ext2/3/4 do not support disk block sizes greater than the page size. On x86-32 and x86-64, the page size is 4K, so that's the largest block size you can use. (OTOH, if you can dig up a DEC Alpha machine, I hear those had 8K pages.)
From your comment, you want to speed up mkfs. There are a couple of approaches you could take:
- If you don't need all the inodes (remember, adding more will require reformatting), you can decrease the number of inodes. You can use the
-i
or-N
options to do that. Or use one of the other types (e.g.,-T largefile
). - You could use ext4 instead. ext4 can leave block groups uninitialized until they're needed (
-O uninit_bg -E lazy_itable_init
, which are usually the default). Also has the advantage of greatly speeding up fsck.
If you really want to reduce the number of block groups (which you shouldn't), there is a -g
option for that.