<ul>
<li><span><span>Arm</span></span><span><span>推出业界首款搭配</span></span><span><span>Cortex-M</span></span><span><span>使用的微神经网络处理器</span></span><span><span>(microNPU)</span></span><span><span>，搭配最新的</span></span><span><span>Cortex-M</span></span><span><span>处理器，可使机器学习性能提升</span></span><span><span>480</span></span><span><span>倍</span></span></li>
<li><span><span>全新的</span></span><span><span>IP</span></span><span><span>可让无处不在的数十亿个下一代更小型、低功耗的设备具备更强的机器学习处理能力</span></span></li>
<li><span><span>软件开发流程针对提高开发人员的生产力进行了优化，并针对这些全新处理器的各种配置将机器学习性能最大化</span></span></li>
</ul>
<p><span><strong>(2020</strong></span><span><strong>年</strong></span><span><strong>2</strong></span><span><strong>月</strong></span><span><strong>1</strong></span><span><strong>1日</strong></span><span><strong>)</strong></span><span><span><span> </span>Arm</span></span><span><span>今天宣布其人工智能</span></span><span><span>(AI)</span></span><span><span>平台新增重要生力军，包括全新机器学习</span></span><span><span>(ML) IP</span></span><span><span>：</span></span><span><span><span>Arm® Cortex®-M55</span></span></span><span><span>处理器和</span></span><span><span>Arm Ethos™-U55</span></span><span><span>神经网络处理器</span></span><span><span>(NPU)</span></span><span><span>，后者也是针对</span></span><span><span>Cortex-M</span></span><span><span>平台推出的业界首款微神经网络处理器</span></span><span><span>(microNPU)</span></span><span><span>。这样的设计</span></span><span><span>(Cortex-M55</span></span><span><span>结合</span></span><span><span>Ethos-U55)</span></span><span><span>为微控制器带来</span></span><span><span>480</span></span><span><span>倍的机器学习性能飞跃。全新的</span></span><span><span>IP</span></span><span><span>与搭配的开发工具，能够让</span></span><span><span>AI</span></span><span><span>硬件与软件开发人员以更多的方式进行创新，从而为数十亿个小型、低功耗的物联网与嵌入式设备带来前所未有的终端机器学习处理能力。</span></span></p>
<p><span><span>Arm<span> </span></span></span><span><span>资深副总裁暨车用与物联网事业部总经理</span></span><span><span>Dipti Vachani</span></span><span><span>表示：“要让</span></span><span><span>AI</span></span><span><span>无所不在，设备制造商与开发人员必须在数十亿、乃至数万亿个设备上实现终端机器学习能力。我们的</span></span><span><span>AI</span></span><span><span>平台增添这些生力军后，即便在最小的设备上，终端机器学习也将成为新常态，不会再有遗珠之憾，从而让</span></span><span><span>AI</span></span><span><span>的潜力可以在范围宽广的、正在改变我们生活的应用当中得以安全、有效地发挥。”</span></span></p>
<p><span><span>随着物联网的发展与</span></span><span><span>AI</span></span><span><span>和</span></span><span><span>5G</span></span><span><span>的发展相互交织，更多的终端智能意味着更小型且成本敏感的设备会愈来愈智能、功能也愈来愈强，同时因为对云端与网络的依赖较小，也将具备更高的隐私性与可靠度。</span></span><span><span>Arm</span></span><span><span>通过新的设计为微处理器带来安全的智能，为想要有效提升终端数字信号处理</span></span><span><span>(DSP)</span></span><span><span>与机器学习能力</span></span><span><span>(ML)</span></span><span><span>的产品制造商降低芯片与开发成本，同时加快他们的产品上市速度。</span></span></p>
<p><span><strong>Arm Cortex-M55: Arm</strong></span><span><strong>人工智能能力最强大的</strong></span><span><strong>Cortex-M</strong></span><span><strong>处理器</strong></span></p>
