That's single-core/thread performance, which isn't quite the same as per-core performance, because you usually won't get N times the score by running N copies of the workload across N cores.