Please note that in both cases it obviously depends on where your bottleneck is. E.g. if you're IO bound, you're not going to see a lot of improvements. Similarly if most of your time is spent doing index scans there'll not be that much efficiency gains. Although in both cases parallelism can help.

If instead you're doing a bunch of aggregates over table, and the bottleneck is the computation of the transition values, computation of hashes, tuple deforming, etc... you might see quite some benefits (first two some in v10 already, all three hopefully in v11).