Consumers have an insatiable demand for more functionality in their smart electronics – whether that be their mobile phones, consumer electronics devices, or auto entertainment systems. This is where storage selection has been, and will continue to be, absolutely critical. All of these systems have become extremely complex over the last decade, and designers continually need more density in their Flash memory to store the code that will provide the next wave innovative capabilities and features. The challenge, however, is adding that density without increasing the cost of the device or system too much. As history has proven, consumers want more features and functionality, but they are not willing to pay much more for it. That puts the task onto the device makers to figure out how to get faster memory functions so that consumers have better user experiences, but without significantly increasing the cost for both the manufacturers and the users.
Markets and applications
All electronic systems today use some type of Flash memory and the demands for higher performance memory are increasing at a dramatic rate due to the emergence of new and exciting applications. In particular, the markets and applications highlighted below are experiencing rapid advancements that can only be made possible through new and better memory subsystems that can handle faster read, program, erase functionalities – and instant boot up.
- Automotive – Advanced Driver Assisted Systems (ADAS), Clusters, Push Button
- Artificial intelligence – training models, database storage, autonomous driving
- Over the Air (OTA) – consumer, automotive, gaming consoles
In the automotive market, throughput is very important when selecting memory, particularly in applications such as advanced driver assisted systems. The cluster of instruments being added to the digital displays of automobiles are designed to ensure that the driver is comprehensively and reliably informed at all times. These displays need to turn instantly on and render 2D/3D images quickly and accurately. Current cluster users typically leverage SPI NOR (Octal or Quad) or SPI NOR (Octal or Quad) +eMMC for cluster usage. However, the features/functions are growing more sophisticated which makes the code size exponentially bigger. This expanding code size requires more density of memory which increases the cost.
In addition, although the code size is bigger, the booting time needs to be the same as before or even shorter. This is very important in the automotive market for features such as the push button where multiple screens need to show up instantly when the user pushes the button.
For OTA applications, the major requirement is erase/program time because the purpose of OTA is to update the codes so that higher erase/program speed is better. Automobiles, such as E-cars and many consumer devices are leveraging OTA more and more today due to its flexibility and ease of use for the user. These devices are all mobile and they require an easy and efficient way to receive critical updates.
Other applications that also need fast data and high throughput include AI and face recognition. AI applications typically need to run a wide variety of training models to make a decision for the end device and then tell the end device what it needs to do. To do this, they need to store the training models inside and often these are very large models that need to be processed. In the case of autonomous driving, these models need to be loaded into DRAM from very high throughput NAND because they need to make a decision in a very short time (in milli-seconds) – and they need to be accurate decisions such as avoiding a pedestrian or a car. Data throughput here is absolutely critical and it also has to be instant.
NOR vs. NAND
Most systems today use NOR flash in their designs. Some applications such as automotive and industrial are reaching the limits of NOR flash technology’s ability to scale cost effectively while being assembled in industry standard packages. In applications such as automotive graphics displays and Advanced Driver Assistance Systems (ADAS), boot code size is often larger than 512Mbits. Customers have traditionally used NOR Flash for code storage because it is highly reliable, and its fast read throughput supports fast boot performance. At densities above 512Mbits, however, NOR flash scales poorly, and die size and cost become excessive. Because NAND Flash technology scales well, its cost is more attractive at densities of 1Gbit and higher, but conventional NAND flash has been too slow for functions such as shadowing boot code to DRAM.
Some vendors are trying to solve this problem by introducing Octal interfaces for NOR that provide 8 I/Os instead of 4. However, these solutions are proving to be too costly because the cost of NOR memory is so high. However, by switching to Octal interfaces for NAND, manufacturers can solve these problems by having both a better cost structure and higher throughput. Octal interfaces for NAND Flash memory enable much higher data throughput than general NAND Flash memory and much higher program/erase speed than NOR. It’s also much cheaper because the die size of NAND is much smaller than NOR. In fact, one gigabit of NOR Flash 8 I/O is double what an 8 I/O NAND solution would cost.
Another advantage between NOR and NAND is programming speed. NAND always has shorter program time than NOR which is critical in applications such as OTA in automotive applications where they need short time to program the code and firmware inside the system.
Clearly, there is a huge market for solutions that can eliminate the tradeoff between density, performance and cost. According to IDC, there will be 50 million L1/L2 cars in 2024. The CAGR of EV cars is approximately 18% which means there will be huge demands for memory ICs because every system needs a boot-up sequence to enable the system. Not to mention that in addition to boot-up, other features need to be loaded for each system.
The gaming market is also a significant opportunity, with the total gaming console shipment in 2020 at approximately 47.8 million. For better gamer experiences, the firmware updating speed for consoles is critical and this requires performance and throughput from the memory subsystem.
Leveraging Octal interfaces for NAND flash memory will enable automotive, AI, consumer electronics, and industrial manufacturers to tap into this market growth opportunity by providing code storage in high density without having to pay a premium for NOR flash, a fast memory technology which scales poorly at densities above 512Mbits.
Wilson Huang is Flash Memory marketing manager at Winbond.
>> This article was originally published on our sister site, EE Times Europe.
- Flash 101: NAND Flash vs NOR Flash
- Flash 101: Types of NAND Flash
- NAND flash paces memory controller requirements
- How flash memory can support functional safety requirements
- Latest flash storage spec aids automotive, edge AI
For more Embedded, subscribe to Embedded’s weekly email newsletter.