<p><span><span>Cortex-M</span></span><span><span>处理器已经成为开发人员运算平台的最佳选择，</span></span><span><span>Arm</span></span><span><span>的合作伙伴已经出货超过</span></span><span><span>500</span></span><span><span>亿片基于</span></span><span><span>Cortex-M</span></span><span><span>的芯片，用于各种客户应用。新增的</span></span><span><span>Cortex-M55</span></span><span><span>是</span></span><span><span>Arm</span></span><span><span>历来</span></span><span><span>AI</span></span><span><span>能力最为强大的</span></span><span><span>Cortex-M</span></span><span><span>处理器，也是首款基于</span></span><span><span>Armv8.1-M</span></span><span><span>架构、内建</span></span><span><span>Arm</span></span><a href="https://www.arm.com/company/news/2019/02/next-generation-armv8-1-m-arch… href="https://www.arm.com/company/news/2019/02/next-generation-armv8-1-m-arch…，它可以大幅提升</span></span><span><span>DSP</span></span><span><span>与</span></span><span><span>ML</span></span><span><span>的性能，同时更省电。与前几代的</span></span><span><span>Cortex-M</span></span><span><span>处理器相比，</span></span><span><span>Cortex-M55</span></span><span><span>的</span></span><span><span>ML</span></span><span><span>性能最高可提升</span></span><span><span>15</span></span><span><span>倍，而</span></span><span><span>DSP</span></span><span><span>性能也可提升</span></span><span><span>5</span></span><span><span>倍，且具备更佳的能耗比。</span></span></p>
<p><span><span>此外，客户也可以使用</span></span><span><span>Arm Custom Instructions (Arm</span></span><span><span>自定义指令集</span></span><span><span>)</span></span><span><span>延伸处理器的能力，对特定工作负载进行优化，而这也是</span></span><span><span>Cortex-M</span></span><span><span>处理器的全新功能。</span></span></p>
<p><span><strong>Ethos-U55: Arm</strong></span><span><strong>针对</strong></span><span><strong>Cortex-M</strong></span><span><strong>推出的首款</strong></span><span><strong>microNPU</strong></span></p>
<p><span><span>针对需求更高的</span></span><span><span>ML</span></span><span><span>系统，可将</span></span><span><span>Cortex-M55</span></span><span><span>与</span></span><span><span>Ethos-U55</span></span><span><span>搭配，后者是</span></span><span><span>Arm</span></span><span><span>首款微神</span></span><span><span>经</span></span><span><span>处理器</span></span><span><span>(microNPU)</span></span><span><span>，与现有的</span></span><span><span>Cortex-M</span></span><span><span>处理器相比，两者结合后可使产品</span></span><span><span>ML</span></span><span><span>性能提升</span></span><span><span>480</span></span><span><span>倍。</span></span></p>
<p><span><span>Ethos-U55</span></span><span><span>具有高度的可配置性，旨在加速空间受限的嵌入式与物联网设备的</span></span><span><span>ML</span></span><span><span>推理能力。它先进的压缩技术可以显著节省电力并缩小</span></span><span><span>ML</span></span><span><span>模型尺寸，以便运作之前只能在较大型系统上执行的神经网络运算。</span></span></p>
<p><span><strong>简化的软件使得所有开发人员都能获取安全的</strong></span><span><strong>AI</strong></span><span><strong>功能</strong></span></p>
