Leading memory for AI, IoT edge and autonomous driving vehicles
Electronic Specifier had a chance to meet with memory solutions provider Micron at the recent electronica exhibition in Munich. The company, with headquarters in Boise, Idaho, specialises in DRAM, NAND Flash, NOR Flash, Hybrid memory solutions and Solid State Drives (SSDs), which have applications in a broad line of industries such as automotive, consumer, embedded, industrial, mobile, networking and financial.
Memory and storage is the biggest part of the semiconductor industry, and Micron is focused on the following business segments – computing data centres, mobile phone storage and embedded business. The focus of the show for Micron was on embedded applications, which included consumer IoT, Industrial IoT and automotive solutions.
The memory and storage market has increased substantiously, becoming bigger year-on-year. For example, in 2017 22 billion Gigabits of data was created, and this is expected to rise to 62 billion in 2021. The modern needs of the industry are for continuous capturing, processing, moving and storing of data, generating ever-increasing demand for memory and fast storage for automotive, mobile, and industrial applications.
Speaking of IoT, the huge trend right now is in Edge intelligence. In modern industry, infrastructure and assets (smart factories, cities, transportation, infrastructure and others) create an enormous amount of data. Edge intelligence is happening between Cloud and the Things, where the data is converted into meaningful results. Many Edge devices are coming onto the market – smart data aggregators, gateways and others. And it creates a lot of opportunities for companies like Micron to evolve and innovate.
For example surveilance. By 2021 160 million cameras will be shipped annually, 28% of them will have 4k resolution or even more, 15% of them will be AI cameras with significant ammount of storage. A typical AI enabling camera will contain 32/64GM eMMC, 4-8Gb DDR4/LP4 with 256Gb -1TB of edge storage.
The data processing will occur in the edge server, communicating with the Cloud orsmart city datacentre. For example, in 2017 a standard IP camera with edge storage contans 0.5GB of DRAM or 64GB of NAND memory. By 2021 an IP camera with edge storage will contain 8GB of DRAM or 1TB of NAND storage.
The challengefor innovation is keeping the price for these devices reasonable with the amount and quality of the memory components having significantly increased. All the datafrom the device can be aggregated, processed and stored at the edge level, and only then stored into the Cloud. So intelligent devices can be a part of the edge process. When the devices are aggregated together, the edge process can be integrated into smart city datacentres. It’s here we can connect the cameras, infrastructure devices and others.
Security of these devices is also a big challenge. The typical key attacks points can be brown outs, credit card theft, ransomware, control hijacking and others. To secure its solutions for IoT, Micron developed the Micron Authenta Technology, which covers different security profiles for silicon manufacturing, component distribution, IoT device manufacturing, deployment and field lifecycle.
By using this technology the designer can utilise existing flash memory, hardening the system level security without additional hardware components. The Micron Authenta NOR flash device offers hardware-based security, protecting the software that is also running on the device.
From a security point of view Micron is working with companies like Rubicon Labs, Spyrus, TrustiPhi, Karamba Security, and Move by TataCom to provide the best security for its solutions for IoT.
Automotive is another strong focus for Micron. New autonomous vehicles require a huge amount of computing, where memory plays a very important role. When autonomous vehicles are driving, they collect a huge amount of data about their environment, and this requres a lot of data storage and processing power.
So the cockpit of today’s cars contains HD maps, 4K screens, voice and gesture and infotainment control. In future, let's say by 2021, autonomous vehicles will require a lot of storage to contain and process information from the 40+ sensors on the vehicle, high resolution maps, information from ADAS, sonar and other sensors, communication information between vehicles, and all this data will be stored locally in the vehicle.
In 2017 for L1/2 types of vehicles DRAM and NAND occupied only 8GB of data per vehicle. By 2021 vehicles are expected to evolve to the L3 level, containing 16GB of DRAM and 256GB of NAND data per vehicle. By 2025 vehicles will evolve to the L5 level, containing 74GB of DRAM and 1TB of NAND data per vehicle. NAND is the high resolution content, maps, and data from sensors regarding events.
Density and bandwidth memory is progressing, in so far as the ADAS sensors are becoming more advanced and complex, as is the data required to be processed in real time, so memory requirements are also growing. Micron has solutionsfor these situations. During electronica PCIe was announced, which gives increased performance and power consumption, performing 1,600MB/s and 2,000MB/s PCIe 2 and 3 Gen and PCIe 4 Gen correspondingly.
Regarding memory bandwidth, Micron offers GDDR6 @14Gb/s performing the x128, x256 and x512 bus width for the bandwidth 224Gb/s, 448Gb/s, 896Gb/s correspondingly. PCIe delivers twice the sequential performance, reaching the sequential reading performance of 2,000MB/s and sequential writing performance 110MB/s, compared to the UFS 2.1 and 3.0. The energy consumption for 1TB Read/Write for the PCIe significantly decreases from 5 Joles for UFS 2.1 to 2 Joles for PCIe SSD.