AI-enabled applications, including autonomous decision-making systems and real-time data analytics [1]. This positions the server as a crucial component for organizations looking to enhance their AI infrastructure capabilities.
Market Availability
The official unveiling of the G4527G6 is scheduled for COMPUTEX 2025, taking place from May 30 to June 2, 2025, where MiTAC Computing will showcase the system at Booth M1110 [7]. With a track record in server solutions dating back to the 1990s, MiTAC Computing’s latest offering builds upon decades of expertise in AI, HPC, and cloud computing infrastructures [3].