Direct answer

Thermal or power-limit throttling reduces the CPU frequency under sustained demand; cgroup throttling restricts scheduling time. They require different fixes.

How to check it

  1. Record CPU model, power plan, cooling conditions, and whether lag happens only after prolonged load.
  2. On Linux inspect lscpu; if available use sensors or your host hardware monitoring for temperatures and clocks, and compare with CPU utilization.
  3. On a container, also read cpu.stat / cpu.max; high nr_throttled indicates quota pressure, not necessarily overheating.

What to change

lscpu
# Optional if installed and supported:
sensors
cat /sys/fs/cgroup/cpu.stat
  1. For a self-hosted machine, address dust, cooling, firmware power limits or background workloads based on actual temperature/frequency evidence.
  2. For rented hardware, ask the provider about thermal or power caps rather than guessing.
  3. If there is no throttle evidence, profile Minecraft tick work before changing hardware.

Software and version notes

Compatibility and safety

Clock frequency readings can be misleading under virtualization, boost behavior and CPU frequency governors. Do not disable thermal protection or force unsafe voltages to chase TPS.

Verify the result

Run the same sustained workload after a safe hardware/power change and compare temperature, observed frequency and MSPT.

References