Real Processing in Phase Change Memory

Informazioni relative al progetto

Real-PCM-PIM

ID dell’accordo di sovvenzione: 101069336

DOI

10.3030/101069336

Progetto chiuso

Data della firma CE 22 Marzo 2022

Data di avvio 1 Aprile 2022

Data di completamento 30 Settembre 2023

Finanziato da

• European Research Council (ERC)

Costo totale

Nessun dato

Contributo UE

€ 150 000,00

Coordinato da

TECHNION - ISRAEL INSTITUTE OF TECHNOLOGY
Israel

Periodic Reporting for period 1 - Real-PCM-PIM (Real Processing in Phase Change Memory)

Periodo di rendicontazione: 2022-04-01 al 2023-09-30

Sintesi del contesto e degli obiettivi generali del progetto

This PoC aims to enable the development of a new type of processor, the mMPU that will be based on commercial memory technologies. The mMPU system improves the performance and energy of computing systems by two to three orders of magnitude compared to von Neumann architecture as we and other groups have demonstrated for different data intensive applications such as image processing, DNA sequencing, databases, and deep neural networks. We demonstrated the implementation of an mMPU using a memristive technology that is already on the market (i.e. PCM). PCM is considered the state-of-the-art technology for solid-state drives (SSD) and persistent memory. The results of this PoC project enable, for the first time, computation within such products (e.g. Intel's Optane) with minimal modifications and will accelerate the commercialization of the mMPU. The results will also demonstrate the superiority of the mMPU over standard state-of-the-art processing products (e.g. GPU, CPU, ASIC), which are separated from the memory and suffer from the memory and power walls.
RRAM-based stateful logic was proposed and demonstrated in several techniques, in addition to MAGIC. On the other hand, no PCM-based logic has been proposed to date (not to mention demonstrated in experimental work and/or integrated within memory). The unique ideas behind this PoC are based on our expertise and knowledge acquired from using RRAM-based logic. The PoC opens an alternate path for mMPU design and makes it less dependent on a specific memristive technology (note that RRAM is still being explored in numerous companies as an attractive future persistent memory device).

Lavoro eseguito dall’inizio del progetto fino alla fine del periodo coperto dalla relazione e principali risultati finora ottenuti

In this research, we have made significant advancements in the field of Phase Change Memory (PCM) technology, particularly in the development of PCM stateful logic gates. Our team has successfully designed and analyzed various logic gates that are not only based on PCM devices but are also fully compatible with the PCM crossbar array structure. This compatibility is a crucial step forward, as it ensures that our designs can be integrated into existing PCM systems, thereby enhancing their applicability in real-world scenarios.
To validate our designs, we conducted comprehensive evaluations of these newly developed logic gates. This involved a series of meticulous tests and demonstrations on a discrete PCM setup. Each logic gate in our study was carefully constructed using two to four PCM devices. The experimental setup was designed to mimic real-world conditions as closely as possible, ensuring the reliability and validity of our results. The testing process was carried out using a sophisticated probe station, which allowed us to connect and measure the performance of the PCM devices with high precision. This meticulous approach to testing has been instrumental in verifying the efficacy of our logic gates.
Additionally, our research has been bolstered by proactive engagement with the industry. We have initiated and maintained productive discussions with potential users and customers, providing them with insights into the practical applications and benefits of our PCM-based logic gates. These conversations have not only helped in understanding the market needs but have also been vital in refining our designs to meet real-world requirements.
Furthermore, we have reached out to PCM manufacturers, fostering a collaborative environment that bridges the gap between academic research and industrial application. These interactions with manufacturers are crucial as they provide valuable feedback and help in aligning our research objectives with industry standards and expectations.
In summary, the main achievements of our research include the successful development and comprehensive evaluation of PCM stateful logic gates, along with establishing meaningful collaborations with industry stakeholders. These accomplishments mark a significant contribution to the field of PCM technology, opening up new possibilities for its application and advancement.

Progressi oltre lo stato dell’arte e potenziale impatto previsto (incluso l’impatto socioeconomico e le implicazioni sociali più ampie del progetto fino ad ora)

Our research has notably advanced PCM technology by developing stateful logic gates that integrate seamlessly with existing PCM crossbar array structures. This breakthrough enhances real-world applicability, bridging the gap between theoretical design and practical use. Through rigorous testing and industry collaboration, our work not only validates the effectiveness of these gates but also paves the way for innovative applications in memory and computing technology, potentially revolutionizing the PCM landscape. The impact of the mMPU will be seen in different applications such as genomics, artificial intelligence, image processing, and databases.
We still need to experimentally demonstrate and validate the gates within a large memory crossbar array and based on that, in collaboration with an industrial partner, integrate the gates in a commercial manufacturing process.