You can directly measure the core-to-core latency.