I don't know about the exact definition of "mainstream bytecode-based VM", but just for the record: both the OCaml interpreter and the GForth interpreter usually score pretty well, too.
Probably the de facto fastest interpreter is the one of the Sun HotSpot Java virtual machines. There is a paper by Robert Griesemer detailing some information, but AFAIR it is a hand-coded optimized assembly interpreter that does not too bad. LuaJIT's interpreter is doing quite well, too. (Mike Pall said in the LTU thread about trace-based compilation that the LuaJIT2 interpreter is mostly on par with the LuaJIT1 compiler, or at least not much slower.)
EDIT: replaced "LuaJIT1 interpreter" with more accurate compiler as pointed out by haberman.
Ah, sorry I should have been more specific and said "for dynamic languages."
The LuaJIT2 interpreter is on par with the LuaJIT1 JIT compiler. I don't believe LuaJIT1 had a separate interpreter -- it was just a JIT compiler that used the main Lua interpreter.
Ok, well even for dynamic languages threaded code is likely to give you a noticeable speedup. The only negative example I'm aware of was adding threaded code (+superinstructions) to the Tcl interpreter, which resulted in slowdowns sometimes.
Even though the potential is not going to be as big as for Java, Forth, and OCaml interpreters (where people frequently report 2x speedups), for example Python gains between 20 and 45%. But somebody already replied to a similar inquiry and said that ANSI C compatibility is more important than the increase in performance. (Python uses conditional compilation to achieve both.)
Probably the de facto fastest interpreter is the one of the Sun HotSpot Java virtual machines. There is a paper by Robert Griesemer detailing some information, but AFAIR it is a hand-coded optimized assembly interpreter that does not too bad. LuaJIT's interpreter is doing quite well, too. (Mike Pall said in the LTU thread about trace-based compilation that the LuaJIT2 interpreter is mostly on par with the LuaJIT1 compiler, or at least not much slower.)
EDIT: replaced "LuaJIT1 interpreter" with more accurate compiler as pointed out by haberman.