ARM架构服务器之路仍漫长崎岖

重大颠覆性的机会并不常有，它不会没事从天上掉下来。也别期待ARM与英特尔(Intel)之间很快会展开什么大规模的激烈竞争。例如，在迈向基于ARM架构服务器的这条发展道路，目前看来比原先预期的更漫长也更险峻。 两年前，ARM与其合作伙伴擘划了大举进军云端运算市场的愿景，然而，在这个领域中，英特尔以其 Xeon处理器持据占据主导地位。 Applied Micro公司在2011年ARM Tech上大张旗鼓地推出其基于64位ARM核心的XGene系列SoC。不久之后，从AMD到三星等一连串ARM合作伙伴均纷纷计划或悄悄地加速成立其ARM服务器设计团队。 惠普公司(HP)日前表示该公司已经推出分别配置Applied、Calxeda与德州仪器(TI)基于ARM架构芯片的服务器主板，预计将在明年出货。而惠普的竞争对手戴尔(Dell)也积极打造ARM架构服务器原型，使得这一处境更加晦暗不明。 戴尔平台技术与架构资深总监Robert Hormuth指出，基于ARM架构的服务器“可望在2014年年底或2015年上市”。 戴尔和惠普都表示，基于ARM架构的系统将更集中于储存等专用市场，而非主流的服务器市场。两家公司已经出货采用Avoton的低功耗服务器了。Avoton是今年初才推出的英特尔第二代Atom服务器SoC。 然而，目前ARM所掌握的最新时机显示，在ARM系统推出以前，英特尔将会先行布局新一代Atom服务器SoC。也就是说，ARM架构中先进的产品如Applied的XGene可能必须与英特尔采用类似先进外围模块与14nm制程制造的组件相互竞争。Applied的XGene采用的是40nm 制程。 对于ARM阵营来说，更糟的是英特尔已经透露一系列采用低功耗服务器SoC的计划了，那不再是Atom，而是搭载其x86核心的最新一代Haswell微架构SoC。有些分师已经表示，这款Xeon级英特尔Broawell芯片很可能让许多着眼于ARM服务器市场的公司计划败退。 这或许就能解释Nvidia公司(十几家ARM服务器SoC的支持厂商之一)为什么不再谈2011年发布的“丹佛计划”(Project Denver)了。同样地，华为(Huawei)、高通(Qualcomm)与三星(Samsung)等公司悄悄地增强其ARM服务器团队，但却对相关细节保持沉默。 博通公司(Broadcom)最近宣布其自定义的64位ARM核心与SoC。然而，就像Cavium对于其ARM架构的Thunder芯片计划一样，这些公司据称多半瞄准于通讯设备，而非主流的服务器领域──至少目前看来如此。 因此，ARM看来不太可能在很短的时间找到什么令人刮目相看的方式来抢占英特尔的云端领域了。不过，对于ARM来说，好消息是英特尔也未能在移动领域展现任何具有突破性机会的迹象。 颠覆性的绝佳机会说起来十分美好，但难以实现。 本文授权编译自EE Times，版权所有，谢绝转载 编译：Susan Hong 参考英文原文：ARM Servers Face Long, Winding Road，by Rick Merritt

相关阅读：
• AMD发布X86/ARM架构混搭处理器，目标笔电和平板
• 全球75%企业高管关注物联网新机遇
• ARM阵营从可穿戴设备到服务器通吃7bfesmc

{pagination} ARM Servers Face Long, Winding Road Rick Merritt, SiliconValley Bureau Chief Big, disruptive opportunities don't come along very often. Don't expect one anytime soon in the mega battle between ARM and Intel. For example, the road to ARM-based servers now looks longer and narrower than previously thought. Two years ago, ARM and its partners had visions of significantly encroaching on the cloud computing space Intel dominates with its Xeon processor Applied Micro launched to much fanfare its XGeneSoC based on a custom 64-bit ARM core at ARM Tech Con in 2011. Soon after, a laundry list of other ARM partners, from AMD to Samsung, tipped plans or quietly ramped up ARM server design teams. This week, Hewlett-Packard said it has server boards using ARM-based chips from Applied, Calxeda, and Texas Instruments running in the lab and will ship them next year. But HP rival Dell, which has also been prototyping ARM-based servers, added a little darker color to the situation. The ARM server launches "feel like the end of 2014 or early 2015," said Robert Hormuth, who is leading the investigations at Dell. Both Dell and HP say the ARM-based systems will target relatively narrowly focused markets such as storage rather than mainstream servers. Both companies already ship low-power servers using Avoton, Intel's second-generation Atom-based server SoClaunched earlier this year. The updated ARM timing means Intel will be able to field a whole new generation of the Atom server SoCs before ARM systems launch. That means leading-edge ARM products such as Applied'sXGene may have to compete with Intel parts using similarly advanced peripheral blocks and made in a 14nm process. Applied uses 40nm for its XGene. What's worse for the ARM camp is Intel has telegraphed plans for a low-power server SoC using, not Atom, but the new Haswell generation of its full blown x86 core. Some analysts are already saying the Xeon-class Intel Broadwell chip could sink the plans of many companies that had eyed a significant market for ARM-based servers. That might explain why Nvidia, one of a dozen ARM server SoC wannabes, has gone quiet about its Project Denver plans announced in January 2011. Similarly, Huawei, Qualcomm, and Samsung quietly ramped up ARM server teams but then kept mum about them. Broadcom recently announced its plans for a custom 64-bit ARM core and SoCs based on it. But like rival Cavium with its plans for an ARM-based Thunder chip, the companies are said to target communications gear as much as, if not more than, mainstream servers -- at least for now. So ARM is not likely to encroach on Intel's cloud territory in any significant way soon. The good news for ARM is Intel is not showing signs of a breakout opportunity in mobile either. Disruption is wonderful to talk about, but hard to do.