隨著 5G 問世，人工智慧（AI）、邊緣運算、大數據等也將面臨大幅變化。但是這些由 5G 推動的應用程式，對記憶體和儲存空間具有什麼意義呢？I sat down with Micron Mobile Business Unit’s VP of marketing, Chris Moore, to talk about how he sees memory and storage powering the brave new world of 5G.
Me: The 5G era requires not only changes in storage technology but also changes in computing methods. Edge computing will be widely used. In this process, there will also be challenges across privacy, security, reliability, power consumption and other factors. How do you think about solving these pain points from the perspective of storage?
Moore: Edge computing certainly has huge implications for memory and storage: 5G-connected edge devices and the internet of things (IoT) will need the computing horsepower that was once exclusive to servers, PCs and smartphones. At Micron, we’re tackling this need by offering scalable, high-bandwidth, low-latency storage and memory for the network edge — necessary for enabling real-time analysis and processing in edge devices and for handling the rapidly growing volumes of data we’ll see. Another challenge with 5G and edge computing is definitely power consumption, especially since research states that 5G infrastructure will need increased amounts of power compared to 4G. And it’s already predicted that information and communication technology will account for up to 20% of total electricity consumption worldwide by 2030. Therefore, it’s important that those at all layers of the 5G ecosystem design energy efficiency into their infrastructure from the start.
We’re constantly innovating to increase the power efficiency, performance and speed of our products and believe these benefits go hand in hand. One example is our low-power DDR5 DRAM, which offers more than 20% improvement in power efficiency and a 50% increase in data access speeds compared to previous generations. These capabilities allow 5G smartphones to process data at peak speeds of up to 6.4 Gb/s, speeds that are critical for preventing 5G data bottlenecks. And this capability is well-suited to devices at the edge too, for example, helping autonomous vehicles to compute and process data in real time. It’s also important to note that edge computing saves energy since, by enabling data processing at the edge itself, it eliminates the need to send data back to the data center hundreds of miles away, thereby reducing energy expenditure.
As far as security, with the rise of the edge, the key is protecting the data precisely where it is generated, more specifically, embedding security directly in flash memory hardware to better protect data at the system level. We’re tackling this approach with our proprietary AuthentaTM technology, which offers chip-level security for edge devices with a hardware-based root of trust, guarding 5G-enabled devices against unauthorized access and malware. This silicon-level security is key because today the flash has become a target for many hackers to create advanced persistent threats that can mask themselves from higher levels of code. By moving essential security primitives in memory, it becomes simpler to protect the integrity of code and data housed within the memory itself.
Me: As 5G rolls out, how does Micron plan to support the growing need to communicate large datasets in real time?
Moore: The need to solve the issues around large datasets and near instantaneous responses from 5G has put significant pressure on the underlying hardware to ensure no loss of performance even under the heaviest use conditions. As the first vendor to volume production on the 1z node, this will be a key focus for Micron as we look to the rewards that this technology lead brings. Our 1z-based DDR5 lays the critical foundations to enable the next generation of technologies for 5G and beyond. DDR5 is uniquely suited to handle advanced data center workloads from rapidly expanding datasets and data-centric applications, with a more than 85% increase in memory performance and with the increased reliability, availability and serviceability that modern data centers require.
Me: How will storage and memory demands evolve in 5G smartphones?
Moore: Rising entertainment, social networking, camera and system needs will all drive requirements for higher system capacity in 5G smartphones. For example, 5G will enable gaming like you’ve never seen on a handset, but it will require more DRAM. Downloading full-length HD movies to stream to your TV will require more NAND storage. 5G will enable tomorrow’s video streaming to be more immersive, allowing you to select camera angles and other content in real time, driving the need for more DRAM. Using your 108MP+ camera and recording 8K video content to share with your friends on social media require both a large amount of DRAM to encode the data and quite a bit of NAND storage to save the data to share later.
On top of that, today’s systems use multiple AI engines for enhanced image and video capture, and this drives even more DRAM to support the system needs. Finally, the real benefit of 5G in the modern cell phone is that a user can do all these activities at once, truly multitasking. Of course, that means you’ll need even more DRAM content to use all these features in parallel. We believe that very soon Micron’s 12GB LPDDR5 DRAM coupled with 256GB UFS3.1 NAND storage will be the minimum requirement in a full-featured 5G handset.
