Having read it, my understanding is not that it's slower, just not twice as fast (as you would normally expect) because of the controller bottleneck.
I did a ton of research before I configured things this way. For many typical workloads, RAID 0 doesn't confer much advantage because sequential read/write is not a bottleneck, and don't have high queue depths. However, I do software development and have a ton of VMs. I have to figure this must be exercising those items because RAID 0 has always provided me a pretty dramatic improvement, no benchmarking required. This
video convinced me it's worth doing. The tl;dr there is that the board I have has a faster controller, so it takes more to saturate, and even after saturation, it keeps helping with latency at high queue depths. Not sure how that is physically possible, but latency is something that a user notices, and it definitely conforms with my subjective experience.