When looking at their
performance claims, it's very important to look at what they did, here:
In this slide, they're picking the point where they offer the same performance as the competitor's peak, in order to compute an efficiency metric. That much is fair, I suppose. I mean, it doesn't exactly reflect a real-world scenario, but at least it's a sort of apples-to-apples comparison.
However, they go on to compute the ratio of their performance to the competitor's, at that same point, and claim they're 54% faster (at same power). However, that doesn't mean they're 54% faster than the competitor's peak performance, just at the point where they equal the competitor's peak performance.
If we unwarped the image (is it any wonder these are monitor screenshots and not PDFs or JPEGs they sent out??), we could probably use an image editing program to find their peak performance is only like 30% faster, in this case.