While our VMs were using the same disk, we found significant benefits from placing data, log, tempdb data, and tempdb log files on separate drives. It appears that the Windows OS views them differently when on separate drives and this in turn greatly reduces disk latency.
For some instances, we found it very helpful to create multiple data drives and spread data across them, spreading the I/O work, essentially.
When you say "separate drives", are you talking about actual physically different drives or not?
Separate Windows drives on the same VM and using the same underlying disk. For example, when we measured disk latency on some servers on their data (mdf) drive D:, we created two more data drives E: and F: and then in SQL Server created extra data files for the databases formerly only on D:.