Page 23 - EE|Times Europe Magazine - December 2020

EE|Times EUROPE — The Memory Market 23

Emerging Memories May Never Go Beyond Niche Applications

ENTER OPTANE the lab to the production line are “enormous.”
Phase-change memory in the form of 3D Xpoint, jointly developed by Deep pockets and experienced engineers are required. So, too, are
Intel and Micron Technology, might have finally reached a tipping point long-term commitments and patient capital, according to Michael
that allows it to graduate from the “emerging” category. PCRAM has a Yang, senior director for memory at research firm Omdia. Few com-
long history, but only in the last decade has all that research moved into panies other than Intel have been able to push an emerging memory
commercialization. PCRAM is nonvolatile and exploits the unique behav- technology across the finish line. Optane “can turn a corner because
ior of chalcogenide glass. So far, its one real commercial success has been there is a market,” Yang said.
in Intel’s Optane. Like 3D NAND, PCRAM has had its growing pains. The market is not replacing DRAM or flash but using alternative
However, 3D Xpoint has proven more cost-effective to manufacture. memory where it makes sense. Yang said that there’s a lesson for other
Moreover, chipmakers other than Intel and Micron might be able to emerging memory makers in Intel’s Optane approach: Find a profitable
build products around it. Intermolecular Inc., a subsidiary of Germany’s market for technologies rather than try to displace a proven memory.
Merck, recently disclosed research around a new combination of mate- “For [technologies aspiring] to replace DRAM or flash, that train left
rials that could enable a 3D vertical nonvolatile memory architecture, the station years ago,” he said.
shaping the second iteration of 3D Xpoint technology. Most MRAM companies have been around too long to be considered
Intel currently offers Optane as both DIMMs and SSDs and has posi- startups, making little progress beyond embedded markets — let alone
tioned it as storage-class memory. The company seems comfortable supplanting DRAM or flash.
with the technology replacing DRAM and flash in some use cases, but Claiming that you can replace DRAM or flash might make for great
not all. For example, the Optane Persistent Memory 200 Series comes marketing pitches, but it is not realistic from an engineering perspec-
in the form of DIMMs aimed at replacing NAND SSDs for real-time tive. Most companies simply can’t be the John Wick of memory, said
analytics applications. The reason is Yang, but there is business potential in niche memory markets such as
speed: 225× faster than a main- embedded and low-density devices. Hence, it may be time to jettison
stream SSD. the “emerging” label, whether it’s MRAM, ReRAM, or FRAM. “They’re
One advantage of SSDs was their just little memory,” Yang said.
ability to handle random inputs/ Thomas Coughlin has been tracking the emerging memory mar-
outputs, but that was at a time when ket for the past three years in collaboration with Objective Analysis.
CPUs had just a few cores and vir- Coughlin said that embedded is the best opportunity for many of the
tualization was just getting started. emerging memories. Even if it’s a standalone, a device is going to be
With multiple cores, virtualization, low-density for applications that require low power consumption.
and software containers taking off in “There are a lot of applications that would benefit enormously from a
the cloud, the challenge has become little bit of nonvolatile memory,” he said.
feeding all these cores. DRAM and NOR flash may eventually hit scaling limitations. That
As DRAM approaches scaling would create an opening in embedded applications for FRAM, MRAM,
Intel’s Christopher Tobias limits, there is an opportunity for and ReRAM once they pass qualification screening.
Optane to augment DRAM, serving Whether the moniker is still apt or not, MRAM and ReRAM are still
CPUs in a DIMM format, Tobias said. “We’ve been producing it at scale, emerging memories, but Coughlin said that Optane has emerged, as
and it serves a critical need. We can augment memory in feeding all Intel has likely reached a break-even point for Optane manufacturing.
these virtual cores, and compute becomes denser.” Meanwhile, its customers are realizing the need for fast, inexpensive
Optane has reached the point at which it’s cost-effectively filling memory, like DRAM, he said. “It’s an inexpensive memory expansion
a layer between NAND SSDs and DRAM, whether it’s complement- capability. It allows lower-cost, memory-intensive applications.”
ing DRAM or meeting latency demands that traditional SSD cannot, While Optane may be a billion-dollar product given its suitability
according to Tobias. for workloads like real-time analytics, said Yang, other memory makers
“The big cloud companies have a real need to fill in these layers,” he need to come to grips with the fact that their market potential is
said, adding that hurdles remain, as software and architectures need to smaller but healthy.
catch up to fully take advantage of Optane’s performance gains. “It’s not black or white,” he added. “You can be successful at
“Bringing a new silicon technology to market is very daunting,” said US$30 million. It’s all different shades of success.” ■
Tobias. “There are multiple challenges to overcome,” such as scaling
manufacturing; the hurdles to transitioning a working prototype from Gary Hilson is a contributing editor for AspenCore.

NEWS

SEMI: 300-mm Fab Investments Rose in 2020, Will Boom in 2023

Global demand for 300-mm capacity remains robust, driven by growth by a 20%, or US$70 billion, record high in 2023. The year 2024 will,
in cloud services, servers, laptops, gaming, and health-care technology. however, experience a 4% downturn, SEMI said.
SEMI expects the final tallies for 2020 to show 13% growth in 300-mm SEMI has made the conservative projection that at least 38 new
fab investments year over year, with a record high forecast for 2023. 300-mm fabs will open from 2020 to 2024. Asia will be well-repre-
In its latest report, the trade organization covers the installed base sented, as Taiwan and China are set to add 11 and eight 300-mm
of 300-mm fab capacity, current and near-term market trends, and volume fabs, respectively. The chip industry will then command 161
forecasts to 2024. After a surge in 2020, semiconductor fab investments 300-mm volume fabs by 2024, SEMI said.
are expected to continue to increase in 2021 but at the reduced rate of Over the same time interval, SEMI expects per-month fab capacity to
4% year over year. SEMI then predicts a 2% slowdown in 2022, followed increase by about 1.8 million wafers to reach over 7 million. ■

www.eetimes.eu | DECEMBER 2020