Ampere推出全球首款专为云设计的处理器

近日,芯片公司Ampere(安晟培半导体科技有限公司)发布了全球首款云原生服务器--80核ARM架构64位处理器Altra。

Ampere产品高级副总裁Jeff WitTIch告诉记者,Ampere致力于打造面向云计算和边缘的新一代处理器。“我们研发的目标是开发面向于云计算的新一代云原生CPU,这里讲的“云”是指从超大规模云到分布式边缘云。与其他芯片厂商不同,我们的产品纯粹针对云计算模型进行设计,并与行业里最广泛的GPU、FPGA的供应商进行合作,能够非常灵活地为客户提供其所需要的解决方案组合。”

1、数字化进程的加速发展带来新型处理器平台的旺盛需求

由于互联网和数字经济的快速崛起,中国的数据中心进入了新一轮的建设周期,特别是云计算、大数据、人工智能等新技术的引入,加速了数字化进程。然而,随着工作负载的增加和实时计算的兴起,传统架构终将无法满足未来的需求。

Jeff表示,从软件的角度来说,现在的需求更多的是面云原生比如容器、微服务这类的计算模型快速转变,而不是面向以前那种企业级内部的计算模型。同时,能效也是很重要的一个问题,不仅影响到运营成本,且大规模部署与绿色可持续发展也是目前受到关注的全球性问题。

人工智能不断发展,机器学习使得计算能力在整个基础设施的各个地方都有非常大的需求,包括在数据中心和边缘。越来越多的异构型基础设施要求有云计算的能力,能够被定制,能够提供确定性、可预测性的性能表现,使服务实现最低延时。

“针对当下云计算的需求,需要非常高效的CPU内核,需要整个基础设施实现非常低的功耗,且具备非常好的扩展性。

2、Ampere Altra应运而生

1)实现了可预测的高性能、高扩展性和高能效

作为世界首款专为云全新设计的处理器,可以满足云计算在性能、扩展性、能效的需求。

性能方面:Jeff Wittich表示,当前数据中心消耗了世界3%的发电量,2030年可达11%。Altra比当前AMD最先进的64核EPYC 7742处理器功耗减少,性能快4%。对比Intel 28核至强可扩展(Cascade Lake)铂金处理器8280,性能更是快了1.23倍。

扩展性需求方面:针对云应用场景,比如搜索、AI推理、音视频解码、云游戏、数据库等,Ampere Altra可以满足从超大规模到边缘的扩展性,在相同功率范围下最大化数据中心容量,每一个机架能够部署3500多个Altra的内核。由于性能更好,功耗更低,所以Ampere Altra处理器的总体应用成本(TCO)也更低。

高能效方面:Ampere Altra凭借其单核单线程的内核设计将提供可预测的高性能、用户的安全隔离、全平台的灵活扩展性以及全新水平的高能效。Ampere Altra处理器具有80个内核和210W的功耗,为日新月异发展的前沿科技领域如数据分析、人工智能、数据库、数据存储、电信堆栈、边缘计算、Web主机与云原生应用等提供突破性的能效提升。Ampere Altra功耗非常低,单核功耗更低或者说每瓦性能是业内最高的。Ampere Altra具备先进的系统安全、电源管理系统,提供服务器级RAS,Arm v8.2+架构符合服务器基础架构规范(SBSA)4级,让服务器更具可管理性,云提供商可以非常简单地把Ampere Altra处理器整合到他们的云环境当中。

2)开始向云服务和边缘计算客户出样

Altra基于ARM Neoverse N1平台,台积电7nm工艺。目前已开始向云服务和边缘计算客户出样,包括微软Azure、甲骨文、联想等,服务场景包括数据分析、人工智能、数据库、存储支撑、通信堆栈、web主机等,预计2020年中旬开始规模量产。

3、未来,将不断构筑合作共赢的生态圈

针对于Ampere自身在生态方面的部署,Amper与全球众多伙伴合作,包括:IP提供商、软件生态伙伴、开源社区及开源社区项目之间建立以生态为基础的合作。

软件生态系统进展方面,Ampere支持Linux、BSD、windows等所有操作系统,以及云工具docker、KVM、vmware、kubernetes等。软件生态建设方面已与所有其他开源社区、包括软件开发商共同进行完善。

Jeff Wittich表示,我们的专家团队始终在性能、扩展性和能效等方面保持领先,一直投入进行不断的创新,并与生态伙伴一起合作共赢,从而赢得客户。

Recently, the chip company ere released the world's first cloud native server, the 80 core arm 64 bit processor Altra.

Jeff WitTIch, senior vice president of Ampere products, told reporters that Ampere is committed to building a new generation of processors for cloud computing and the edge. "The goal of our R D is to develop a new generation of cloud native CPU for cloud computing. Here," cloud "refers to the transition from ultra large scale cloud to distributed edge cloud. Different from other chip manufacturers, our products are designed purely for cloud computing model, and cooperate with the most extensive GPU and FPGA suppliers in the industry, which can provide customers with the solution combination they need very flexibly. "

1. The rapid development of digital process brings the strong demand of new processor platform

Due to the rapid rise of the Internet and digital economy, China's data center has entered a new round of construction cycle, especially the introduction of cloud computing, big data, artificial intelligence and other new technologies, accelerating the digital process. However, with the increase of workload and the rise of real-time computing, traditional architecture will not be able to meet the needs of the future.

