Kirk D. Lamb

Server-class DDR3 SDRAM memory buffer chip

IBM System i®, System p®, and System z® servers require an efficient ultrareliable high-performance memory subsystem. The fourth-generation IBM advanced memory buffer (AMB) chip provides industry-leading performance, scalability, and reliability for the double-data-rate 3 (DDR3) synchronous dynamic random access memory (SDRAM) subsystems employed across a wide range of server platforms. The new IBM AMB employs a cyclic redundancy code-protected packet-protocol-based 6.4-Gb/s host channel, as well as dual 9-byte/10-byte wide 800 to 1,333-Mb/s SDRAM interfaces with dynamic calibration for optimal signal integrity under varied device and system environmental conditions. Applications support industry-standard dual inline memory module (DIMM) and low-latency high-capacity proprietary DIMM packages in conventional multichannel and redundant array of independent memory system architectures. A fully configured daisy-chain topology contains up to 256 GB of memory per host channel. This paper describes the IBM AMB chip architecture, design, and key engineering aspects.

Use of binary decision diagrams in the modelling and synthesis of binary multipliers

Previous research has proved that the size of binary decision diagrams (BDD) which model the outputs of a binary multiplier grows exponentially with the number of multiplier inputs. This paper presents a method of partitioning the multiplier in a manner which restricts the complexity to the carry bits, which results in BDDs which grow no faster than the square of the number of inputs. These carry functions are themselves interesting, in that their computation amounts to computing subsets of power sets. The concept of largest possible BDD is introduced.

Event monitoring in highly complex hardware systems

This paper presents a novel approach for monitoring disjunct, concurrent operations in heavily queued systems. A nonobtrusive activity monitor is used as an on-chip tracing unit. For each pending operation the monitor uses the hardware implementation of an event-triggered operation graph to trace the path of the operation through the system. In contrast to conventional tracing units, which collect and record information from one or more functional units for later analysis, the presented solution directly records the path taken by the operation through the system, making possible an immediate analysis of operation inconsistencies. For each followed path a unique signature is generated which significantly reduces the amount of trace data to be stored. The trace information is stored together with a time stamp for debugging and measuring of queueing effects and timing behavior in the system. The method presented has been successfully applied to the memorybus adapter chips in the S/390® G5 and G6 systems.