Hardware Wiki: GPU Wiki CPU Wiki

Tag AMD EPYC

AI & Compute

AI & Compute roundup: Microsoft expects Maia 300 to deliver 40% more tokens per megawatt than leading GPU systems

Two columns drawn to scale, the second 40 percent taller than the first, for Microsoft's expected 40 percent more tokens per megawatt from Maia 300 than leading GPU-based systems.

Microsoft's Maia 300 rack holds 96 accelerators and 27.6 TB of HBM in a 167-kilowatt envelope, due in 2027. NVIDIA's MGX 1.2 supports 250kW racks, and AMD plans Helion racks for November.

Saeid Yazdani
October 15, 2026

AI & Compute

Intel reports Xeon 6 at up to 1.66 times the agent throughput of 5th Gen EPYC in a one-hour test

Intel Xeon 6 agent test: up to 1.66× vs 5th Gen EPYC

Intel reports Xeon 6 at up to 1.66 times 5th Gen EPYC on agent tasks, in a benchmark that replays a recorded run with no model calls on 48 vCPU AWS instances.

Saeid Yazdani
October 12, 2026