[Info-vax] Unix on A DEC Vax?

[Wilm Boerhout]

John Wallace mentioned  on 19-1-2013 11:18:

[snip]

> Who can say, with a question as non-specific as that? An Alpha
> emulator on a modern x86 being faster than an Alpha Multia might not
> be unreasonable to expect. How would one find out?
>
> I would hope that reputable suppliers would ask more specific
> questions than the one you just asked, and that some of the better
> suppliers might even offer quantifiable benchmarks (formal or
> informal) for specific workloads in given configurations, so that
> potential customers could consider performance before making a
> decision as to whether to even investigate further.
>
> But my hopes are not always realised.

Commercial emulators on modern server hardware (~3 GHz, multicore) runs 
rings around any and all VAX hardware, including SMP VAXen.

For Alpha, this same hardware gives you about 600-700 MHz equivalent 
Alpha performance. So, everything up to EV67 is on a par.

Even replacing EV68 and up (DS20E, ES45) we've seen /application/ 
performance better than on the original hardware, for VMS as well as 
Ultrix-32.

The past few years we've replaced systems that control (e.g.) warehouse 
systems. The warehouse hardware (cranes, belts etc. ) have a long 
lifespan. The VAXes and Alpha's (and the occasional PDP-11) are sturdy, 
but maintenance is very expensive and uncertain. Hardware replacement 
and emulation never, I repeat, never fails.

The software in this systems is coupled to the layout of the warehouse. 
Rebuilding the software on Linux/C++ platform is costly and the expense 
is never justified. So unless and until the warehouse is fundamentally 
rediesigned, the VMS or Ultrix application S/W stays in place. Happily, 
on an emulator.

/Wilm