Platform tuning: Apple Silicon, NVIDIA, AMD and CPU-only machines
The operating-system and hardware knobs that change local inference speed: the Apple GPU memory limit and power modes, NVIDIA power limits and clocks, AMD backend choice, and threads, NUMA and memory channels on CPUs. Plus how to spot thermal throttling.
15 min·Verified 2026-11Pro+50 XP
Share
Two machines with the same GPU can differ by tens of percent because of power limits, driver settings, thermals and memory configuration. This lesson covers the machine-level settings that matter for inference. For each one: what it does, why it helps, and how to measure it.
Pro content
Keep reading with Inference Academy Pro
Unlock the rest of this lesson, plus every advanced lesson, hands-on lab, the larger playground models and completion certificates.