MiTAC showcases 6th Gen AMD EPYC server platforms and AMD GPU-based AI infrastructure at Advancing AI

MiTAC Computing Technology unveiled a new AI and data-center infrastructure lineup at AMD Advancing AI, anchored by its M2810Z6 and M2610Z6 multi-node servers built on 6th Gen AMD EPYC processors. The company said the platforms are designed to scale AI, cloud, HPC and enterprise workloads while improving performance, efficiency and deployment flexibility through support for AMD's Zen 6 Venice architecture, PCIe Gen 6 and CXL 3.1. The M2810Z6 is a 2U, two-node, single-socket-per-node platform based on AMD SP7, supporting one 6th Gen AMD EPYC processor per node with cTDP of up to 600 W, 16 DDR5 DIMM slots, up to 4 TB of DDR5-8000 memory and front hot-swappable NVMe storage options including E3.S, E1.S and U.2. The M2610Z6 targets cost-sensitive enterprise, edge, storage, virtualization and telecom operator deployments with an AMD SP8 single-socket-per-node design, support for AMD EPYC 9006 SP8 processors up to 400 W cTDP, 16 DDR5 DIMM slots, up to 4 TB of DDR5-6000 memory and OCP 3.0 TSFF connectivity. MiTAC also highlighted broader AI infrastructure built around AMD accelerators and networking. Its TN85-B8261 is a 2U dual-socket AI server compatible with AMD Instinct MI350P GPUs and the AMD Pensando Pollara 400 AI NIC, and supports two AMD EPYC 9005- or 9004-series processors, up to 6 TB of DDR5 memory, four double-width PCIe 5.0 GPUs, eight hot-swappable NVMe U.2 bays and Titanium-efficiency redundant power supplies. For denser deployments, MiTAC said its 52U liquid-cooled rack integrates 12 G4826Z5 AI servers with up to 96 AMD Instinct MI355X GPUs and AMD Pensando Pollara 400 networking, delivering 50% higher GPU density per rack than conventional AI deployments while reducing data-center footprint by up to 33%. The company also presented the G8825Z5 with Akash Systems' Diamond Cooling technology, which it said can provide up to 50% more AI tokens per watt under standard operating conditions while sustaining performance at inlet temperatures above 95°F. MiTAC said it is also offering three AI racks equipped with AMD Instinct MI350 GPUs and AMD EPYC processors as shared remote test environments for up to 96 engineers and developers conducting validation and benchmarking work.