TORONTO — As cars get smarter and demand more memory, many technologies are angling for the driver's seat, but it's safe to say NOR flash at least gets to ride shotgun.

As a successor to EEPROM in many applications thanks to its programmability capabilities, NOR flash is finding new opportunities in application areas that need fast, non-volatile memory, including communications, industrial and automotive. The latter, of course, is getting a lot of attention thanks to autonomous vehicle development.

Macronix International, which describes itself as the leading supplier of NOR flash overall, find itself in the third position for automotive. But Anthony Le, senior director of marketing, ecosystem partnership and North America automotive, said the company is confident it will lead that segment in the next two to three years.

Le's career has been well-aligned with the evolution of NOR flash, beginning with his first technology program in the mid-1990s when already NOR flash was able to withstand the high temperatures needed for use under the hood. Today, said Le, that heat tolerance combined with data retention for up 20 years puts Macronix in a good position to capitalize on all the automotive opportunities.

NOR flash got its start in radio — an automotive application that didn't need a lot of memory, said Le. But in the last decade, telematics and all of features and functions in the center console have upped the requirements for memory. Today, you can't sell a car without a digital display. Instead of radios that need 1 megabit of NOR flash, its clusters that need 12 megabits, even as much as 1 gigabit thanks to all the graphics, he said.

“In a high performing system, like under hood, you can't use any other non-volatile memory besides NOR flash," Le said. 

NOR Flash applications in cars

NOR flash got its automotive start in radio, but now there are opportunities throughout the vehicle thanks to its fast boot up, reliability and endurane.

Performance is ultimately what's putting NOR flash in the driver seat today, said Le. “We're getting to the point where you can get almost instant on capabilities because we're able to really boot up the processor at about 500 megabytes per second," he said. When you get into the car and turn on the key, he added, you want that display and that rear-view camera to turn on within a second.

Jim Handy, principal analyst with Objective Analysis, said that “instant-on" is why NOR flash is used in dashboards because it can bring up the basic functionality right away — NOR is better for executing code because special software is needed to manage having your code in NAND flash. “And what you do with that software is that it moves the code from the NAND to a DRAM. And then everything is executed out of the DRAM," he said. “You can execute code right out of the NOR flash, but you can't execute right out of the NAND."

NOR is already in a lot of different parts of the car today, and it will find its way into more because it's good for subsystems, said Handy. If it's got very little code, then it will use a microcontroller that has the NOR inside the microcontroller. But if it's got a larger amount of code because it's a more sophisticated system, then it will have an external NOR flash chip. “You look at today's self-driving cars and all of them have at least one lidar, and each one of those lidars is going to have a NOR flash in it to control the various functions inside that," Handy said.

There was a time when NAND flash was expected sideline NOR flash, but NOR has persevered for several reasons. One is that while NAND does better in applications where you need to store a lot of data, there is a minimum cost for a NAND flash chip, said Handy, so when it comes to smaller amounts of code, you can't buy a fifty-cent NAND flash chip. “You can buy a fifty-cent NOR flash chip," he said. "And it's just more economical for manufacturers of NOR to make a small chip than it is for manufacturers of NAND to make a small chip."

Meanwhile, even automotive-grade NAND can't handle the heat that NOR can, and it doesn't have the reliability for mission-critical automotive applications. Longevity is also a key characteristic of NOR flash, said Macronix's Le, as automakers like stuff that will last a decade to support replacement parts. “We're hearing a lot of pressure for 15 years now. These cars are sitting around, and they still need these parts."



Cypress Semiconductor Corp. has pole position when it comes to NOR flash in automotive and has been proactive in addressing the instant-on demands of applications such as advanced drive-assistance systems (ADAS). The company recently introduced a new family of NOR flash, dubbed Semper.

Sam Geha, executive vice president of Cypress' memory products division, said it's the first memory architected and designed to meet the automotive industry's ISO 26262 functional safety standard for building fail-safe embedded automotive systems. And like Macronix's NOR flash offerings, Semper claims superior endurance and data retention at the extreme temperatures common in automotive applications.

Geha said the safety standards for an automotive were a key driver for building a new NOR flash architecture. “You're an automotive supplier, do you want to buy a product that meets the safety standards, or do you want to buy somebody else's low-end product?" He also sees Cypress as being able to differentiate with its MCU and SOC expertise.

Of course the instant-capabilities are critical, given that every vehicle now has at least a rearview camera for the ADAS—the fast boot requirements of the instrument cluster have been a key market for Cypress' Hyperbus technology. “Now we're going toward autonomous, there won't necessarily be a person in the car, you want to make sure this chip doesn't screw up the car," said Geha.

But it's not just instrument clusters and cameras where NOR flash is needed. Cypress also sees the powertrain, the engine controls and the chassis controls as a big market, and reliability is critical due to the high temperatures, making NOR flash the preferred technology. Geha said what's differentiating Semper is its architecture, which is starting a point for Cypress to do more with NOR flash for automotive, likening to Lego in that IP blocks can easily be added for more functionality. “We're going to be able to add other very important features with this platform," Geha said. 

ARM Cortex M0 featuring NOR Flash

Cypress created its Semper Flash architecture from scratch to address requirements of the automotive industry so it can functionality over time.

NOR does have its disadvantages, and a critical one today has nothing to do with technology, said Macronix's Le. Because the market is roughly about $1.8 to $2 billion dollars, many players had dropped out. And despite the opportunities on the near horizon, the road for NOR flash doesn't go on forever. Le acknowledges it's hitting a wall at this point in its evolution because it's difficult to strengthen the technology. “It's almost impossible. We're struggling with it." He said the primary barrier is the limitations of floating gate technology. The roadblocks are not dissimilar to what planar NAND faced—once you start shrinking, the read reliability drops and you need to add wear leveling and bad block management almost, right? “We're close to the very edge of NOR," said Le. “We're still able to do some tricks to squeeze out the technology."

There are some paths forward. Like NAND, there's potential for 3D NOR, said Le, which is being investigated. Macronix is doing multi-chip packaging to get the same densities in a single chip and die stacking. He expects NOR flash will be fairly dominant into 2025 in part because the longevity and qualifying that automotive requires. “Beyond that, we are looking at other technologies to continue."

Handy said that while there are people working on 3D NOR architectures, not sure where that will end up. “One thing to keep in mind is that even though Macronix is talking about reaching the end of the road by 2025, that's for new parts. “NOR flash has this great huge longevity in that a part gets introduced and 20 years later it's still shipping in high volume," he said. With NAND, every time there's a new generation, it pretty much works to squeeze out the older generation. “A legacy NAND part might be a five-year-old part. A legacy NOR part might be a 30-year-old part," Handy said.

—Gary Hilson is a general contributing editor with a focus on memory and flash technologies for EE Times.