IBM 3081
| Company type | Mainframe computer series |
|---|---|
| Industry | Computer hardware |
| Predecessor | IBM 3033 |
| Founded | November 12, 1980 |
| Defunct | August 4, 1987 |
| Headquarters | Armonk , United States |
| Products | Mainframe computers (3081, 3083, 3084) |
| Parent | IBM |
IBM 308X Mainframe Series
The IBM 308X line consists of mainframe computers built on the IBM System/370 architecture, later incorporating the System/370 Extended Architecture (S/370-XA) starting in 1983. This architectural enhancement expanded addressing capability from 24 bits to 31 bits, allowing for access to larger memory spaces. The S/370-XA platform was compatible with IBM's MVS/SP V2 operating system and introduced the Start Interpretive Execution (SIE) instruction, which enabled virtualization through the Virtual Machine/eXtended Architecture Migration Aid.
The first released unit, the Model 3081 Processor Complex, debuted on November 12, 1980, consisting of a 3081 Processor Unit alongside its supporting hardware. Subsequently, IBM introduced the 3083 on March 31, 1982, followed by the 3084 on September 3, 1982. IBM promoted these three models collectively as "System/370-Compatibles." Production of the series ceased on August 4, 1987.
These systems marked a significant step forward in IBM's mainframe development, enhancing processing capabilities and memory addressing to meet enterprise demands for extensive data processing and virtualization support.
IBM 3081
The initial 3081 variant, known as the 3081D, delivered a performance of 5 MIPS. The subsequent 3081K model increased throughput to 7 MIPS. On September 3, 1982, the 3081G was introduced alongside the 3084, replacing the original 3081D while maintaining a 5 MIPS rating.
Within the 308X family, the 3081 provided scalable performance options positioned between the entry-level 3083 and the more powerful 3084 models. It incorporated a dual-processor configuration within a single chassis; however, it was not designed to be partitioned into separate independent systems.
The physical dimensions and components of all 3081 and 3083 models were consistent: each system included the CPU unit (either a 3081 or a 3083), a service processor equipped with channels (the 3082), and a water pump/coolant assembly (the 3087). These systems operated on a power frequency of 400 Hz, which could be supplied either by an optional motor/generator unit (model 3089) or from an external power source.
Successor to IBM 3033
In comparison with its predecessor, the IBM 3033, the IBM 3081 incorporated several technological improvements:
- Utilization of approximately 800,000 circuits implemented through large scale integration, with chips containing up to 704 logic circuits to boost performance and reliability
- Introduction of the Thermal Conduction Module (TCM), which eliminated one packaging layer
- Adoption of water cooling technology for enhanced heat dissipation surpassing traditional air cooling methods
- Machine cycle time reduced to 26 nanoseconds (equivalent to a clock speed of about 38 MHz)
- Lower power consumption—23 kilowatts for a 3081-D16 compared to the previous model's 68 kilowatts for a 3033-U16
- Nearly double the instruction execution speed
- Dual central processors sharing access to as many as 24 channels and up to 32 megabytes of main memory
The Thermal Conduction Module was designed as a flat ceramic package housing approximately 30,000 logic circuits distributed over up to 118 custom TTL chips mounted face-down using flip-chip technology. It contained thirty-three metalized layers that routed signals and power through around eighteen hundred pins. Each module was sealed with a helium-filled metallic cap featuring pistons that pressed against each chip to efficiently transfer heat to a water-cooled cold plate maintained at roughly 24 °C. This engineering allowed heat flux densities approximately ten times greater than those achieved by the IBM 3033.
Internally, IBM referred to the IBM 3081 by its codename Adirondack.
IBM 3083
The IBM 3083 was originally conceived without plans for actual production, aimed specifically at the ACP/TPF segment, which demanded a rapid uniprocessor machine.
Within the 3083 lineup, the CX model exhibits the slowest instruction throughput. Next in performance ranking are the E and EX variants, followed by B and BX, with the J and JX models delivering the highest speeds.
According to IBM's performance data:
- The CX model operates at roughly 75% of the instruction rate of the EX model.
- The B and BX models run approximately 1.4 to 1.5 times faster than the E and EX models, respectively.
- The J and JX models achieve speeds ranging from 1.8 to 2.0 times those of the E and EX counterparts.
In summary, the peak performance of the fastest 3083 is nearly 2.67 times greater than that of the slowest variant.
IBM 3084
Introduced in September 1982 and discontinued by August 1987, the IBM 3084 came with main memory options of 32, 48, or 64 million bytes; subsequent versions (3084X) expanded capacity up to 128 MB. The top-tier 3084QX model reached a processing speed of 31 MIPS.
Essentially formed by pairing two IBM 3081 systems, it functioned as a four-processor symmetric multiprocessing (SMP) configuration capable of operating either as two separate dual-processor units or as a unified four-processor system.
Larger in physical size than previous models, it incorporated dual CPUs, two cache modules (3082), and a pair of floating-point units (3087). Systems utilizing optional 3089 processors required two such units.
The architecture supported two operational modes:
- PP-mode (Partial Processing): where it ran as two distinct systems.
- SI-mode (Single Image): where all four processors worked cohesively as one system.
See also
Further reading
- — Chapter 9 (pp. 241–254) discusses the details of the 308X series.
