Processor In Memory: Return of the Transputer
The Memristor has recently brought about a lookback at early Processor In Memory attempts, one of which was the Transputer. INMOS, a UK semiconductor company from the early 1980’s, (later the familiar ST Electronics, NYSE:STM) invested heavily in concurrent computing microprocessor design, which, while missing the target on a viable “Plastic Pal who’s Fun to Be With [link]”, laid a lot of the groundwork for conceptually reframing the microprocessor models that have dominated till today.
The main goal of the Transputer model is to reduce memory latency and increase bandwidth, while also reducing the power requirements of a system. Memory stalls in current processor designs are responsible for much of the power consumption due to the complexity required to avoid them.
In the late 1980s, the Transputer was hyped up to be the next greatest design for the future of computing. However, experimental research has continued to this day, most recently centered around Processor in Memory designs reliant upon DRAM substrates. But work has begun on theoretical applications of memristors to overcoming the logic memory complexity Transputers started on long ago.