Data General Corporation
From as far as I can tell, here it is - Data General's final NOVA; the end of the line. It's the fastest of the Novas here and is almost as fast as the Eclipses. The CPU card is a derivation of the Eclipse S/140 .
The Nova 4 is implemented around four AMD 2901 bit- slice chips and, unlike all its Nova brethren, is microcoded. The ALU on this example is the full sixteen bits wide and uses look- ahead carry. Everything gets done on the single CPU board (in slot 1, as per DG tradition) save for a floating- point option which resided in slot two. Unlike other Novas, the boards in this one are inserted from the front rather than from the side; the backplane is in the rear and is protected by a swing- out panel with a power supply and various connectors mounted on it.
Most of what were at one time considered "options" became standard features on the Nova 4; these include power- fail detection, a real- time clock, and the primary serial- line controller. There's also a field- service- only cassette reader interface built onto the single CPU card.
Three models of Nova 4 were produced, the Nova 4/C, 4/S and the 4/X. The 4/S was the "standard" machine without a memory map and supported a maximum of 32 kW of mainstore. The 4/X, for "eXtended", had a map and could have up to 128 kW of mainstore. Hardware multiply/ divide was optional on both models. The map and MDV options were "adds" to the microcode control store. 4s came in both 5 and 16 slot backplane models. The Nova 4/C is a "compact" 4 which has the mainstore on the CPU card; the others used separate memory boards.
Rumour has it that the Nova 4 is so closely related to the Eclipse S/140 that certain enterprising individuals actually made "replacement" microcode ROMS for the Nova 4 to turn it into an S/140. This, obviously, drew the ire of Data General; I don't know whether lawsuits were filed or not as a result of any of the "modifications".
This example is a Nova 4/S - a "standard" unmapped Nova with 32 kW of MOS mainstore and the hardware multiply/ divide option in a 16- slot enclosure. In its life before coming here it served with distinction as a cyclotron- controller for TRIUMF in British Columbia - a worthy calling if you ask me.
Gone is the familiar "lights- and- switches" front panel; that is replaced by the "Virtual Console" (VC) in this machine which you "talk to" over the TTY interface. This machine is thoroughly modern in design and construction. The power supplies are switching- mode and run off a 300 VDC line and the memory (MOS only; core wasn't offered) has a battery back- up option. The CPU sports an eleven- deep instruction pipeline.
Evidence presents itself that DG was "getting desperate" with the Nova's instruction set by the time this machine was introduced. The manuals, which had previously stated memory capacity and addresses in words, now give everything in "bytes" (octets) and two new instructions are present in this machine to deal with byte pointers and operate as I/O instructions communicating with device 1 (the memory map). The "TRAP" instruction makes its entry with the Nova 4 in a nod to the 16- bit Eclipse line and its method of instruction expansion (using the "no-load"/"no-skip" notation in an ALC instruction).
This is definitely a "large- format" Nova; the boards are all of the traditional 15 x 15 inch form factor and the machine sports the customary Nova I/O and memory busses.
The colour scheme on this machine is interesting; prior to this, all the
DG machines I've seen used "mustard", orange, or blue for the primary
colours; the switches on this one are brown (my wife tells me that this
was part of the "Earth- tone" scheme DG was migrating toward), although
the trim panel is blue. I rather like it. (There are also hints that
this Nova 4 is actually living an an Eclipse S/120 enclosure.)