MiTAC Computing Technology has introduced a new portfolio of server platforms powered by 6th Gen AMD EPYC processors. The systems incorporate advanced liquid and air cooling designs to support intensive agentic AI, cloud, and enterprise workloads. But corporate clients will have to wait for full pricing details before deploying these high density racks.
For high density data centers, the new M2810Z6 and M2610Z6 platforms offer flexible configurations. Both systems are built on 2U 2 node chassis and support AMD Venice Zen 6 architecture with PCIe Gen 6 connectivity. The M2810Z6 supports SP7 CPUs up to 600W and up to 4TB of DDR5 8000 memory. The M2610Z6 targets cost conscious virtualization and edge tasks, running SP8 processors up to 400W with DDR5 6000 memory. MiTAC is not cutting corners when it comes to power delivery, ensuring these boxes can handle heavy thermal loads.
The hardware is designed for scaling out workloads. MiTAC president Rick Hwang stated that "Agentic AI and next generation enterprise workloads require infrastructure that is fast, scalable and efficient." To support this, the systems run OCP 3.0 TSFF expansion slots and hot swap NVMe storage form factors including E3.S, E1.S, and U.2. Dan McNamara of AMD explained that "enterprise AI has moved from experimentation to production" and that these new processors help scale increasingly complex data pipelines.
To accelerate GPU heavy training, MiTAC has prepared the TN85 B8261 server. This 2U dual socket box supports AMD Instinct MI350P GPUs and the AMD Pensando Pollara 400 AI NIC. It fits dual AMD EPYC 9005 or 9004 processors, up to 6TB of system memory, and 4 double width GPUs. The server includes 8 hot swap NVMe U.2 bays and Titanium efficiency redundant power supplies to keep operations online.
On the liquid cooling front, MiTAC is showing off a massive 52U high density rack. It packs 12 G4826Z5 AI servers with up to 96 AMD Instinct MI355X GPUs. This layout offers 50% higher GPU density while shrinking the data center footprint by 33%. For even higher thermal efficiency, the G8825Z5 model uses Akash Systems Diamond Cooling tech. It runs AMD Instinct MI350X GPUs and generates 50% more AI tokens per watt under typical workloads. Let's be real: scaling up AI is incredibly expensive, so getting these efficiency numbers right is a big deal.
For teams looking to benchmark their software, MiTAC also offers remote testing racks. Up to 96 developers can log in to these shared access setups to test their models before making a purchasing decision. This shared access sandbox helps developers test their code before spending a single dollar on deployment.
