Using the bunching method, writing 8,000,000 entries takes 7s (on my computer). For 8,000,000,000 entries, the time taken will be 2hrs, which I suppose is not too bad to represent every person on the planet with 1 byte on a home PC.
For best performance (which you're going to want with a file that size), first shell out to "fsutil fsinfo ntfsinfo c:" (or whichever drive you're saving to), and read the output of the "Bytes Per Physical Sector" (if it exists).
This will tell you the drives sector size and allow you to minimize disk access calls as much as possible. Old drives (and older OSes -- before Win 8, I think), primarily used 512 bytes per sector, so that's the size you want to write to disk. Some newer drives have 4096 bytes per sector, so use that when possible. (I think all the new drives with 3TB+ storage use 4096 bytes per sector, if you have/use any of those.)
When in doubt, I'd default to a dumpsize of 4096 bytes, as that's a perfect multiple of 512 anyway....
My example used a size of 1000, just out of laziness, but it's not the most efficient. If 512 is the sector size, we'd write 512 bytes at once on the drive, then only write 488 the second time... We're still making more access calls than absolutely necessary.
It's a little thing, but done billions of times, it makes a noticeable difference for us. :)