Fact-checked by Grok 2 weeks ago

Data General Nova

The Data General Nova was a pioneering series of 16-bit minicomputers introduced in late 1968 by Data General Corporation, a company founded earlier that year by former (DEC) engineers including Edson de Castro, who had led the PDP-8 design team. Designed as a compact, general-purpose system for and on-line applications, the original Nova featured a multi-accumulator architecture with four 16-bit accumulators (two indexable), a load/store design, and medium-scale integration () TTL logic circuits—making it one of the first minicomputers to incorporate 10 to 100 gates per for enhanced performance and reliability. It supported up to 32K words (64 KB) of core memory with cycle times as low as 2.6 µs, 16 bidirectional I/O data lines capable of addressing up to 62 devices, and a high-speed channel, all housed in a single rack-mount chassis consuming about 400 watts. Priced starting at around $8,000 for a 32K configuration, the Nova offered competitive performance for its era, with a 1.8 µs and an extensive instruction set including over 100 operations for arithmetic, logical, shift, and skip functions, appealing to original equipment manufacturers (OEMs) through quantity discounts. The system quickly gained popularity in scientific research, industrial control, , educational settings, and process monitoring due to its , ease of interfacing, and software support including assemblers, debuggers, and operating systems like RDOS. The Nova family evolved through the , spawning models like the (1973), (1975), and Nova 4 (1978), which maintained the core 16-bit architecture while adding features such as memory management units, faster semiconductors, and multiprocessing capabilities, ultimately influencing later personal computing designs including the and . By the late , Data General had delivered over 50,000 Nova systems, cementing its role in the revolution before transitioning to 32-bit and MV lines.

History

Founding and Origins

Data General Corporation was founded on April 15, 1968, by Edson de Castro, Richard Sogge, , and Herbert Richman, all former employees of (DEC) except Richman from . The company was established in , initially operating out of a rented beauty parlor, with the aim of entering the burgeoning minicomputer market. De Castro, who had served as the product manager for DEC's influential PDP-8 —a 12-bit system introduced in 1965—grew frustrated with DEC's direction. In late 1967, he proposed the PDP-X, a 16-bit designed as a successor to the PDP-8, emphasizing compatibility, upgradability, and the use of emerging medium-scale integration () technology to capture a larger share of the market. DEC's management rejected the PDP-X proposal, viewing it as too risky and incompatible with their existing product lines, prompting de Castro and his colleagues to leave the company. This rejection directly inspired the conceptual roots of the , positioned as a more advanced 16-bit alternative to the PDP-8, targeting similar applications but with improved performance and cost-efficiency. The founders secured initial funding totaling $800,000, including contributions from and financial backer Frederick Adler, who invested $50,000 personally. Their motivation was to develop a faster and cheaper than DEC's offerings, leveraging monolithic integrated circuits to serve scientific and industrial markets underserved by larger mainframes like IBM's System/360. The early team included de Castro for logic design and leadership, Burkhardt for , Sogge for memory and , and Herb Richman as vice president of sales. Development of the began in April 1968, with the first prototype completed by December 1968 for demonstration at the Fall Joint Computer Conference, and initial shipments occurring in February 1969. The company went public later that year, raising an additional $5 million.

Initial Design and Introduction

The initial design of the Data General Nova stemmed from the engineering team's vision to create a cost-effective, high-performance 16-bit that could outperform contemporaries like the PDP-8 while targeting and on-line applications. Key goals included a load-store with four 16-bit accumulators—two of which could function as index registers—to minimize accesses and enhance programming efficiency for tasks such as process control and . This multi-accumulator approach addressed limitations in single-accumulator designs by allowing temporary without frequent references, enabling faster execution of and logical operations. The supported a 15-bit , accommodating up to 32K words of , and prioritized simplicity and reliability for both original equipment manufacturers (OEMs) and direct end-users. The Nova's central processor was implemented using discrete medium-scale (MSI) transistor-transistor logic () circuits, consolidated onto a single 15-inch by 15-inch to reduce costs and improve compactness compared to multi-board rivals. This single-board CPU design fit into a low-profile 5.25-inch-high rack-mount , measuring 19 inches wide and 21.5 inches deep, which facilitated easy into industrial environments. utilized magnetic core technology with a cycle time of 2.6 microseconds for access in a configuration, corresponding to an approximate clock speed of 0.4 MHz, allowing basic operations like addition to complete at rates up to 384,615 per second. The architecture's emphasis on avoided more expensive components like transformers in , contributing to the system's economy and efficiency. Data General demonstrated a of the Nova at the 1968 Fall Joint Computer Conference and commercially introduced it in 1969, marking its public debut as a revolutionary priced at approximately $3,950 for the basic unit without memory or peripherals. The company positioned it as an affordable alternative to DEC systems, with full configurations including words of core memory and interfaces costing around $8,000. First units shipped in February 1969, achieving rapid market traction; by 1970, Data General had delivered 700 Novas, demonstrating strong initial sales to OEMs for applications and end-users in scientific and industrial settings. This success validated the design's focus on performance-per-dollar, helping establish Data General as a key player in the industry.

