Baseline is 100% and we're comparing SLI scaling from 100% to 200%. So if it scales 92% worse, then it's still at 100%. If it scales 8% worse, it's at 184%. My math was right from that standpoint. But I guess the question was x16/x16 vs x8/x8 and I was answering x8/x8 & x16/x16 vs. single card. Although crossfire efficiency, which is what I was referencing, only goes from 100% to 200% (except for when it goes outside of that range) and so it was 8% less efficient as a crossfire setup.
So was my math right? Depends on what you think I was trying to say.