How many features/attributes/columns are in each dataset?
You can see, for either the "particle snapshots" or the "group catalogs", the attributes in the data specifications page. Sum the number of dimensions for each row.
For example, for the TNG300 simulation alone, one full snapshot is ~4 TB, which contains ~ 30 billion entries, each labeled by e.g. 50 attributes (for gas). At the catalog level, i.e. comparable to galaxy survey data, this snapshot will have 15 million entries with ~180 attributes each. In both cases, we then also sample across ~100 epochs in temporal evolution.