Monday, July 8, 2019
Emerging memory technologies have been emerging for decades, but this year’s report by Objective Analysis and Coughlin Associates, Emerging Memories Ramp Up, suggests they have reached a critical point where they make sense in more applications than ever before.
The report found that the emerging memory market is poised to reach $20 billion of combined revenues by 2029, largely by displacing today’s less efficient memory technologies such as NOR flash and SRAM—even displacing a share of DRAM sales. They will also be more competitive with incumbent memory technologies, as future process shrinks and improved economies of scale will drive down prices, enabling adoption both as standalone chips and embedded within ASICs, microcontrollers, and even compute processors.
The three key emerging memories to keep an eye on are PCRAM, MRAM, and ReRAM; NOR flash and SRAM are prime candidates to be replaced. (Source: Coughlin Associates)
But although the $20 billion figure sounds impressive, it’s important to remember that it’s the market for all emerging memories, with MRAM and PCRAM primarily in the form of Intel Optane being the dominant types, according to Jim Handy, principal with Objective Analysis. “DRAM reached the highest level that it may ever reach in its lifetime last year at just shy of $100 billion,” he said in a joint telephone briefing with co-author Thomas Coughlin, founder of Coughlin Associates.
The report pegs the revenue growth of 3D XPoint — again, PCRAM in the form of Intel Optane — at $16 billion by 2029, thanks to its sub-DRAM prices. Stand-alone MRAM and STT-RAM revenues, meanwhile, will approach $4 billion, or more than 170 times MRAM’s revenues in 2018. Along with ReRAM, MRAM is expected to compete to replace the bulk of embedded NOR and SRAM in SoCs, fueling even greater revenue growth.
Emerging memories cover a breadth of technologies, but the key ones to watch out for are MRAM, PCRAM, and ReRAM, said Coughlin. There’s been discrete MRAM devices out for some time, but there’s been a lot of talk about foundries building ASICs with specialized chips and replaced volatile memories with non-volatile options, he added. “That's going to be one of the big biggest drivers.”
andy said NOR flash’s inability to scale past 28 nanometers is also shifting attention to alternatives. “In the past, the only reason to have an emerging memory in an embedded application like on an MCU or on an ASIC is because you needed some technical attributes that it had, but it always added cost.” The prospect of being able to replace NOR flash using a smaller process node is spurring interest, he said.
The three key emerging memories to keep an eye on are PCRAM, MRAM, and ReRAM as NOR flash and SRAM are prime candidates to be replaced. (Source: Coughlin Associates)
Overall, the economics of the emerging memories are also improving, said Coughlin, as foundries don’t necessarily have to add another back-end process but make it a part of the existing CMOS processing, which lowers the manufacturing costs as volumes go up and yield improves. He said it’s a “golden age for MRAM” because it’s getting an opportunity to prove itself in volume as foundries are looking to build it into embedded chips, while PCRAM has some support behind it, thanks to Intel Optane. Meanwhile, ReRAM is getting a lot of attention for artificial intelligence and machine learning applications. Even FRAM remains in the mix, as a dark horse contender for some applications.
However, Handy believes it’s too early to name a winner, and despite the healthy forecast ahead for emerging memory technologies, it’s still hard for them to get a leg up on entrenched technologies. “They're really hard to knock out of their leadership position even if the economics are improving. If you're not the cost leader then all these wonderful technical benefits you have over the entrenched technologies don't really mean an awful lot.”
Copyright © 2019 CST, Inc. All Rights Reserved