Understanding Tendermint performance

In the COSMOS whitepaper it is stated that Tenderimint consensus has been tested in benchmarks with “7 datacenters on 5 continents, on commodity cloud instances”, and shown very good transaction throughput and reliability.

a) Is the claim that Tendermint consensus, when compared fairly, has better performance than Nakamoto style consensus algorithms?

b) if yes in a), what should one attribute this improved performance to? How should one understand the difference?