Model Evolution

The SuperNOVA, introduced in 1970, represented the first significant upgrade to the original Nova design, achieving over three times the performance through the use of Schottky TTL logic and supporting () for improved I/O efficiency. It featured a core memory cycle time of 800 nanoseconds, with optional enabling a 300-nanosecond cycle, and the SuperNOVA SC variant pioneered in commercial minicomputers. In 1971, Data General expanded its lineup with the cost-reduced Nova 1200 and Nova 800 models, targeted at entry-level markets and OEM applications. These systems used slower magnetic core memory cycles of 1.2 microseconds for the Nova 1200 and 800 nanoseconds for the Nova 800, while maintaining compatibility with the original Nova's architecture to support 4K to 32K words of memory. The Nova 2 series, launched in June 1973, incorporated more integrated medium-scale integration () logic to reduce component count and cost, available in 4-slot (Nova 2/4) and 10-slot (Nova 2/10) chassis configurations. It offered memory cycle times of 1 or 800 nanoseconds with up to 32K words, emphasizing affordability for industrial and laboratory use while preserving binary compatibility with earlier models. Released in October 1975, the Nova 3 family advanced performance with options, including a standard 700-nanosecond cycle time and compatibility across its variants like the 4-slot Nova 3/4, 12-slot Nova 3/12, and protected-memory Nova 3/D. An optional cache-like high-speed memory board provided faster access, expanding capacity to 128K words and introducing for multi-user environments. The Nova 4 series, announced in 1977, further increased integration with models such as the single-board Nova 4/C, higher-performance Nova 4/S, and extended Nova 4/X, supporting up to 1 MB of memory through advanced boards. It achieved a 400-nanosecond memory cycle time on the CPU board, with options for multiply/divide hardware and , while ensuring full with prior Nova software. In 1977–1978, Data General introduced the microNOVA family for applications, featuring single-chip processors like the mN601 (NMOS-based, 8.3 MHz clock) and later mN602 ( 4-compatible). These emphasized I/O interfaces over buses, with two 8-bit ports for simplified connectivity in distributed systems, and supported up to 128K words of . By the early 1980s, the Nova family had sold over 50,000 units worldwide, establishing Data General as a leading vendor during the .

Decline and Legacy

By the late , the Data General faced intensifying competition from Digital Equipment Corporation's VAX line of 32-bit s, introduced in 1977, which offered greater addressable memory and processing power for emerging applications. This shift toward 32-bit architectures contributed to declining sales of the 16-bit , exacerbating Data General's financial difficulties as the market demanded more scalable systems. In response, Data General pivoted to its series, initially launched as a 16-bit successor in 1974 but expanded with the 32-bit MV line by 1980 to directly challenge the VAX. Production of models effectively ceased around this time, marking the end of the line as the company focused on for commercial and scientific . The Nova's legacy endures as a cornerstone of the minicomputer revolution, democratizing access to powerful computing by delivering affordable 16-bit performance that rivaled larger systems at a fraction of the cost. Its innovative single-board and use of medium-scale circuits set benchmarks for reliability and compactness, inspiring competitors such as Prime Computer's 16-bit systems and Hewlett-Packard's series in the 1970s market. By enabling widespread adoption in scientific, educational, and industrial settings, the Nova helped transition computing from mainframe dominance to distributed environments. Culturally, the Nova era shaped Data General's high-stakes engineering ethos, which chronicled in his 1981 Pulitzer Prize-winning book The Soul of a New Machine, detailing the intense development of the Eclipse MV/8000 but rooted in the innovative, deadline-driven culture forged during the 's rapid evolution. In the , preservation efforts have revitalized interest in the , with hobbyists undertaking restorations such as the 2024 project on a Nova II system deployed in Arctic research environments. channels like Tech Tangents have documented repairs and software revival for the 1975 microNOVA variant, culminating in a functional demonstration video on May 24, 2025. Online resources, including the datageneral.uk archive, provide extensive documentation, schematics, and emulator tools to support community-driven maintenance of Nova hardware. The Nova's accumulator-based , featuring four 16-bit accumulators for efficient data manipulation, influenced early designs by emphasizing compact, versatile register sets that echoed in accumulator-centric chips like the Intel 8080. This conceptual foundation contributed to the transition from minicomputers to personal computing, with Nova-inspired elements appearing in systems like the and early Apple prototypes.

Architecture

Processor Design

The Data General Nova processor employed a 16-bit word , enabling efficient handling of in whole words rather than bytes, which accelerated transfers and arithmetic operations. At its core, the design featured four 16-bit accumulators, designated AC0 through , which served as the primary registers for manipulation and temporary . AC2 and AC3 could additionally function as registers to support indirect addressing modes. The processor also included a 15-bit for sequencing instructions, limiting the initial to 32,768 words (64 KB of ). The Nova adopted a load-store , distinguishing it from register-memory designs prevalent in contemporaries like the PDP-8; memory access instructions (such as load and store) were separate from arithmetic and logic unit (ALU) operations, promoting a cleaner separation of data movement and computation. This structure required explicit loads to transfer data from memory into the accumulators before ALU processing could occur, enhancing modularity but increasing instruction counts for complex tasks. The ALU supported fundamental operations including , , logical , and XOR, all performed between accumulators or an accumulator and immediate values, with results stored back in the accumulators and a single-bit for overflow detection. Early models relied on hardwired logic implemented with transistor-transistor logic () integrated circuits, providing straightforward execution paths for these operations. Later iterations, particularly the Nova 4 introduced in 1978, incorporated horizontally microcoded to enable prefetching and overlap, improving efficiency through more flexible sequencing of micro-operations. Clock speeds evolved significantly across the Nova family, reflecting advances in semiconductor technology and integration. The original (1969) operated with a basic time of approximately 2.6 microseconds using core , constrained by circuitry. The Nova 1200 (1970) improved to a 1.2-microsecond time, still based on . Subsequent models like the (1970) achieved around 800 nanoseconds, while the Nova 3 (1975) used NMOS technology for times around 700 nanoseconds. The Nova 4 culminated this progression with a 400-nanosecond time, leveraging NMOS and four-way interleaving for sustained high performance. Central to the processor's design was a multiplexed address/ bus in the core, which shared lines for both addressing memory and transferring , optimizing pin count and board space in the compact while supporting () for I/O devices. This bus structure integrated seamlessly with the load-store paradigm, facilitating efficient communication between the CPU and memory subsystems.

