[Simh] Relative efficiency of (simulated) VAX architectures?

[Bailey, Scott (Global Midrange Server Services)]

It's Friday afternoon, so it seems like a perfect chance to ask a weird
question that's been bobbing around in the back of my mind:

Which model of SIMH VAX is the fastest? (or most efficient, if there's a
difference)

I ask because I am a pure hobbyist looking to get the most mileage out of the
nostalgia-inducing VMS installation running on my Linux host (which is a Core
i7-based system running 64-bit Debian). I used all of the real VAXen back in
the day and have much more attachment to the operating system than any specific
[virtual] hardware model.

That is, do I get faster execution if I run a VAX-11/780 and let SIMH emulate
the full VAX instruction set, or if I run a MicroVAX 3900 and let SIMH emulate
the architectural subset of commands (and then emulate the software emulation
of the missing instructions)? Does the VAX 8600 end up running at the same speed
as the 780? Obviously this is likely to be workload dependent, but frankly most
of my workload is booting VMS and playing old games written in BASIC or FORTRAN.

:-)

Yeah, I could break down and figure out some sort of benchmark and investigate it
myself, but I'm curious if anybody has done any of this work already, or if the
learned members of the list have theoretical understanding that would illuminate
the matter.

TGIF,
Scott