In normal workflows you end up with the files you're accessing in cache so there's no I/O to physical devices and most operations do get CPU bound, especially read only ones. For performance purposes git pretty much assumes that most of the time you're running from hot cache.