Memory and I/O Systems

The Data General Nova initially featured magnetic core memory, with standard configurations ranging from 4 KB to 16 KB. Later models transitioned to semiconductor RAM for improved speed and reliability, beginning with the SuperNOVA in 1970; the Nova 2 series supported up to 64 KB and expanding to 1 MB in the Nova 4 series. This shift allowed for larger addressable spaces while maintaining compatibility with existing core-based systems in hybrid configurations on select models like the Nova 3/12 and 3/D. Memory addressing in the Nova supported direct access to a 32 KB using 15-bit logical addresses from the . For expansion beyond this limit, memory banking was implemented via the (MMU) or Memory Management and Protection Unit (MMPU), which divided memory into 1K-word pages and enabled to access up to 128 K words. Later models, such as the NMOS variants, included optional bits—one per 16-bit word—for error detection, while the MMPU in the /D added protection features like memory access validation, I/O safeguards, and runaway deferral to prevent unauthorized or erroneous operations. The Nova's I/O system utilized 6 to 12 direct channels, configurable through the system's slots, allowing programmable interfacing with external devices. A standard (DMA) channel was included across all models for efficient data transfers between memory and peripherals, with enhancements in the and later series providing higher-speed options up to 1 MHz transfer rates in the Nova 4. Standard peripherals connected via these channels included teletypes for console interaction, paper tape readers and punches operating at speeds up to 400 characters per second, and disk drives with capacities such as 25 fixed-head units or cartridge-based subsystems akin to the RL01 offering 10 per drive. Custom expansion boards were also supported for industrial control applications, integrating sensors and actuators directly into the I/O framework. System expansion relied on a modular backplane design within 19-inch rack-mountable chassis, accommodating up to 20 slots for memory, I/O, and processor modules, with options for additional chassis to extend capacity in larger configurations like the Nova 3/12's 12-slot base. The processor bus connected these modules, facilitating shared access to memory and I/O resources without requiring extensive rewiring.

Programming Model

The programming model of the Data General Nova centered on a minimalist register set designed for efficient 16-bit operations, emphasizing accumulators for data manipulation and control. The original Nova processors featured four 16-bit accumulator registers (AC0 through AC3), which served as the primary locations for arithmetic, logical, and data transfer operations; AC2 and AC3 could additionally function as index registers for addressing. A 15-bit program counter (PC) managed instruction fetching, while a single-bit carry flag (often denoted as CRY) captured overflow from arithmetic and logical instructions. There was no dedicated skip flag register; instead, skip conditions were handled via conditional instructions that incremented the PC by two, effectively skipping the next instruction based on flags like zero, carry, or sign. Later models, such as the Nova 3 introduced in 1975, extended this set with a 15-bit stack pointer (SP) and frame pointer (FP) to support hardware-managed stacks, maintaining backward compatibility with earlier code. Addressing in the Nova was flexible yet constrained to a 15-bit memory space (32,768 words), with modes optimized for locality and to minimize instruction size. Direct addressing used an 8-bit field (octal 000 to 377, or -128 to +127 in signed interpretation) relative to the current page or page zero, allowing access to 256 words without additional overhead. Indirect addressing enabled deferred resolution by fetching the effective address from , supporting multi-level up to eight levels in some implementations. Autoindexing (or indexed addressing) added the signed to the contents of AC2 or before accessing , facilitating array traversal; deferred indirect combined with indexing for complex pointer operations. These modes were encoded in the 's upper bits, with the X field (bits 13-12) selecting between page zero/relative (00/01) and indexed via AC2/ (10/11). Stack operations relied on software conventions in early Nova models, as there was no stack pointer initially; programmers typically used or dedicated locations (e.g., 0 and 1 for vectors) for temporary storage and returns. The Jump to Subroutine (JSB) provided support by storing the return address (PC+1) in and jumping to the target, enabling nested calls through software stack management; interrupts did not automatically use the stack, requiring explicit handler saves. With the , stacks became integral, using the to point to the top of a 256-word LIFO region in for operations, while the delimited frames for variables and parameter passing during subroutines—enhancing efficiency without altering base JSB semantics. This evolution preserved subroutine compatibility across the family. The Nova handled data primarily as 16-bit two's-complement signed integers or unsigned logical words, with no native byte operations—bytes were extracted via masking and shifting within words, assuming big-endian layout. Arithmetic and logical instructions operated on these full words, supporting addition, subtraction, AND, OR, and shifts, with the propagating for multi-word operations. was absent in hardware and emulated in software using double-word (32-bit) formats with a 7-bit exponent and 24-bit , or extended 64-bit versions; such was common in scientific applications but incurred performance overhead. Binary compatibility was a of the Nova design, ensuring that from earlier models (e.g., original Nova or Nova 2) executed unchanged on successors like the or microNOVA, as extensions such as the stack registers and new instructions were optional and did not modify the core 16-bit architecture or base instruction encoding. This allowed seamless upgrades in deployments, with later processors emulating any missing features via where needed.

Instruction Set

The Data General Nova's instruction set is designed for a 16-bit , where each instruction occupies a single 16-bit word. The high three bits (bits 15-13) typically determine the instruction class: 000 for memory reference instructions (with low 13 bits encoding operation, , and an 8-bit ), 011 for I/O instructions (with low 13 bits specifying accumulator, transfer type, control action, and a 6-bit code), and arithmetic/logical instructions identified by bit 0 set to 1 (with bits 15-1 encoding source/destination accumulators, operation, shifts, carry handling, and skip conditions). for memory reference include (page zero or current page), indexed (using AC2 or AC3), and indirect, supporting a 12-bit effective range. The set comprises approximately 50 instructions through combinations of these fields, without exposure of underlying operations.

