Here's an interesting graph that shows the performance differences between the various Python runtime strategies I've tried - this pertains only to the project's critical section, not the project as a whole; lower durations are better:

• 32 bit:
  
• 64 bit:
  
• Click the images to download a scalable pdf.

A note about performance:

• For your first fullsave of a given filesystem, pypy is probably fastest, because it can chunk up blocks rapidly. This is a CPU-intensive task, and Pypy shines on CPU-heavy workloads.
• For subsequent fullsaves, I like CPython 3.[34], because they have a built-in lzma module, and the task is more I/O heavy and CPU-light. For I/O-heavy loads, CPython tends to be faster than Pypy.