Actually, the KL10 was not the reference machine, the VAX 11/780 was. 
There was no reference machine for the MIPS benchmark, the 780 was the
ref machine for the SPEC benchmarks. 

VUPS on the other hand is purely a DEC thing. I like to consider it as 
meaning Virtually Useless Performance Statistic, but in actuality it's not
any worse than SPECmarks or MIPS or dhrystones as a benchmark of speed. 
The 11/780 is marked at 1 VUP. The uVII is marked at .9 VUPs, and so on 
and so forth. The 11/780 actually weighs in somewhere around 1.1 MIPS
by usual tests. Especially with an instruction set as rich as the VAX's, 
MIPS doesnt mean very much. It means equally little on RISC machines for 
the opposite reason :-)
Modern systems, of any type, are all an incestuous mix of techniques 
anyway. (i do not consider RISC a 'technology' any more than i consider 
a particular kind of tire tread pattern a 'technology' They're 
techniques. The chemistry of the rubber and the processes of manufacture, 
that's technology. The technology in the processors is CMOS VLSI in just 
about every mainstream piece of circuitry we'd be concerned with, and 
each is rather the same with its contemporaries in terms of technology.. :)

when all is said and done, though, a 4000/90 is a quite fast machine. I
daresay it would seem faster than its 32 VUPs would make one think. The 
VAXen i have used (older ones mostly, nothing newer than an SOC in my 
personal collection) tend to be more efficient all around and so the 2.7 VUPs
of my uVIII feels a lot more like a 4 or 5 MIPS machine than the 3 it 
supposedly is. Benchmarks are merely numbers, though- test various 
systems with your own programs and see how they perform. 

isildur

 On Thu, 27 Jul 2000, hg/jb wrote:

> Vup is equivalent to approx .9 of the mythological MIP (where the KL10
> was 1.0 MIP in 1980).  If you want to compare compile times on a 4000/90
> versus a uVII, the 90 is faster than lightening!!
> 
> 
> Magnus Gr|nlund wrote:
> > 
> > >
> > > I have a 4000/60
> > >
> > > I found on the netbsd web pages a page about performances of VAXen
> > >
> > > there is written speed factor 12 and compared to the 4000/90 which is 32
> > >
> > > is it possible that 4000/90 is really twice faster than 4000/60 ??
> > > anyway how to  compare this (unknown scale) speed factor to something
> > > resembling SPECint95 ???
> > >
> > The speed factor is VUPS (Vax Unit of Performance) which is almost
> > like MIPS.
> > For such low mips values you can compare it to SPECint95 by dividing
> > it with 50. (It all ends up as rounding errors anyway. :-)
> > 
> > So for the 4000/60 thats about 0.2-0.3 SPECint95 and for the 4000/90
> > about 0.6-0.7 SPECint95.
> > 
> > /Magnus
>