Memory Reference Instructions

These instructions handle data movement between memory and the four accumulators (ACs), as well as program flow control and memory modification, operating on the specified effective address computed from the displacement and mode bits (bits 8-10: 00 for page zero unsigned, 01 for current page signed, 10 for AC2 index, 11 for AC3 index; bit 7 for indirect if set). They form the core of load/store and branching operations in the programming model.
MnemonicOctal Opcode PrefixFormat (Bits 15-13 = 000)Behavior
JMP000 [mode] [disp]Bits 12-10 = 000Unconditional jump to effective address; PC set to that location.
JSR001 [mode] [disp]Bits 12-10 = 001Jump to subroutine: PC + 1 stored in AC3, then jump to effective address.
ISZ010 [mode] [disp]Bits 12-10 = 010Increment memory at effective address by 1; skip next instruction if result is zero.
DSZ011 [mode] [disp]Bits 12-10 = 011Decrement memory at effective address by 1; skip next instruction if result is zero.
LDA[AC]00 [mode] [disp]Bits 15-13 = 000, bits 12-10 encode AC and load (e.g., 000 for AC0 LDA), bit 9 = indirect, bits 8-7 = mode (00 page zero, 01 current page, 10 AC2 index, 11 AC3 index), bits 7-0 = 8-bit signed displacementLoad contents of effective address into specified AC; memory unchanged.
STA[AC]01 [mode] [disp]Bits 15-13 = 000, bits 12-10 encode AC and store (e.g., 100 for AC0 STA), bit 9 = indirect, bits 8-7 = mode (00 page zero, 01 current page, 10 AC2 index, 11 AC3 index), bits 7-0 = 8-bit signed displacementStore contents of specified AC into effective address; AC unchanged.
These instructions execute in 2-3 memory cycles, depending on and .

I/O Instructions

I/O instructions manage data transfer and signals to up to 64 peripheral s, using three 16-bit buffers (A, B, C) per . The format specifies the accumulator (bits 12-11), transfer type (bits 10-8), action (bits 7-6: 00 none, 01 set busy/clear done, 10 clear busy/done, 11 pulse), and code (bits 5-0, 00-77). Transfers move data between the AC and the selected buffer; skips test status flags (busy or done). The CPU is treated as 77 for functions.
MnemonicOctal PrefixTransfer (Bits 10-8)Behavior
DIA3[AC]1 [ctrl] [dev]001Input data from device buffer A to specified AC.
DOA3[AC]2 [ctrl] [dev]010Output data from specified AC to device buffer A.
DIB3[AC]3 [ctrl] [dev]011Input from buffer B to AC.
DOB3[AC]4 [ctrl] [dev]100Output from AC to buffer B.
DIC3[AC]5 [ctrl] [dev]101Input from buffer C to AC.
DOC3[AC]6 [ctrl] [dev]110Output from AC to buffer C.
SKP3[AC]7 [cond] [dev]111Skip next instruction based on condition: 00 if busy=1 (SKPBN), 01 if busy=0 (SKPBZ), 10 if done=1 (SKPDN), 11 if done=0 (SKPDZ).
These provide asynchronous device handling without dedicated I/O processors.

Arithmetic and Logical Instructions

These operate on the four 16-bit accumulators, performing and bitwise logic between source and destination ACs (bits 1-2 source, bits 3-4 destination). Bit 0 = 1 identifies the class. Additional fields include function (bits 5-7), shift (bits 8-9: 00 none, 01 rotate left 1, 10 rotate right 1, 11 swap bytes), carry control (bits 10-11: 00 ignore, 01 clear to 0, 10 set to 1, 11 complement), skip control (bits 13-15 encoding 8 conditions like always skip, skip if result zero, skip if carry set), and a no-load bit (bit 12=1 to skip loading destination). Results may affect a 1-bit .
MnemonicFunction (Bits 5-7)Behavior
ADD000Destination += source (+ carry if set); supports signed/unsigned.
SUB001Destination -= source (+ carry if set).
AND010Destination = destination AND source.
IOR011Destination = destination OR source (inclusive).
XOR100Destination = destination XOR source.
COM101Destination = ~source (complement).
NEG110Destination = -source ( negate).
MOV111Destination = source (move).
INC011 (variant)Destination += 1 (increment).
CLR000 (with clear)Destination = 0 (clear).
Shifts and skips enhance conditional execution and manipulation efficiency. These execute in one , emphasizing accumulator-centric operations.

Special and Control Instructions

Special instructions are encoded as I/O instructions to the CPU ( 77) or operate variants, handling control without reference. They include no-operation (, operate with AC0 to AC0 and no ), halt (HLT: DOC 00 to 77, stopping execution), and shifts like SHL/SHR (via rotate controls in operate). Interrupt enable/disable (/INTDS: NIO S/C to 77), mask set (MSKO: to 77), and acknowledge (INTA: DIB from 77, loading to AC) manage the interrupt . Read switches (READS: DIA from 77 to AC) loads front-panel . These ensure basic management within the base set.

System Features

Interrupts and Handling

The Data General Nova featured a 16-level interrupt system designed to handle operations efficiently, with levels numbered 0 through 15. Each level corresponded to a bit in a 16-bit mask , allowing software to enable or disable selectively by setting or clearing the appropriate mask bits using the MSKO . Devices were assigned to specific levels based on their wiring to the I/O bus positions, ensuring prioritized handling where higher levels (lower numerical values) took precedence. Interrupt requests were managed through a daisy-chain on the I/O bus, where devices were connected in series to establish among those at the same level; the device closest to the in the chain gained access first if multiple requests occurred simultaneously. The ION (Interrupt On) and IOFF (Interrupt Off) instructions, encoded as NIO S 0,CPU (octal 060177) and NIO C 0,CPU ( 060277) respectively, controlled the global enable flag, disabling further interrupts upon entry to a handler to prevent nesting until explicitly re-enabled. Upon an occurrence, the automatically saved the (PC) to memory location 0 () and performed an indirect jump through location 1 () to the start of the common routine. For handling across the 16 levels, the common routine dispatched to level-specific handlers using software pointers in page zero, based on the interrupting level determined from the device or mask. The handler then performed a software context save, storing the accumulator registers (AC0-AC3), , and other state to a designated area, as the hardware provided no automatic stacking beyond the PC; restoration occurred symmetrically before returning via a JMP to the saved PC. In the SuperNOVA model, () operations integrated seamlessly with the system, using dedicated data channels that could request upon transfer completion or error, maintaining priority over standard program while adhering to the same and daisy-chain protocols. Additionally, a non-able power-fail was provided via an optional power monitor circuit, which detected impending power loss and forced an to allow software to perform emergency saves during the brief delay before shutdown, bypassing the standard and daisy-chain. This design, while efficient for its era, imposed limitations such as reliance on software for complete context management in multilevel scenarios, potentially increasing handler in complex environments.

