HP/DEC/Compaq

A while back I got a Personal DECstation 5000/25. While it's a solid little machine, I have yet to really find the interest. Yes, it was a member of the competitively fast DECstation lineup that kept DIGITAL a bit more relevant until Alpha and NVAX could be rolled out the door, but what, if anything, really made them interesting in design or use? It seems kind of like a generic MIPS box with a variant of the generic "proprietary 32-bit expansion bus with index-card size cards" done up with a decent implementation of BSD.

The DECstation series is interesting in that it was a RISC adaptation of the VAXstation 4000 system architecture. I seem to remember that it was developed and released in a very short timeframe. TurboChannel was open, like SBus, in that its documentation and licenses were easily available, but unlike SBus it was not widely used by third parties since there were no TC clone machines. (Kubota Graphics and maybe E+S did develop 3D subsystems for the TC.) The VAX influence can be seen everywhere, but most saliently by the use of VAX keyboards and mice. How ironic, then, that they did not support the architectural features needed to run VMS.

There was some interesting work on computer music and teleconferencing that was supported by DEC WRL, leading to the release of the 'LoFi' base rate audio I/O TurboChannel card and its associated external amplifier box. This was contemporaneous to other DSP experiments at NeXT and SGI, as were the ISDN features similar to the ports on the SS10 or the Indy. Quite a number of researchers on the MBone, or using real-time kernel extensions, used them. The Open Group used DECstations extensively, and an early release of OSF/1 exists on them. Other OS researchers also used them, like Ousterhout's Sprite team and the developers of the Chorus microkernel that was the basis for Cellular IRIX.

The Personal DECstation was also a rare machine for the era that used one serial bus for all its input peripherals. At the time, only ADB (used by Apple and NeXT) had that architecture; other workstations used multiple rs232 or rs423 signals for each device, or, like PCs, had a single serial bus per device. The Access.Bus interface was later refined into VESA DDC, losing the ability to connect input devices. The system bus was flexible enough to accommodate 64-bit R4000s when they were available, unlike the situation with the Indigo, which required a new IP20 design.

3D on Ultrix and OSF/1 were an afterthought, though, so the systems are not as interesting as SGIs.

robespierre wrote: The Personal DECstation was also a rare machine for the era that used one serial bus for all its input peripherals. At the time, only ADB (used by Apple and NeXT) had that architecture; other workstations used multiple rs232 or rs423 signals for each device, or, like PCs, had a single serial bus per device.

My Vectra had that HIL (Human Interface Loop ) or the HP-IL (can never keep them straight) ... that would be about 1985 ? Earlier than this or later ? The keyboard, mouse, and trackball were all on one 4-wire interface. Pretty sure the HP workstations used that same system.

The HP version of DOS reserved 4K at the top of conventional memory to deal with the input loop, that often caused difficulties with HIMEM.SYS and some programs .... remember all the fun with CONFIG.SYS and AUTOEXEC.BAT and reordering device drives to maximize memory (by 2k) and reduce conflicts ? Good times

Pretty sure the HP workstations used that same system.

They did. My HP 9000/350 has HP-HIL, for example. It was available at least as early as the /310 (1985) and I'm pretty sure the /200s have it also, so it predates ADB by at least several years.

I like the scientific papers from the DEC research lab that contain MIPS-based DEC hardware and information about the
inner workings of X11. I once played with the thought to buy a DECstation just because I wanted to learn more about
X11 internals. Now, some Ultrix versions are open-source and this makes a DECstation even more attractive.
I like the simple and elegant DECstation graphic card architecture which offloads complexity to the processor.

Get the X11 Volume Zero book from O'Reilly, that will explain all about X11 on the wire.

http://shop.oreilly.com/product/9781565920835.do

Thanks for pointing out the HP-HIL counterexample. I don't want to quibble, but isn't it more of a ring?
HP seem to have dropped HIL sometime in between the C180 and C200 workstations, in 1997.
(For comparison, the Personal DECstation had been discontinued for three years by then.)
And of course daisy-chain wasn't a new idea, HP and Commodore had been using it since the '70s.

robespierre wrote: Thanks for pointing out the HP-HIL counterexample. I don't want to quibble, but isn't it more of a ring?
HP seem to have dropped HIL sometime in between the C180 and C200 workstations, in 1997.
(For comparison, the Personal DECstation had been discontinued for three years by then.)
And of course daisy-chain wasn't a new idea, HP and Commodore had been using it since the '70s.

In hardware, yes. In software HIL doesn't work with any 64-bit version of HP-UX, just 32.

Are you still fighting with that thing?

pentium wrote: Are you still fighting with that thing?

Off and on. If there had been more "who cares" posts I might have passed it on. I find I use VAX and Alpha more, so I was looking for what the MIPS DECs did well/made them "special"

SAQ wrote: I find I use VAX and Alpha more, so I was looking for what the MIPS DECs did well/made them "special"

Faster Ultrix while keeping the |d|i|g|i|t|a|l| logo, mostly. Which was great when DEC was easier to get than Sun due to generous university subsidies. But Ultrix was a lagging flavor, feature-wise (dynamic shared libraries anyone?) and you could only step up to OSF/1 when you could afford to move to AXP.

PMAX made a great Project Athena workstation, but it wasn't a software platform unique to that hardware. And the research systems you mentioned were very interesting, but many such systems were brought up on multiple platforms, so Sprite for example supports SPARCstations and DECstations, and I think started out on Sun-3 and VAX.

They were fairly popular in academia. The platform hardware was relatively simple, and well documented (ftp links to the official docs are available here for anyone curious: http://www.linux-mips.org/wiki/DECstation ). And DEC didn't make it too difficult for Universities to get official Ultrix source code access. So it featured in a lot of OS research/development, networking (e.g. network stack research), OS benchmarking, and open source OS porting projects. The original MIT Exokernel was developed for the DECstation; it was also a major platform for CMU Mach, and even used for early Windows NT MIPS work.

I have a DECstation 5000/240 which is now maxed out with RAM. It's a great box and you can definitely tell the difference in speed compared to a comparable era VAX (or indeed early Alphas). The graphical display especially just feels snappy. I am currently running Ultrix on it.

kramlq wrote: They were fairly popular in academia. The platform hardware was relatively simple, and well documented (ftp links to the official docs are available here for anyone curious: http://www.linux-mips.org/wiki/DECstation ). And DEC didn't make it too difficult for Universities to get official Ultrix source code access. So it featured in a lot of OS research/development, networking (e.g. network stack research), OS benchmarking, and open source OS porting projects. The original MIT Exokernel was developed for the DECstation; it was also a major platform for CMU Mach, and even used for early Windows NT MIPS work.

They were fairly popular in mining and related fields for some reason. Back when they would show up in surplus places they usually had a asset sticker from a mining related company, mining related school program, or government agency that dealt with mining.

Mining without Interlocked Pipeline Stages