Jeff said that from the perspective of software, the current demand is more the rapid transformation of the computing models of the cloud, such as containers and microservices, rather than the former enterprise level internal computing models. At the same time, energy efficiency is also a very important issue, which not only affects operating costs, but also large-scale deployment and green sustainable development.

With the continuous development of artificial intelligence, machine learning makes computing power have great demand in all parts of the infrastructure, including data center and edge. More and more heterogeneous infrastructure requires the ability of cloud computing, which can be customized, can provide deterministic and predictable performance, and make the service realize the minimum delay.

"For the current demand of cloud computing, it needs a very efficient CPU core, a very low power consumption of the entire infrastructure, and a very good scalability.

2. Amber Altra was born

1) Predictable performance, scalability, and energy efficiency

As the world's first processor specially designed for the cloud, it can meet the needs of cloud computing in performance, scalability and energy efficiency.

Performance: according to Jeff Wittich, data centers currently consume 3% of the world's power generation, up to 11% by 2030. Altra is 4% faster than current AMD's most advanced 64 core epyc 7742 processor. Compared with Intel 28 core platinum 8280, the performance is 1.23 times faster.

Scalability requirements: for cloud application scenarios, such as search, AI reasoning, audio and video decoding, cloud games, databases, etc., amber Altra can meet the scalability from large scale to edge, maximize the capacity of data center in the same power range, and each rack can be able to deploy more than 3500 Altra cores. Due to better performance and lower power consumption, the total application cost (TCO) of the ere Altra processor is also lower.

Energy efficiency: with its single core and single thread kernel design, ere Altra will provide predictable high performance, user safety isolation, flexible scalability of the whole platform and a new level of energy efficiency. The ere Altra processor has 80 cores and 210W power consumption, which provides breakthrough energy efficiency improvement for the fast-growing cutting-edge technology fields such as data analysis, artificial intelligence, database, data storage, telecom stack, edge computing, web host and cloud native applications. Ampere Altra has a very low power consumption, a lower single core power consumption or the highest per watt performance in the industry. Ampere Altra has advanced system security and power management system. It provides server level RAS. The arm v8.2 + Architecture conforms to the server infrastructure specification (SBSA) level 4, making the server more manageable. Cloud providers can easily integrate ere Altra processors into their cloud environment.

2) Start to sle cloud services and edge computing customers

Altra is based on arm neoverse N1 platform and TSMC 7Nm process. At present, we have begun to sle cloud services and edge computing customers, including Microsoft azure, Oracle, Lenovo, etc. the service scenarios include data analysis, artificial intelligence, database, storage support, communication stack, web host, etc. it is expected to start mass production in the middle of 2020.

3. In the future, we will continue to build an ecosystem of win-win cooperation

Aiming at the ecological deployment of amber, amber cooperates with many partners around the world, including IP providers, software ecological partners, open source communities and open source community projects to establish ecological based cooperation.

In terms of software ecosystem development, ere supports Linux, BSD, windows and other operating systems, as well as cloud tools such as docker, KVM, VMware, kubernetes, etc. In terms of software ecological construction, it has been improved together with all other open source communities, including software developers.

Jeff Wittich said that our expert team has always been leading in performance, scalability and energy efficiency, and has been engaged in continuous innovation and win-win cooperation with ecological partners to win customers.

--------------------------------------------------------------
主机参考，收集国内外VPS，VPS测评，主机测评，云服务器，虚拟主机，独立服务器，国内外服务器，高性价比建站主机相关优惠信息@zhujicankao.com
详细介绍和测评国外VPS主机,云服务器,国外服务器,国外主机的相关优惠信息,商家背景,网络带宽等等,也是目前国内最好的主机云服务器VPS参考测评资讯优惠信息分享平台

这几篇文章你可能也喜欢：

• ToToTel- 真正的黑色星期五活动。 始终先打开服务器。 满意后再付款！ 支持PAYPAL/支付宝
• CSTserver 宣传香港/美国优化线路云服务器为 2.2 美元/月，裸机为 19.9 美元/月，高防御为 30 美元/月，10G 高带宽为 39 美元/月，站点组为 105 美元/月。
• 华纳云双11超值购买，香港云主机4H4G5M，每年支付888元送1年免费.com域名
• AkileCloud美国流量型云服务器，不限流量，可选500M或1000M带宽，235元/月，DNS解锁流媒体
• zlidc（智联IDC）：韩国原生IP云服务器，35.9美元/季度，4核/4G内存/50G SSD/300M优质网络@2.5T月流量

本文由主机参考刊发，转载请注明：Ampere推出全球首款专为云设计的处理器 https://zhujicankao.com/12733.html