Front Panel and Diagnostics

The front panel of early Data General Nova models, such as the Nova 1200, provided a direct physical interface for manual system operation and troubleshooting, featuring a row of 16 LEDs (or incandescent lamps in early models) that displayed memory or contents in format for visual inspection of 16-bit words. Complementing this were 16 toggle sense switches positioned below the grid, used to set specific addresses or data values for entry into the system. These elements allowed operators to interact with the machine at a low level, examining and modifying internal states without external peripherals. Key controls on the panel included push-button switches for EXAM/DEP (examine/deposit), which enabled reading the contents of a selected location or onto the and writing values from the switches into , respectively; RUN/STEP for initiating continuous execution or single-stepping through instructions; to halt the and clear pending operations; and an enable button to arm the system for external signals. Accumulator selector switches (for AC0 through ) worked in conjunction with the deposit and examine functions, directing switch inputs to specific registers or displaying their values on the LEDs. For instance, activating the AC0 deposit switch loaded the sense switch pattern directly into accumulator 0, while the examine counterpart transferred the accumulator's contents to the display. Diagnostics were integrated into the through dedicated error lights signaling conditions like parity failures, carry overflow, or direct channel activity, providing immediate visual feedback on hardware issues. Built-in exercisers for and CPU testing could be invoked via controls, cycling through read/write patterns to detect faults in core or processor logic without requiring software intervention. User interaction extended to bootstrap loading, where operators manually entered a short loader routine using the sense switches and deposit functions to initialize the system from paper tape readers or ROM-based devices, establishing the initial before full operation. Subsequent models simplified the front panel design to reduce complexity and cost. The , for example, employed a more compact layout with a single row of 16 data LEDs, status indicators for run mode, memory power, errors, and conditions, and streamlined toggle switches for essential examine/deposit and step operations, while retaining compatibility with core functions. In the embedded microNOVA variant, the panel became optional, often replaced by a compact hand-held unit featuring only power/reset/run switches, basic indicator lights, and minimal capabilities to suit space-limited applications.

Performance Characteristics

The Data General Nova series exhibited progressive improvements in performance across its models, primarily through reductions in memory cycle times and enhancements in instruction execution efficiency. The original Nova, introduced in 1969, featured a core memory cycle time of 2.6 microseconds per word, enabling basic fixed-point arithmetic operations such as addition in approximately 5.9 microseconds. Subsequent iterations accelerated this: the Nova 1200 series reduced the cycle time to 1.2 microseconds, the Nova 2 series to 1.0 microsecond, and the Nova 800 and Supernova models to 0.8 microseconds for core memory, with bipolar semiconductor options in the Supernova achieving 0.3 microseconds. By the Nova 4, released in 1978, the cycle time reached 400 nanoseconds with four-way memory interleaving, allowing typical instruction executions like stores in 400 nanoseconds and adds in 200 nanoseconds. These advancements translated to effective throughput rates approaching 0.5 million (MIPS) for basic operations in later models, though actual performance varied with memory type and workload. Benchmark results underscored the Nova's capabilities relative to its era. In Whetstone benchmarks, which measure floating-point performance, the original Nova achieved around 0.016 million Whetstone instructions per second (MWIPS) using software-implemented floating-point arithmetic, scaling to 0.068 MWIPS with hardware support in models like the Nova 840. The Nova 3/12 variant performed at 0.014 MWIPS in Fortran single precision, while later Eclipse derivatives (successors to the Nova line) reached 0.45 MWIPS with hardware floating-point, representing roughly a 4x improvement over the original Nova. Direct memory access (DMA) throughput also improved, from 285,500 words per second input in the Nova 800 to 1.25 million words per second in Supernova models with high-speed options. Compared to contemporaries, the Nova series provided superior integer mathematics performance over the DEC PDP-8, benefiting from its 16-bit architecture versus the PDP-8's 12-bit design, which enabled roughly 2-3 times faster processing for equivalent tasks at similar cycle times. It remained cost-competitive with the DEC PDP-11, offering about 40% lower pricing for comparable configurations while delivering equivalent or better price/performance in applications. Key performance factors included initial memory access latencies of 800 nanoseconds in early core-based systems, reduced to 700 nanoseconds in Nova 3 , which minimized bottlenecks in compute-intensive workloads. Later models like the incorporated overlapped fetch-execute cycles, further boosting effective speed by up to 50% over predecessors. Power consumption typically ranged from 175 to 725 watts across configurations, averaging around 500 watts for a fully equipped rack-mounted system, which supported efficient deployment in and industrial environments with standard single-phase power. This heat dissipation, while manageable via rack , highlighted the trade-offs of TTL-based logic in minicomputers.
ModelMemory Cycle TimeExample Add Time (Fixed-Point)Relative Performance (vs. Original Nova)
Original Nova2.6 μs5.9 μs1.0
Nova 12001.2 μsN/A2.5
Nova 21.0 μs1.0 μs4.5
Supernova (Core)0.8 μs1.6 μs7.0
Nova 40.4 μs0.2 μs~10x (est.)

