I wonder how does the size of dataset affect the question or dashboard operating speed.
For example, when the dataset is 1 thousand and 1 billion, the operating difference is significantly different, but when the dataset is 10 thousand and 100 thousand, or 1 million, is there any obvious operation speed difference?
Metabase will only request 2,000/10,000 rows, when displaying question results - exactly for performance reasons.
The actual database size and how your database handles big datasets depends on your database and your setup - not really related to Metabase. Try doing a query locally on your database with datasets of 1k, 1m and 1b records and notice the difference in speed comes from the database.