登录/注册
汽车是半导体行业中增长最快的领域之一. 到2025年， 将售出9700万辆汽车 每一个都有大约90gb的RAM + NAND. 软件 footprints are expected to expand soon from 100 million lines of code in today’s high-end vehicle to 1 billion lines — 比阿波罗11号多1000倍.
因此，难怪在最近美光委托的报告中， “比特、字节和联网汽车，” S&P Global Mobility估计，汽车内存和存储业务的收入将增长到11美元.到2026年将达到60亿美元，复合年增长率(CAGR)为20%至24%.
随着汽车增加数据丰富的功能，如车载信息娱乐(IVI)，, 自治, 订阅服务及更多, we can expect memory and storage requirements to increase by nearly 10 times over the next seven years. 以下是该报告的一些亮点:
- 今天，平均每辆车使用90GB的RAM + NAND. By 2026, 平均每辆车将使用大约278GB的DRAM和NAND, 在高端汽车上，这个数字更接近2tb. 到2030年，这一容量预计将增长到4TB.
- 联网汽车是当今最复杂的软件驱动机器, 8000万到1亿行代码，2.5亿到3亿行用于自动驾驶.
- 内存 capacity will increase ~445% from over 385 million GB (~385,000 TB) by the end of 2022 to 2.10亿GB (~2,100,2026年先进驾驶辅助系统(ADAS), 电子控制单元(ecu)和模数转换器.
- Roughly 73 million new passenger vehicles will feature an 信息娱乐 head unit by the end of 2023. 到2026年，这一数字将增加到8200多万.
- IVI memory and storage content is expected to increase from 6 billion GB of memory and storage by the end of 2022 to 7.到2026年将达到20亿GB.3%的复合年增长率.
- Approximately three million new vehicles will be equipped with cockpit domain controllers this year, 膨胀到16.到2026年将达到500万.
很明显, 随着汽车变得更加智能和软件驱动, 内存和存储是实现面向未来的创新所需的基石. 的年代&报告还指出，今天正在发生的变化. 从历史上看, 汽车利用成熟的半导体技术可能早在十年前就开始了, 但是今天, 由于汽车创新的速度很快, state-of-the-art multicore system-on-chip (SoC) processors are being designed for vehicles that will be on the road within the next two years.
那么，我们所知道的重塑汽车的主要趋势是什么呢? 总的来说，这些趋势可以概括为以下几个大趋势:
大趋势1:自主化
汽车工程师协会将自动驾驶定义为六个级别, 从0级到5级, L0没有主动辅助系统，L5可以自动驾驶. By 2030, 预计将有近300万辆汽车完全自动驾驶, and more than 15 million will support L2+/L3 自治 — which already demands significantly more memory and storage.
最近, 在微米, 我们越来越多地听到生成式人工智能在自动驾驶汽车中的应用, where AI models are speeding up the car’s learning curve with tighter optimization loops around the intelligence of vehicle-response systems. 事实上，我们已经看到了 像Waze这样的导航系统使用生成式人工智能 实时提供高度个性化的路线推荐. 在美光与客户的谈话中, we’ve observed keen interest in understanding how much memory and storage will be needed to support voice processing, natural 语言 processing and massive 语言 models — and how to move data from memory and storage to chipsets and processors more quickly.
生成式人工智能正在改变汽车行业 目前的市场规模为3.12亿美元，并将增长到2亿美元.到2032年将达到70亿. 随着它扩展到整个车辆的应用, 对更多更快的内存和存储的需求将呈指数级增长. 来启动他们的人工智能创新, OEMs will need to ensure vehicles are equipped with the bandwidth and capacities that provide AI models the headroom and horsepower to grow. 作为这些变化的一部分, it’s critical to have memory and storage that are both high performance and low power to support energy-guzzling AI. And that’s just the tip of the iceberg: 汽车 memory will also require rigorous functional safety testing and certification to ensure that autonomous vehicles on the road are safe.
大趋势2:连通性
如今，互联技术已经在汽车中广泛应用. Think about how the ability to connect our phones to our cars’ system has helped streamline the driving experience. 在未来, connectivity will also enable novel technologies such as vehicle-to-vehicle and vehicle-to-infrastructure communication, 允许汽车交换速度信息, 方向, 位置, 通过点对点网状网络的转向意图等因素. These capabilities will help vehicles better coordinate with each other and smart infrastructure to avoid collisions and to optimize traffic flow efficiently.
Most new car models also offer connectivity with over-the-air (OTA) updates to improve functionality, 解决软件问题，上传诊断数据. 虽然OTA更新可能看起来微不足道, 它们通过允许汽车制造商添加新的软件更新来推动汽车创新, 功能，甚至安全补丁都超过了购买汽车的时间. 事实上, OTA updates are foundational to enabling the new automotive subscription business models we’re seeing take flight with the rise of mobility-as-a-service as they allow software to enable and disable features in the car at a hardware level as needed. 想象一下，一个司机在去滑雪的路上增加了加热座椅作为一项服务, or parents lending their teenagers their cars and enabling 位置 services to check that they are headed home before curfew. These features unlock more subscription revenue for OEMs and allow more flexibility for consumers to add services as needed or disable them when not.
面向未来的汽车硬件，并为OTA更新带来的新功能做好准备, oem厂商应该考虑选择具有高容量和更强大芯片组的存储器. 这样做将确保这一点, 因为在汽车的整个生命周期中，新的服务都会被激活, critical automotive hardware has adequate headroom to scale with these growing data requirements.
大趋势3:分区架构
随着这些趋势的兴起, we’re seeing a huge shift in cars from simple subsystems coordinating with different domains to a highly matrixed system with centralized decision-making.
Zonal architectures feature systems grouped by physical zones and located near the ECUs that are controlled by them. This approach is much more efficient than the traditional domain approach where systems are grouped by function (for example, 信息娱乐, 发动机和变速器控制, 和其他人). 尽管这种方法曾经奏效, 随着汽车变得越来越复杂(传感器越来越多), cpu, 摄影机及系统), 领域架构需要更复杂的连接. 区域架构简化了所有这些电子设备的连接方式, 减少所需的布线数量，从而降低成本和重量. 结果是更好的里程数和洞察汽车的时间.
随着汽车走向中央决策, we’re seeing that the memory itself actually becomes more complicated as it needs to multitask to handle so many different systems. 在这里，我们可能最终会看到新的内存标准出现, 例如服务于不同域的虚拟化环境. 在接下来的几年里(想想2025-2028年), 我们可能会看到汽车走向完全的中心化, 这将需要更强大的内存和存储解决方案. The ecosystem may eventually need to consider high-bandwidth solutions such as Compute Express Link or High-Bandwidth 内存 — once reserved for supercomputing — to power all this intelligence on the go and at the edge.
这些只是美光看到的几个大趋势 其他如强化小屋和电气化.
推动汽车创新
虽然曾经处于创新的前沿, the automotive industry has flipped the script and is driving innovation that has the potential to influence other technology sectors. 很明显 the industry is going through one of the most significant transformations since the invention of vehicles themselves.
随着汽车越来越多地被智能等特征所定义, 个性化和自主性而不是制造, 型号和扭矩, 汽车生态系统需要通过确保正确的架构来跟上步伐, 内存和存储将安全、大规模地推动这一创新. 汽车 manufacturers can stay ahead of the curve by directly collaborating with semiconductor suppliers to better understand and meet evolving memory and storage requirements and, 最终, 到面向未来的数据密集型架构, 颠覆性的汽车技术即将到来. Laying this hardware foundation with the right building blocks will unlock more possibilities for innovation and new business models as automotive players reinvent themselves in this new era of auto.