Software and Applications

Operating Systems and Software Ecosystem

The primary operating system for the Data General Nova was RDOS (Real-Time Disk Operating System), released in 1970 as a multitasking environment supporting up to 32 processes and up to 64K words of . RDOS provided file capabilities for disk drives, including support for multiple controllers and up to four drives per controller, enabling efficient data storage and retrieval in multi-user configurations. It facilitated program development through integration with compilers and utilities, emphasizing low-latency task switching suitable for industrial and scientific applications. Data General developed several programming languages for the Nova, including Extended BASIC, an interpreter based on the Dartmouth BASIC standard with extensions for graphics output and direct I/O operations to peripherals like terminals and plotters. and compilers were available, supporting scientific with features like and compatibility with RDOS file systems for . implementations, such as Interactive COBOL, catered to business applications with indexed sequential access method (ISAM) support for database management. The assembler, featuring macro capabilities and a syntax using directives like .ZREL for page-zero relocation, allowed low-level programming with symbolic tools. Utilities for the Nova ecosystem included text editors for source code management, symbolic debuggers for runtime analysis, and cross-compilers running on host systems like larger Data General machines to generate Nova binaries. A library of supporting programs encompassed routines, data conversion tools, and communications packages for network integration, enhancing productivity in development environments. Third-party efforts have preserved the Nova software ecosystem through emulators, such as the Wild Hare Nova and Eclipse emulator released in 2022, which supports running RDOS and legacy applications on modern Windows and platforms under a hobbyist license. The open-source simulator also emulates the Nova, allowing execution of original software distributions for research and restoration. Preservation sites like Bitsavers.org host scanned manuals and binaries, addressing gaps in historical documentation related to early software ports and bugs by providing access to original artifacts.

Programming Examples

Programming examples for the Data General Nova typically involve low-level to demonstrate core instructions for I/O, arithmetic, and logical operations. These snippets highlight the use of the machine's four accumulators (AC0-AC3), memory reference instructions like LDA and JSR, I/O transfers via device codes, and (ALU) operations. The Nova's assembly format uses representations for , with directives like .OCT for constants and comments for clarity. All examples assume a basic configuration without optional hardware extensions like multiply/divide units. A simple output program, akin to a "," uses I/O instructions to send characters to the teletype (TTO device code 10 ). This involves loading message bytes into an accumulator, checking device readiness with SKPBZ (skip if busy zero), and outputting with (data out to buffer A), often with wait loops for the slow peripheral. The message can be stored as packed ASCII (two characters per word), unpacked via shifts or masks, and output byte-by-byte until a terminator. For 16-bit unsigned multiplication without hardware support, a software routine implements the shift-and-add (or Booth) algorithm. It uses separate accumulators for multiplicand, multiplier, partial product, and loop counter. The routine initializes the product to zero, then for each bit of the multiplier (starting from LSB), conditionally adds the multiplicand to the product if the bit is set, shifts the multiplier right, and repeats for 16 bits. This leverages ADD, SHR (shift right), and conditional jumps like JCN (jump on condition, e.g., zero or carry). The 17-bit ALU handles carries, but the result is truncated to 16 bits. To print the accumulator contents in , a routine copies the value to a working accumulator, then iteratively shifts right to isolate each bit (starting from MSB by using LSH or equivalent for positioning), tests the low bit with AND #1 or carry after SHR, conditionally loads and outputs '0' ( 060) or '1' (061) via an I/O subroutine, and loops 16 times using ISZ for counting. The output uses device skips and data out instructions, with a carriage return terminator. Immediate constants are 8-bit limited, so larger values like 16 () are loaded via LDA #20 ().

Notable Deployments

The Data General Nova found significant application in scientific laboratories for data acquisition and analysis tasks. For instance, NASA employed Nova systems in prototype configurations for processing earth resources data, such as in the Earth Resources Data Analysis System (ERDAS) project, where a Nova 3/12 minicomputer with 32K words of core memory handled image and sensor data integration. Similarly, Nova computers supported remote sensing and computational tasks in NASA facilities, leveraging their compact design and I/O capabilities for real-time data handling. In industrial settings, the Nova was deployed for process control in and sectors, benefiting from its robust I/O interfaces for integration. One early example involved a Nova system intended for Oil's operations in , to manage data processing in oil-related applications, highlighting its suitability for harsh industrial environments. The system's modular enabled custom expansions for analog-to-digital conversion and control loops in refineries and production lines. Educational institutions adopted the Nova for its affordability and accessibility in curricula, particularly in teaching programming and operating system fundamentals. Universities equipped departmental computing centers with Nova systems to support hands-on instruction in and graphics, making it a staple in mid-1970s academic environments. For military and embedded applications, the microNOVA variant was integrated into and under 1970s defense contracts. A notable deployment occurred in the U.S. Air Force's Air Intercept Controller , where the microNOVA processed inputs from trackballs and displays for tactical and . Key OEM integrations showcased the Nova's versatility in specialized fields. In , EMI's pioneering scanners from 1973 utilized the 820 with 32K memory for image reconstruction, enabling scan times of about 7 minutes per slice and revolutionizing diagnostic capabilities. This integration demonstrated the Nova's role in high-impact, computational tasks beyond general computing.

