Indeed, multi-variant Stockfish is at least 10% slower than Stockfish. In recent years I've tried very hard to minimize this slowdown because until fishnet v2, it would impact the entire fishnet including servers Lichess pays for.
My machine isn't the best for measuring performance, but last time I attempted to measure official-stockfish is about 15% faster than my fork. But again, I don't have the best hardware, so I guess I'd recommend downloading from stockfishchess.org/ and ci.appveyor.com/project/ddugovic/Stockfish/branch/master (or compiling from source; I don't know how to build it on my Windows machine, although in Appveyor it builds correctly) and doing a comparison on your own hardware, for example running in a terminal:
"stockfish bench"
My machine isn't the best for measuring performance, but last time I attempted to measure official-stockfish is about 15% faster than my fork. But again, I don't have the best hardware, so I guess I'd recommend downloading from stockfishchess.org/ and ci.appveyor.com/project/ddugovic/Stockfish/branch/master (or compiling from source; I don't know how to build it on my Windows machine, although in Appveyor it builds correctly) and doing a comparison on your own hardware, for example running in a terminal:
"stockfish bench"