MiTAC Introduces New AI and HPC Servers at Supercomputing Asia 2025
Singapore, Tuesday, 11 March 2025.
MiTAC unveiled state-of-the-art AI and HPC servers with enhanced memory, GPU support, and energy efficiency, designed to cater to growing European computing demands at Supercomputing Asia 2025.
Next-Generation Server Architecture
MiTAC Computing Technology Corp., a subsidiary of MiTAC Holdings Corp. (TSE:3706), has unveiled its flagship G4520G6 AI server, featuring unprecedented computing capabilities. The system supports dual Intel® Xeon® 6 processors with up to 350W TDP and incorporates 32 DDR5 DIMM slots, enabling a remarkable memory capacity of up to 8,192 GB DDR5 at 6,400 MT/s [1][3]. This architectural advancement represents a significant leap in server technology, particularly in supporting intensive AI workloads.
Advanced GPU Integration and Cooling Innovation
The G4520G6’s distinguishing feature is its support for up to eight NVIDIA H200, H100 NVL, or L40S GPUs, coupled with three PCIe 5.0 x16 expansion slots and three OCP v3.0 mezzanine slots [2]. To manage the thermal demands of such powerful components, MiTAC has implemented an advanced cooling system comprising eight hot-swap system fans and four external hot-swap fan modules, complemented by four hot-swap CRPS power supplies delivering up to 9,600W with 80+ Titanium efficiency [1].
AMD-Powered HPC Solution
Alongside the G4520G6, MiTAC introduced the TN85-B8261, a 2U server specifically engineered for HPC workloads. This system leverages AMD EPYC 9005/9004 processors and supports up to 6,144 GB of DDR5-6000 memory [3]. The server’s versatility is enhanced by its support for up to four NVIDIA Enterprise H100 or L40S GPUs, making it particularly suitable for research institutions and enterprises requiring substantial computing power [4].
Market Impact and Future Outlook
MiTAC’s strategic timing of this release aligns with growing demands in the AI and HPC sectors. The company’s expertise, dating back to the 1990s, has been instrumental in developing these energy-efficient solutions [7]. Since their initial announcement on February 24, 2025 [4], these servers have garnered significant attention for their potential to address demanding AI challenges across edge, data center, and cloud environments [4][7].