Me: Some skeptics think that the use cases of the zero-labor factory and telesurgery don’t really need 5G. What’s your take on this?
Moore: Zero-labor factories and telesurgery are two use cases that both hold huge promise for society. Just think about the current situation today with the COVID-19 pandemic. If these were already possible on a widespread basis, we wouldn’t have had the global factory shutdowns and subsequently supply chain disruptions we’re experiencing today. And if telesurgery were already widely available, we wouldn’t have had to delay surgeries or put doctors and other patients at risk of catching the disease. 5G is absolutely critical for these use cases since they both require many connected systems and devices that can quickly communicate in real time. Just think about how dangerous a lag would be for a doctor operating a robotic arm in a critical surgery or for a factory worker operating industrial machinery.
5G is vital here because it allows many more machine-to-machine (M2M) connections for three reasons:
- Ultralow latency, much lower than 4G (1 ms vs. 10 ms)
- 10 times more connections per square kilometer (1 million vs. 100,000)
- 20 times faster data rate (20 Gb/s vs. 1 Gb/s)
Because there is more data bandwidth for more devices at low latency, there can be more autonomous coordination among intelligent devices and between infrastructure and those devices. The 1 ms latency is important because it enables real-time control across the network, eliminating video lag in operating a remote device. These elements will be crucial for enabling telesurgery and zero-factory labor.
Me: At present, augmented reality (AR) and virtual reality (VR) are the two technologies that the industry is generally optimistic about but has not yet fully realized. For AR and VR to take advantage of 5G, how do they need to be improved?
Moore: From what I’ve seen, the most popular use case for AR/VR today is gaming. Certainly, my children use the technology every day playing Pokémon Go! But with the ultralow latency that 5G provides, more enterprise applications will come. In the current environment of COVID-19, it’s clear that remote education, remote doctoring, even remote mechanics to work on one’s car are needed. Coupling these concepts with artificial intelligence is even more exciting. Imagine a four-year-old child sitting in a virtual classroom learning to read from a personalized AI-based teacher that customizes the lesson to the child. We are not far from that as a reality.
All these applications will require a lot of DRAM to process the video and AI, as well as NAND to store big data in the cloud and mobile data in the handset. The key to these applications’ success is that everything must be seamless to the end user. There cannot be glitches or lag time to make the virtual environment appear anything less than real; otherwise, the end user will not accept it. Certainly, the four-year-old child will lose interest and focus if her virtual teacher takes too much time to respond to her questions! This type of system will drive the need for faster compute, better neural engines and more DRAM memory and NAND storage than we see today.
Me: What kind of 5G applications do you think will explode in the next two years besides smartphones?
Moore: We see 5G as a disruptive technology that will influence consumers and enterprises in many ways, including mobile phones, autonomous driving, lights-out factories, AR/VR, the intelligent edge, AI and, indeed, things we haven’t even dreamed of yet. These are all important areas of focus for us, as our memory and storage products are core technologies that are suited to powering many 5G applications across the board — because, ultimately, these all rely on data.
But one area we find particularly exciting is AI. With 5G, our devices will be able to process all the data required for AI tasks instantaneously, bringing intelligence to consumers’ fingertips. Here, memory and processing requirements will evolve. AI applications will need more memory, storage and much faster speeds than they have now to accommodate the exponentially increasing amounts of data 5G will bring. That’s where Micron comes in. Micron technologies will combine the best of 3D NAND — high data storage capacity — with the most desirable features of DRAM — superfast processing — to enable AI.
But for me, what’s even more exciting is what we don’t see yet. I often think of the university student sitting in his or her dorm coding a new application that will push today’s hardware to the limit. Imagine yesterday’s college student inventing the video chat, social network, or e-tail usage model that we wouldn’t have thought possible yet we consider commonplace today.
So when you ask about 5G and the future for mobile memory and storage, I get really excited. This is why I come to work every day, to build the platforms that enable the future. What will that platform need? 1TB of UFS4.0 super high-speed storage? 24GB of LPDDR5x memory? New memory and storage technologies? Disruptive technologies like 5G drive innovation.
A version of this content first appeared as an interview with EEFocus.