SAQ wrote:VAXstations are also reliable, though they use more desktop-grade support componentry than the bigger VAXes (NCR 53C90 SCSI, LANCE Ethernet, the like) so they don't run as well under heavy users.
There are often, not always, differences to the larger boxes, and each company's engineering standards certainly come into play - probably depends on which year you pick too - but you'll see these controllers up into the mid-range systems at least. Very often the bus behind the controller is the difference, and then the number of these controllers - low-end systems making due with one controller on the single system bus, high-end systems supporting a dozen on a dedicated I/O bus...
They provided a VAX-compatible workstation that was originally competitive speed-wise and would interface with the DEC services and run VMS but with a graphics head and 2-user "workstation" license. Much of the silicon was shared with the MicroVAX 3100 series as well, so there wasn't that much of a design penalty.
Often the only
difference was whether or not the framebuffer was plugged in, and a jumper being switched on the motherboard.
Early VAXstations* were performance- and cost- competitive with other CISCy UNIX workstations, but they faded pretty quickly as RISC reached the market. By 1989 DEC introduced the VAXstation 3100 and the MIPS-based DECstation 3100 just about side-by-side, with the MIPS-based DECstation about 3 times faster. The DECstation line was absolutely needed to stay competitive in light of Sun's SPARC, HP's PA-RISC, etc.
If you wanted VMS, you needed a VAX. Many VAXstations got added to clusters anchored by larger models, allowing the primary user(s) to have a graphical workstation and increasing the horsepower of the whole cluster. VMS sites often used the built-in batch job management features to distribute workloads to available CPUs/systems throughout the day, so you could get a lot of bang for your workstation dollar in that kind of environment vs. cyclic utilization of many commercial UNIX workstations. (While people found lots of ways to address that, with VMS it was there out of the box. Well, after you paid for a cluster node license...
VMS allegedly had good real-time support, and I worked with groups in an academic setting where they would use VAXstation II's for process- or instrument-control and data recording, then afterwards have the machine in question reboot into a cluster to off-load the data and start processing it in batch mode using faster CPUs in the cluster. The control node would then reboot into standalone mode the next time it needed to run one of these real-time tasks. Seemed pretty neat in 1989-90.
* VAXstation I - 1984; VAXstation II - 1985; VAXstation 3200 - 1988