2017/08/15 STTMRAM is moving to large scale commercialization, at last!

According to Yole Développement report Emerging Non-Volatile Memory 2017, the market for emerging NVM will grow at a 106% CAGR between 2016 and 2022 to reach around US$3.9 billion by 2022. The new SCM hierarchy and embedded MCUs will drive this.

Source: Emerging Non-Volatile Memory 2017 report, June 2017, Yole Développement

Yole Développement had the opportunity to speak with David Eggleston, Vice President of Embedded Memory at GlobalFoundries. We talked about the company’s technologies, position in the industry and outlook on the market. The discussion follows.

Yole Développement: Can you please discuss about the background of GlobalFoundries involvement with emerging NVM technology?

David Eggleston: GlobalFoundries has investigated multiple emerging NVM technologies, and determined that eMRAM provides the versatility required by our customers. Therefore, GlobalFoundries selected Everspin as a development partner, and we have co-developed eMRAM technology in our GlobalFoundries fab over the last 3 years. We are now at the point of offering a commercial eMRAM solution on our 22FDX platform. GlobalFoundries expects to offer eMRAM at future smaller process nodes beyond 22nm, where the cost of scaling eFlash becomes prohibitively expensive.

YD: Emerging NVM technologies have long history of development. But somehow their adoption to market remain limited due to various factors such as limited density product availability and high density products introduction delayed by emerging NVM pioneers , technical issues, etc. However at Yole, we see favorable factors are now emerging that will accelerate the adoption of emerging NVM: Introduction of product by big players, emergence of new SCM segment, involvement of top foundries, new application drivers (such as IoT, big data, AI etc.). What is your opinion on this?

DE: Yes, we do see accelerated adoption of emerging NVM not only from foundries but also from customers. The emerging NVM technologies are successfully addressing the technical requirements such as higher memory density, low power consumption, and high performance that are critical to these new application drivers. In addition, the major fab equipment vendors began investing in the development of mass production equipment specifically for emerging NVM, and the appearance of these new tools mean that emerging NVM is now also a sufficiently cost effective solution for customers.

YD: There are many emerging NVMs in development but three candidates have clearly emerged as winner, each having backing from top players: MRAM/STT-MRAM, RRAM and PCM. What are the advantages & disadvantages of each technology in terms of scalability, manufacturing challenge and performance? 

DE: eMRAM has the greatest versatility of all the memory types, with superior performance and reliability. However, eMRAM is a complex technology to master, and requires a significant investment to develop and deploy. We do see that four leading players have invested in eMRAM at 28nm node or below, and expect eMRAM to be sourced from multiple foundries in the 2018-19 timeframe.

RRAM is also being developed as a simpler, lower cost option, and competes directly with eFlash at the 40nm node. RRAM does not offer the versatility, performance, and reliability of eMRAM, but perhaps is suitable for very cost sensitive markets like chipcard. Two or three players which have not, or cannot afford to invest in eMRAM are the primary drivers of RRAM. RRAM will appear in the market at 40nm in 2017-18, and has been reported to appear at more advanced nodes (such as 28nm) in 2019+.

PCM is unlikely to emerge as a popular embedded NVM, as it does not offer either superior performance and reliability (like eMRAM), or the ability to compete with eFlash as a low cost option (like RRAM). Only one player is still pursuing PCM, and they have indicated they will no longer invest beyond the 28nm node. Therefore, GlobalFoundries does not expect PCM to play any role in the emerging NVM market outside this single player.

Bottom line: eMRAM has the lead, and is widely supported. RRAM will continue as a low cost option. PCM has a very limited future.

Source: Emerging Non-Volatile Memory 2017 report, June 2017, Yole Développement

YD: Do you think that these emerging NVM can replace traditional memories DRAM and NAND in next 10 years?

DE: The future of DRAM is DRAM, the future of NAND is NAND. It is very difficult to replace these incumbent standalone memory technologies in the existing applications. What is more probable is the opening up of a new application that emerging memory is uniquely suited to solve. 
However, in the embedded NVM space, the scaling difficulties (cost, performance, power) of eFlash make the commercialization of emerging NVM much more probable.

YD: What do you think is the key manufacturing challenge for STT-MRAM for embedded & standalone application?

DE: MRAM being an emerging technology, continually improving the bit error rate (BER) and yield are the major challenges.

YD: What is the expected time line for production of MRAM at GlobalFoundries for both embedded & standalone application?

DE: In 2016, GlobalFoundries announced it will complete qualification of 22FDX eMRAM in 2018, and we will release production for our lead customers in late 2018, ramping in 2019. GlobalFoundries will not comment on the standalone MRAM timeframes, our partner Everspin will report on those projections.

YD: We have identified 8 applications for emerging NVM: Embedded application (MCUs, mobiles devices SRAM; SRAM cache HPC and embedded NVM SoC); Standalone application (enterprise storage SCM; client SCM; mass storage, industrial, transportation, and consumer electronics); . What applications you are targeting in the short, medium and long term? 

DE: GlobalFoundries 22FDX eMRAM targets the IoT, GP-MCU, and automotive applications. We expect adoption across all those markets, in the order given.

Source: Emerging Non-Volatile Memory 2017 report, June 2017, Yole Développement

YD: Storage class memory (SCM) has emerged as the key business segment for emerging NVM PCM based product is already introduced in market targeting this application and RRAM has strong potential to capture this business. How do you see this as challenge for STT-MRAM to gain the SCM market?

DE: GlobalFoundries is focused on the embedded NVM opportunities outlined above, and therefore we have no comment or view on the SCM segment.

Sources: Yole Développement - GlobalFoundries

Interviewee: Dave Eggleston, Vice President of Embedded Memory at GLOBALFOUNDRIES 

Dave Eggleston is the Vice President of Embedded Memory at GLOBALFOUNDRIES, which he joined in 2015. Dave has responsibilityfor the embedded volatile and non-volatile memory technologies at GLOBALFOUNDRIES, as well as the related strategic direction and initiatives. Dave is the former CEO and President of Unity Semiconductor, an RRAM industry pioneer acquired by Rambus. He has held technical executive management roles at Rambus, Micron (where he built and spearheaded the NAND systems engineering organization), SanDisk, and AMD. He holds 28 patents in NAND flash and next-generation ReRAM memory, storage system usage, and high volume manufacturing. He currently serves on the Board of Directors of two NVM start-up companies. He received his MSEE from Santa Clara University and his BSEE from Duke University.