{"title":"Performance Enhancement of Edge-AI-Inference Using Commodity MRAM: IoT Case Study","authors":"M. Suri, Abhishek Gupta, V. Parmar, Kangseol Lee","doi":"10.1109/IMW.2019.8739709","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739709","url":null,"abstract":"In this paper we demonstrate how performance of Edge-AI-Inference hardware can be enhanced by effectively using emerging commodity MRAM chips inside specific accelerator pipelines. We consider the special case of IoT-centric ‘normally-off/low-frequency’ AI inference workloads to benchmark the proposed approach. The proposed Non-Volatile AI Inference Accelerator (NVIA) is realized using FPGA and off-the-shelf MRAM chips. NVIA is benchmarked for Human activity recognition (HAR) dataset. Significant power gains, ∼9x (with Toggle-MRAM (180 nm)) and ∼750x (with STT-MRAM (22 nm)), were achieved compared to volatile SRAM.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115324724","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Quantification of Compositional Runaway during Electroformation in TaOx Resistive Switching Devices","authors":"Yuanzhi Ma, Jonathan M. Goodwill, M. Skowronski","doi":"10.1109/IMW.2019.8739727","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739727","url":null,"abstract":"In this work, we demonstrate that the electroformation of resistive switching devices occurs in two steps where the device first goes through thermal runaway followed by formation of the permanent conductive filament by elemental diffusion. The current transients caused by composition change exhibit slow-fast-slow dependence of rate on time, which is characteristic of a positive feedback runaway process. In this case, the feedback loop is between current in the filament and filament composition. The incubation time for the compositional runaway was measured as a function of device temperature and the activation energy for the process was estimated to be 1.4 eV. Plan-view TEM revealed a metal-rich filament formed in the device during the second runaway. Device resistance post-forming was found to be negatively correlated with forming power as a result of thermal effects.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"146 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128112877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Selector-Free Cross-Point Memory Architecture Based on Ferroelectric MFM Capacitors","authors":"W. Chen","doi":"10.1109/IMW.2019.8739707","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739707","url":null,"abstract":"This paper proposes a selector-free cross-point memory architecture. The memory cell consists of a single ferroelectric capacitor without a select transistor or any other selector device. The selection of a memory bit is accomplished by driving half of the program voltage (VPP/2) on selected wordline and another VPP/2 of opposite polarity on selected bitline, where VPP exceeds the coercive voltage of the ferroelectric capacitor.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114214622","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
W. Chien, L. Gignac, N. Gong, C. Yeh, C.H. Yang, R. Bruce, H. Cheng, J. Papalia, H. Miyazoe, C. Cheng, W. Kim, I. Kuo, F. Carta, A. Ray, E. Lai, M. BrightSky, H. Lung
{"title":"Solution for PCM and OTS Intermixing on Cross-Point Phase Change Memory","authors":"W. Chien, L. Gignac, N. Gong, C. Yeh, C.H. Yang, R. Bruce, H. Cheng, J. Papalia, H. Miyazoe, C. Cheng, W. Kim, I. Kuo, F. Carta, A. Ray, E. Lai, M. BrightSky, H. Lung","doi":"10.1109/IMW.2019.8739648","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739648","url":null,"abstract":"A reliability study for PCM and OTS intermixing was addressed. The buffer layer between PCM and OTS plays a key role in preventing PCM/OTS intermixing after BEOL processing thermal treatment. Besides cycling endurance, performance degradation due to interlayer intermixing was observed. Optimal device operation and an improved buffer layer allowed drastically improved cycling endurance from a few cycles to > 1E9 cycles.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"165 7","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132463697","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
F. L. Rosa, S. Niel, A. Régnier, F. Maugain, M. Mantelli, A. Conte
{"title":"40nm embedded Select in Trench Memory (eSTM) Technology Overview","authors":"F. L. Rosa, S. Niel, A. Régnier, F. Maugain, M. Mantelli, A. Conte","doi":"10.1109/IMW.2019.8739731","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739731","url":null,"abstract":"The eStM (embedded Select in Trench Memory) is a floating gate based non-volatile memory cell conceived, developed and industrialized by STMicroelectronics for General Purpose and Secure Microcontrollers embedded applications. Thanks to its unique architecture, the eStM cell gathers the advantages of a conventional split-gate NVM cell together with a more compact cellbit area than a typical 1T Flash Memory cell, and it is claimed to represent the scalability limit for a floating gate based NOR NVM.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117016552","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