References

  1. [1]
    1968 | Timeline of Computer History
    Started by a group of engineers that left Digital Equipment Corporation (DEC), Data General designs the Nova minicomputer. It had 32 KB of memory and sold for ...
  2. [2]
    Data General Nova minicomputer - CHM Revolution
    Data General's Nova was one of the first minicomputers to use MSI (Medium Scale Integration) TTL logic circuits, which had 10 to 100 gates on each IC.
  3. [3]
    [PDF] The best small computer in the world, 1968
    The NOVA 110facility consists of 16bi-directional data lines, 6 lines for device selection, 19 unary control lines from the central processor, and 6 control ...
  4. [4]
    Data General Corporation (DG) | Selling the Computer Revolution
    In 1969 they released the Nova minicomputer and it quickly became a popular machine in scientific and educational markets. Part of its appeal was its cycle ...
  5. [5]
    [PDF] Data General Nova Series
    Nov 20, 1974 · The Nova family of minicomputers is generally used in control/monitoring systems, industrial testing, data acquisition/analysis, and various ...
  6. [6]
    [PDF] Oral History of Edson (Ed) de Castro; 2002-11-22; 2003-01-13
    Jan 13, 2003 · My brother originally started out at Northeastern, in a general science kind of background, but after his first year, his freshman year, decided ...
  7. [7]
    History of Data General Corporation – FundingUniverse
    To start the company Adler put up $50,000 of his own money and raised much of the rest of the founding capital, which totaled $800,000. With that Data General ...Missing: initial | Show results with:initial
  8. [8]
    The Minicomputer -- 1959-1979
    De Castro, promoted to design leader, presented the PDP-X design to management in late 1967. The recommendation echoed many of the same themes as had the ...
  9. [9]
    DG-NOVA
    Data General Nova: To historical notes, it should be noted the architecture of the system is very similar in many respects to the PDP-11, and accusations have ...
  10. [10]
    Data General Nova, serial #1 - CHM Revolution
    Cost: $9,200. This first completed Nova was shipped to Unitech in Austin, TX to be used by Mobil Oil. It never arrived. Braniff Airlines lost it. The computer ...Missing: $4000 units
  11. [11]
    If it moves, it should be Ruggednova - Computer History Museum
    Dec 4, 2012 · De Castro left DEC in frustration and, together with former colleagues, started his new company, Data General, in 1968. Only a few months later, ...
  12. [12]
    None
    ### Evolution of Nova Models
  13. [13]
    [PDF] Super Nova, 1970
    A full memory cycle using core memory takes 800 nanoseconds for the Supernova. Using read-only memory, cycle time is 300 nanoseconds. Supernova overlaps the ...
  14. [14]
    None
    ### Summary of Data General Nova Series (Extracted Details)
  15. [15]
  16. [16]
    [PDF] Data General Nova 4 Series (Preliminary Report) - Bitsavers.org
    A full report on the Nova 4 will be published when full documentation and pricing is released. MARCH 1979. (~) 1979 DATAPRO RESEARCH CORPORATION, DELRAN, NJ ...
  17. [17]
    [PDF] Data General Nova 4 Series - Bitsavers.org
    The Nova 4 series of computers consists of the compact, single board Nova 4/C; the higher-performance Nova 4/S; and the ex- tended-memory Nova 4/X computer.
  18. [18]
    When a Minicomputer becomes a Micro: the DGC microNOVA ...
    Nov 21, 2014 · In 1979 DGC released the mN602, based on the NOVA 4. The power supplies were reduced from 4 to 3, and memory support was now a full 128K using ...
  19. [19]
    [PDF] microNOVA INTEGRATED CIRCUITS DATA MANUAL - Bitsavers.org
    ... Data General's NOVA and. ECLIPSE line computer systems. Each time the CPU fetches an 110 instruction from memory, It transmits the complete instruction over ...
  20. [20]
    DG Eagle MV/8000 / The Soul of a New Machine - Clemson University
    The machine as a four-stage, underpipelined design with 110 nsecs cycle time - (1) fetch instruction, (2) decode, (3) fetch microinstruction, and (4) execute ...Missing: original μs<|separator|>
  21. [21]
    Building the Data General Eclipse MV/8000, or, lessons from ...
    Dec 16, 2014 · The firm's survival was on the line. They were being slaughtered by Digital Equipment Corporation's VAX 11/780 and they needed a winner, badly.Missing: introduction date competition
  22. [22]
    Rise and Fall of Minicomputers
    Oct 24, 2019 · Data General's 1969 16-bit Nova was noteworthy because its clever design put the processor on a single, large, printed circuit board, ...
  23. [23]
    Minicomputers and The Soul of a New Machine - Opensource.com
    Feb 16, 2020 · The Soul of a New Machine was written in 1981 by Tracy Kidder, and chronicles a small group of engineers at the now-former tech company, Data ...<|separator|>
  24. [24]
    Arctic Adventures With A Data General Nova II — The Equipment
    Feb 5, 2024 · The Nova family of computers had 16-bit words, and the front panel had a row of 16 toggle switches. Using these switches, you could store any ...
  25. [25]
    Using a 50 Year Old Computer - Data General microNOVA is Done
    May 24, 2025 · My Data General microNOVA is now fully repaired and ready to use! So lets get it back in the rack, set up an operating system, and try to ...Missing: mN601 mN602 1977
  26. [26]
    About - Data General Nova, Eclipse, MV and AV
    This site is about restoring Data General systems, a hobby for the author, and a resource for others interested in bringing old systems back to life.
  27. [27]
    Data General Nova - Informatics Engineering | Wiki eduNitas.com
    The Data General Nova was a popular 16-bit minicomputer built by the American company Data General starting in 1969. The Nova was packaged into a single ...Missing: logic | Show results with:logic
  28. [28]
    'Nova' minicomputer designed and produced by the Data General ...
    ... Nova sold 50,000 units and made Data General a highly successful company. The Nova reputedly influenced the design of the Xerox Alto and Apple I. Its ...Missing: total sales
  29. [29]
    Data General Nova Class Instruction Set Description
    Sep 13, 1998 · The Nova architecture defines four 16 bit Accumulators ... load an accumulator from memory and store an accumulator to memory respectively.Missing: TTL | Show results with:TTL
  30. [30]
    [PDF] Data General Nova Series
    Feb 9, 1977 · CYCLE TIME: Please refer to the following table. FEBRUARY 1977 ... Data General Nova Series !:> When the original Nova minicomputer-forerunner of ...