<p><span><span>Arm</span></span><span><span>非常清楚，开发人员的经验对于推动</span></span><span><span>AI</span></span><span><span>革命是至关重要的。因此，</span></span><span><span>Cortex-M55</span></span><span><span>与</span></span><span><span>Ethos-U55</span></span><span><span>得到了</span></span><span><span>Arm</span></span><span><span>业界领先的</span></span><span><span>Cortex-M</span></span><span><span>软件开发工具链的全力支持。如此一来，我们针对传统</span></span><span><span>DSP</span></span><span><span>与</span></span><span><span>ML</span></span><span><span>等工作负载，有了一致的开发流程；同时从</span></span><span><span>TensorFlow Lite Micro</span></span><span><span>开始，针对先进机器学习框架进行特定的整合与优化，确保开发人员拥有无缝的开发体验，并能够在任何一种</span></span><span><span>Cortex-M</span></span><span><span>与</span></span><span><span>Ethos-U55</span></span><span><span>的配置上，获取最佳性能。</span></span></p>
<p><span><span>Arm</span></span><span><span>认为安全不应该是事后弥补措施，它对于物联网的普及极为重要。为了确保最安全的设计并顺利通过</span></span><span><span>PSA</span></span><span><span>认证，这些处理器和搭配的</span></span><span><span>Corstone</span></span><span><span>参考设计都可以与</span></span><span><span>Arm TrustZone</span></span><span><span>一起工作，以确保安全性可以更为简易地整合到完整的片上系统中。</span></span></p>
<p><span><strong>赋能生态系统</strong></span></p>
<p><span><span>全新的</span></span><span><span>Cortex-M CPU</span></span><span><span>与</span></span><span><span>Ethos microNPU</span></span><span><span>凸显了</span></span><span><span>Arm</span></span><span><span>对推动物联网发展的承诺：致力于协助芯片设计人员与设备制造商，即便在最小的终端设备上，也能利用</span></span><span><span>Arm</span></span><span><span>架构进行创新，并为物联网点燃新一波创造力与创新性。这一技术正获得生态系统合作伙伴与业界的广泛支持，其中包括</span></span><span><span>亚马逊</span></span><span><span>(Amazon)</span></span><span><span>、</span></span><span><span>Alif</span></span><span><span>半导体</span></span><span><span>(Alif Semiconductor)</span></span><span><span>、恒玄科技</span></span><span><span>(Bestechnic)</span></span><span><span>、赛普拉斯</span></span><span><span>(Cypress)</span></span><span><span>、杜比</span></span><span><span>(Dolby)</span></span><span><span>、</span></span><span><span>Google</span></span><span><span>、恩智浦半导体</span></span><span><span>(NXP)</span></span><span><span>、三星</span></span><span><span>(Samsung)</span></span><span><span>、意法半导体</span></span><span><span>(STMicroelectronics)</span></span><span><span>等。</span></span></p>
<p><span><span>您可以通过</span></span><span><span>Arm</span></span><span><span>资深副总裁暨车用与物联网事业部总经理</span></span><span><span>Dipti Vachani</span></span><span><span>的</span></span><a href="https://www.arm.com/blogs/blueprint/ai-for-iot-devices"><span><span>这篇博…;，了解此新技术如何带给数十亿设备</span></span><span><span>AI</span></span><span><span>体验。</span></span></p>
<p><span><span>您也可以点击以下链接，获取更多有关</span></span><a href="https://www.arm.com/products/silicon-ip-cpu/cortex-m/cortex-m55"><span>… href="https://www.arm.com/products/silicon-ip-cpu/machine-learning/ethos-u55"…。</span></span></p>
<p><strong>关于</strong><strong>Arm</strong></p>
<p><span>作为计算及</span><span>互联革命的核心</span><span>，</span><span>Arm</span><span>技术正改变着人们生活及企业运行的方式。</span><span>Arm</span><span>先进的低功耗处理器设计已应用于超过</span><span>1,500</span><span>亿颗芯片的智能计算，</span><span>Arm</span><span>的技术安全地为电子设备提供支持，覆盖从传感器到智能手机乃至超级计算的多样化应用。结合我们的物联网设备、连接与数据管理平台，</span><span>Arm</span><span>技术帮助客户从互联设备中获得强大且可执行的洞察，从而赋能真正的商业价值。</span><span>Arm</span><span>携手超过</span><span>1,000</span><span>家技术合作伙伴，为从芯片到云端的整个计算领域提供设计、安全和管理方面的技术，并在该技术领域处于业界领先地位。</span></p>