S. Ramesh, S. V. Palayam, E. Rosseel, A. Arreghini, B. Kunert, M. Baryshnikova, L. Zhang, P. Ong, L. Teugels, M. Pak, N. Jossart, T. Raymaekers, J. Stiers, G. Van den bosch, A. Furnémont
{"title":"First Demonstration of $text{In}_{1-mathrm{x}}text{Ga}_{mathrm{x}}text{As}$ Macaroni Channel for Future 3-D NAND","authors":"S. Ramesh, S. V. Palayam, E. Rosseel, A. Arreghini, B. Kunert, M. Baryshnikova, L. Zhang, P. Ong, L. Teugels, M. Pak, N. Jossart, T. Raymaekers, J. Stiers, G. Van den bosch, A. Furnémont","doi":"10.1109/IMW.2019.8739580","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739580","url":null,"abstract":"We demonstrate for the first time epitaxially grown InGaAs macaroni channel down to 15 nm thickness for 3-D NAND with a low temperature (< 575 °C) process. The epitaxial InGaAs macaroni channel is selectively grown inside memory hole, with different compositions of In and Ga, by replacing an a-Si sacrificial channel. InGaAs macaroni channel has improved gate control over the InGaAs full channel. As a result, lower IOFF w.r.t. full channel is demonstrated, while retaining a similar ION of full channel InGaAs and same memory performance.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131073057","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
T. Mikawa, R. Yasuhara, K. Katayama, K. Kouno, T. Ono, R. Mochida, Y. Hayata, M. Nakayama, Hitoshi Suwa, Y. Gohou, T. Kakiage
{"title":"Neuromorphic computing based on Analog ReRAM as low power solution for edge application","authors":"T. Mikawa, R. Yasuhara, K. Katayama, K. Kouno, T. Ono, R. Mochida, Y. Hayata, M. Nakayama, Hitoshi Suwa, Y. Gohou, T. Kakiage","doi":"10.1109/IMW.2019.8739720","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739720","url":null,"abstract":"We have developed neuromorphic computing based on Analog ReRAM, Resistive Analog Neuromorphic Device (RAND), as low power solution for edge application. We have proposed perceptron circuit which has resistive elements to store weights as analog resistance and binarizes output from each layer in order to realize large scale integration and keep high accuracy. We have fabricated 180nm test chip by using mass production process and we have demonstrated MNIST recognition and sensor application in flexible network architecture in which several neural networks can be configured at the same time. We present potential of low power solution by scaling down to 40nm node under development and reliability issues to be considered in neural network processors based on analog nonvolatile memories.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"266 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114980239","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
T. Hsu, H. Lue, P. Du, Wei-Chen Chen, T. Yeh, R. Lo, Hung-Sheng Chang, Keh-Chung Wang, Chih-Yuan Lu
{"title":"Study of Self-Healing 3D NAND Flash with Micro Heater to Improve the Performances and Lifetime for Fast NAND in NVDIMM Applications","authors":"T. Hsu, H. Lue, P. Du, Wei-Chen Chen, T. Yeh, R. Lo, Hung-Sheng Chang, Keh-Chung Wang, Chih-Yuan Lu","doi":"10.1109/IMW.2019.8739692","DOIUrl":"https://doi.org/10.1109/IMW.2019.8739692","url":null,"abstract":"A self-healing Flash with micro heater was proposed [1] to enable very high endurance (100M) for NAND Flash device by heating the device locally during P/E cycling stress. In this work, we propose to use the self-healing 3D NAND Flash with thermal assisted programming/erasing speed and much higher endurance to improve the performances of fast SLC NAND in NVDIMM applications. By applying an in-situly thermal heating of nearly 600C during programming and erasing we demonstrated the feasibility of sub-usec thermal assisted programming, and sub-10usec thermal assisted erasing with 100M endurance for SLC operation. The retention of self-healing 3D NAND Flash after 100M endurance still remains well, suggesting the good healing effect. System-level simulations indicate much reduced latency by the boosted PGM/ERS, while the greatly improved endurance greatly reduces the need for spare blocks for wear leveling (=lower cost) for NVDIMM applications.","PeriodicalId":260146,"journal":{"name":"2019 IEEE 11th International Memory Workshop (IMW)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129005837","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}