What queries do you use? It would be nice to have examples of "bad" ones. Some queries, like grouping, will rather regress than accelerate from sharding.

Also, do you have enough resources (processor, disk performance)? If you run into them, you also would not get any benefits, but only regression due to the fact that aggregation requires additional small resources.

Trying to shard on such a small volume looks like you're trying to do something wrong.