VMS on Raspberry Pi 5

Nov 12, 2023 Last reply: 2 years ago 150 Replies

Just an FYI.



I recently ran VMS (5.5-2) as a VaxStation 3100M38 under OpenSimH on a Raspberry Pi 5!.



I got VUPS rating of 30 or 31 from VUPs.com and VUPs2.com. Not bad on a credit card sized computer for £60!



It ran a digital mapping application from the 1980s successfully, other than a long-standing font corruption problem in the GPX device simulation.



Regards,



The IT world has evolved.

What would 30 780's have costed 45 years ago?

Arne

How does that compare with a real VAX or a PDP/11? My impression of a Pi Zero is that it would knock a PDP/11 into the middle of next week, and then some.

On 12/11/2023 21:05, The Natural Philosopher wrote

Yeah, I would have expected 100s or 1000s of VUPS.

A VAX 11/780 defined 1 VUP.

One VUPS is supposed to be one original VAX CPU running VMS.

Running native sure, but it's running a software emulated processor.

Ok ... impressed !

I always liked VMS - well ahead of its time. HOPING some basement-dwelling guru will create a Pi-runnable version with some more "modern" features added.

Hey, there's a Plan-9 for Pi ... there could be a VMS as well !

BUT - where did you get the OS images ? They USED to be available but the current owners pulled them off the net years ago.

Plan 9 was designed to be portable and released as source code. VMS was designed for the VAX and while it has been ported to the PC the source code has not been released. Any port to the Pi would have to be done by VMS Software who make their money selling commercial VMS licenses.

More than 30.

I don't really know the performance penalty of emulators. AIUI a lot of the performance improvement of modern computers is due to software, working with pipelines, predictive branching, etc. Processor clock is only about 1000 times faster.

I first learnt of RISC when I was given a DECstation 3100 (Mips/Ultrix not a VAXStation) to work on. Circa 1991. Much faster than VMS, but already losing the battle to the PC.

The overhead of a non-JIT instruction set emulation must be huge.

Arne

That's a more realistic figure, and what has been observed (by me).

Right, so this is about a Vax 6000 or so, right?

Note that the Raspberry Pi has a whole lot more I/O bandwidth than comparable 32-bit DEC systems, so the performance will actually be better than the VUPS measurement would indicate.

Mind you, the I/O bandwidth on the vax systems was pretty pitiful and they should never have discontinued the Decsystem-20....

But when the computer security people ask me to describe a raspberry pi, I explain to them that it's about the performance of an old vax, except that it fits in your shirt pocket.

--scott

More like the performance of an old Cray when it's running native code.

Back in my Amiga days, I played with the Transformer, a software emulation of an 8088 on a 68000. It would run MS-DOS, but very slowly - I figured about a 10x slowdown. Once just for giggles I ran Z80MU (a Z80 emulation for MS-DOS) under the Transformer. Under these two levels of emulation I fired up the CP/M BASIC interpreter and typed "PRINT SIN(whatever)". It came back with the correct answer - 7 seconds later.

That sounds like the Sinclair scientific calculator in the early 70s

Observed figures are always more realistic, they are, after all, real.

Apparently, my rPi5 has arrived today, but I'm not going to test the emulator. I'm happy to leave the VMS part of my life, in the past.

Looking back on it now. I think the strangest thing about VMS was the Manual Set.

There's a PC emulation written in JavaScript that can boot Linux (slowly). Presumably it would be possible to run Hercules in it and produce a truly painfully slow mainframe emulation.

No. Original VAX. An 11/780.

A VAX 6000 (one CPU) was about 7 VUPs, as I recall.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required