The performance results referenced in the how-to-par and trees chapters are pretty far off from what a user will experience with MPL v0.5. With automatic parallelism management, these issues are mostly gone. We should update the tutorial to reflect this.