A Low-Cost, Space-Efficient PDP-10 Technical Proposal

This document proposes the development of a low-cost, space-efficient PDP-10 processor, referred to as the KD10, aimed at existing TOPS-10/TOPS-20 customers transitioning to a distributed computing environment. The primary goals include a time-to-market of 18 months or less, minimal software changes, low manufacturing cost (target $15,000-$20,000), attractive and space-efficient packaging, and high reliability, all while leveraging existing designs and low-risk technology. High CPU performance is not a primary objective.

The KD10 system is a single-CPU PDP-10 processor designed to support 20-32 users, featuring 30-bit extended addressing. It is an upgraded version of the KS10 (2020) design, utilizing AMD2901 family and Schottky TTL MSI parts, and fits on two extended hex modules. The memory subsystem supports up to 4 Mwords using 256K MOS RAM parts with ECC. Initial I/O uses a Unibus Adapter (UBA) and UDA50 disk controller, similar to the KS10, with future plans for CI/NI support. Packaging options include an 11/750-size or RA81 cabinet, integrating the processor and disk drives for space efficiency.

Development involves incremental changes to existing KS10 hardware designs, microcode, and software. Key hardware modifications include adding extended addressing to the KS10 processor, upgrading memory modules to support 4 MWords with 256K RAM, and adapting the console and UBA designs. Simulation will be extensive, covering register transfer, gate-level, and timing verification. Microcode will be adapted from the KS10, with changes for extended addressing and performance optimization. Software and diagnostic efforts focus on modifying existing code and developing new drivers for the KD10's specific hardware interfaces.

Manpower estimates project four engineers for 12-15 months for hardware, microcode, and simulation, with additional resources for software, diagnostics, CAD, and module layout. Computer resources like KL10 and VAX systems will be needed for development and verification. Performance estimates suggest the KD10 could halve the KS10 cycle time (from 300ns to 150ns), potentially achieving 0.3-0.5 times the performance of a KL10. The document also includes detailed manufacturing cost estimates for various components over three fiscal years.