  31. [31]
    Nova - Chessprogramming wiki
    The Nova was a 16-bit minicomputer series with a load/store architecture, a 4-bit ALU, and a 15-bit address space. It was packaged in a single rack mount case.<|control11|><|separator|>
  32. [32]
    [PDF] Data General Nova 4 Series - Bitsavers.org
    Introduced in December 1978, the new models include the compact Nova 4/ C, featuring a complete 64K-byte computer on a single circuit board; the standard.<|separator|>
  33. [33]
    None
    Summary of each segment:
  34. [34]
    [PDF] data general nova system - Bitsavers.org
    DATA GENERAL NOVA SYSTEM. NOVA 3. Page 2. Programmer's Reference Series. NOVAR Line Computers. Data General. Data General Corporation, Westboro, Massachusetts ...Missing: handbook | Show results with:handbook
  35. [35]
    None
    Summary of each segment:
  36. [36]
    None
    ### Summary of Interrupts and Related Mechanisms in Data General Nova Systems
  37. [37]
    [PDF] HOW TO USE THE NOVA COMPUTERS - Bitsavers.org
    Load the digital data word from accumulator AC into the D I A converter. Conversion starts automatically and immediately after data has been loaded. DOB ...
  38. [38]
  39. [39]
    Data General Nova 1200 | Vintage Computer Federation Forums
    Jul 23, 2019 · A Data General Nova 1200, populated with a CPU board, 4x 8KW core memory boards (for a full 32KW on board), and a CompuScan OEM I/O board.Data General Nova 1200 - Vintage Computer Federation ForumsData General Nova 1200 | Page 5More results from forum.vcfed.orgMissing: 800 | Show results with:800
  40. [40]
    [PDF] DATA GENERAL CORPORATION - Bitsavers.org
    This manual is intended to aid personnel in the maintenance of the Nova. Included are descriptions of the control logic for the teletype, reader and punch, but ...Missing: front | Show results with:front
  41. [41]
    NOVA 3 - pdp-11.nl
    The NOVA 3 officially can support 128 KW (256 KB) according to Data General, but can really handle up to 256 KW (512KB) with a single-word patch to the RDOS ...Missing: ns cache<|separator|>
  42. [42]
  43. [43]
    None
    ### Performance Characteristics for Nova 4
  44. [44]
    (PDF) Whetstone Benchmark History and Results - ResearchGate
    Jul 28, 2017 · PDF | Several hundred results, from more than 50 computer manufacturers, for systems produced between the 1960s and 1990s, plus PCs to 2014.
  45. [45]
    Computer History | They Create Worlds
    Dec 11, 2014 · The success of the PDP-8 opened up a whole new market for small, cheap machines that soon gained the designation “minicomputers.” With IBM and ...<|separator|>
  46. [46]
    [PDF] Nova 2 Computer Nova 2110
    Data General's Nova 2 combines reliable Nova-line design con- cepts with new ... Address Modes: Direct addressing of 1024 words absolute, relative and ...Missing: manual | Show results with:manual
  47. [47]
    RDOS - ArchiveOS
    Aug 30, 2024 · RDOS (Real-time Disk Operating System) – a real-time operating system released in 1972 for the 16-bit Nova and Eclipse minicomputers from Data General.
  48. [48]
    [PDF] (RDOS) - Reference Manual - Bitsavers.org
    Your RDOS system can support two controllers for every type of disk drive Data General provides. Each controller (except for 6001-6008) can control up to four.
  49. [49]
    [PDF] RDOS System Reference
    This is the primary reference manual for Data General's. Real-Time Disk Operating System, RDOS. It describes all features of the operating system for the NOVA ...
  50. [50]
    [PDF] EXTENDED BASIC - User's Manual
    Data General's Extended BASIC is an implementation of the BASIC language as developed at Dartmouth. College. Extended BASIC includes such features as ...
  51. [51]
    [PDF] Data General Nova 3 Series
    Sep 15, 1979 · DATE ANNOUNCED: Nova 3/4 and Nova 3/12, October. 1975; Nova 3/D, August 1976. ... CYCLE TIME: Please refer to the following table. Module size, ...Missing: cache | Show results with:cache
  52. [52]
  53. [53]
    [PDF] How To Use The Nova - Bitsavers.org
    The basic processor cycle time of the Nova computer is 2.6 microseconds with core memory, 2.4 micro- ... Nova line computers, Data General has available a general.Missing: original | Show results with:original
  54. [54]
    [PDF] Data General Nova 4 Series
    Arithmetic and Logic instructions use bits I and 2 to identify an accumulator containing. a second operand (if present), bits 5-7 to specify primary function, ...Missing: discrete | Show results with:discrete
  55. [55]
    Nova and Eclipse emulator beta release - Simh - Groups.io
    Nov 28, 2022 · Wild Hare Computer Systems is pleased to announce the public beta release of its Data General Nova and Eclipse emulator.Missing: Brain | Show results with:Brain
  56. [56]
  57. [57]
  58. [58]
    [PDF] 19790023556.pdf - NASA Technical Reports Server (NTRS)
    The prototype system consists of a Data General Nova 3/12 minicomputer with 32 kilo words of core memory, a Data General Dual. Floppy Disk with a storage ...
  59. [59]
    [PDF] General Disclaimer One or more of the Following Statements may ...
    Data General NOVA 2, CDC Cyber 74, SEL 32/55. SPECIAL PERIPHERALS: 2 Tape Drives. Disk Files. PFRPOSE: RECCL uses a first order transformation matrix computed ...
  60. [60]
    Thwarted at DEC, Thriving at Data General - CHM Revolution
    When Data General's founders created the company, de Castro was 29 years old and Burkhardt was only 23.Missing: Corporation | Show results with:Corporation
  61. [61]
    Data General Nova 2 - System Source Computer Museum
    It is equipped with two harddisk drives, one twin floppy drive (8" disks!), one teletype, one high-speed paper tape punch reader and one punch card reader (not ...
  62. [62]
    [PDF] Functional Design for Air Intercept Controller Prototype ... - DTIC
    Micronova Input Processing. This function processes TEC inputs (buttons, track ball, etc.) that have been detected by the Micronova. It interacts heavily ...
  63. [63]
    How CT happened: the early development of medical computed ...
    The proposal initially received £2500 DHSS funding on October 28, 1968 ... Data General introduced the first 16-bit Nova in 1968 at $3999, one-fifth ...<|control11|><|separator|>
  64. [64]
    52 year old computer performs fast CT reconstruction | Hackaday.io
    Feb 15, 2025 · The clinical system, released in 1973 contains a Data General Nova computer. I personally first encountered a Data General Nova computer ...Missing: case studies OEM telecom