close
AI

ARM and x86, the decisive battle in the global semiconductor change

e3a944c8-a2c8-11e9-a667-fded1e7260dd.jpg

Recently, the global semiconductor market has been particularly unsettled. There were trade frictions before, which stirred up countless right and wrong of the semiconductor industry in the two countries; after the dispute between Japan and South Korea, it brought a shadow to the global industrial chain. It seems that all kinds of phenomena are showing the fact that in today’s era, the semiconductor industry is already a strategic resource between countries, and the words “self-controllable” are not general. In this case, everyone will inevitably think more. For example, everyone knows that the core of the enterprise network is the server. But today more than 90% of the server chip market in the world is occupied by Intel, built on Intel’s x86 architecture. If a large-scale semiconductor blockade occurs and the chip supply chain is cut off, the data center, which is the cornerstone of the social economy, will be in a very dangerous situation. At present, the main solution to this problem is based on the ARM architecture, and self-developed server chips as a substitute. Although ARM is also a European and American company, it provides chip patents and instruction sets. Chip manufacturers can always use a version of a buyout, do not worry about the problem of direct out of stock upstream. The idea of ​​replacing x86 with ARM on the server chip is not something that is today. However, although ARM has dominated the mobile chip today, it has not gotten its momentum in the server chip market. Today, this task begins to play a leading role in the Chinese chip industry. Why is the ARM+ server chip in a “failed battle” situation? Under the premise that the global semiconductor market is generally tightened, can ARM chips be able to rise? This will be a question that needs to be answered step by step in the semiconductor industry in China and the world for a long time. I want to take the x86 down, it is really difficult since the birth of 1978, today’s x86 architecture in the CPU field, in fact, has also experienced decades of killing battles. The numerous disputes between Intel and AMD have finally created a solid compatibility and application ecology for this common technology architecture. However, when Intel firmly holds more than 90% of the market share, it is harder for the global computing and networking industries to have the right to choose and bargain. So the movement to push Intel into the chip altar has never stopped. Relying on low power consumption and high cost performance, it has proved its ARM in mobile devices. These years have become a new choice for the semiconductor industry. However, in fact, a large number of semiconductor giants have laid out ARM to build server chips, and the results are not satisfactory. For example, Intel’s old rival AMD has invested heavily in the development of ARM-based server chips. In 2012, AMD announced the acquisition of the ARM instruction set license, followed by the launch of the ARM-based Opteron processor in 2014. The result is not long. In 2017, this project basically entered the stage of surviving. Intel continues to occupy the throne in the field of server chips. Other players in the industry, such as Samsung, NVIDIA, Broadcom, and Marvell, have developed or intentionally developed ARM-based server chips, but in the end most of them are gone. One of the things that once made the industry see hope is Qualcomm, which has benefited a lot from ARM in the field of mobile phone chips. In 2014, Qualcomm announced that it will build an ARM-based server chip; by 2016, it announced the first AMR-based server chip product; in 2017, Qualcomm officially released the Centrip 2400 series to enter the server chip market. The result was released at the peak, Qualcomm did not get customer response, and by 2018 it was repeatedly abandoned to abandon the ARM server chip business, and then exposed executives to leave, the team disbanded. To this day, few people believe that Qualcomm will still have a role in the field of ARM server chips. The so-called Feng Shui turns, after the European and American giants generally gave up this project. Today, the main force of building server chips with ARM has become a Chinese company. For example, once famous, it also caused a lot of controversy between Huaxintong and Tianjin Feiteng. Especially noteworthy is that Huawei released the ARM-based server chip Kapeng 920 at the end of last year. The Chinese story of the ARM server chip is still in progress today. It seems necessary to explain here, why x86 can be so hit, so that the European and American giants who brought in the group have challenged the failure.行业普遍认可的说法是,x86最强大的地方在于英特尔积累的服务器芯片生态系统。若干年来,大量软件应用都是基于x86完成开发,而企业市场中软件生态往往更新频次较低,缺乏生态可选择性。而市面上出现过的ARM服务器芯片,普遍在软件兼容性上表现不佳,最终导致用户无法选择尝试ARM,只能固守x86。此外,产业体系的不完成,也是AMR生态的问题。在服务器之下,数据中心涉及计算、存储、网络、数据库等大量底层技术体系,这些东西很难一夜之间被全部替换。然而逐个替换的成本实在太大,企业无法接受舍弃自身核心业务去拥抱新的服务器产品。也有人认为,ARM服务器芯片的核心问题还是自身优势没有被发挥完整。一方面性能依旧长期不及x86,另一方面价格也没有便宜到令市场怦然心动的地步。有一种说法,是只要有20%的价格差距,企业就会选择更换服务器。但ARM生态还没有触发这个临界点。不抛弃,不放弃:ARM的魅力到底在哪?这个话题中令人纠结的一部分就要来了:既然英特尔的x86听上去天下无敌,众高手纷纷被挑落马下,为什么还是不断有企业去寻求在ARM架构上打造服务器芯片呢?这也是一个多方面共同推动的因素。但是核心问题肯定还是开头所说的,自主可控与产业安全问题。今天的全球半导体贸易格局中,处处弥漫着摩擦可能升级的味道。而如果一旦英特尔的芯片被切断供应,数据中心这个国之重器将陷入极大的困境。这个情况下, 2018规模已经达到18000亿人民币以上的中国半导体市场,是一定要找到可替代方案的。而在自主可控和避免重新造轮子之间,ARM是一个切合实际的解决方案。由于ARM的商业模式是IP售卖,可以从根本上规避x86隐藏的专利与供应链风险,那么难点就集中在了芯片性能与生态上。由于目前半导体产业在服务器芯片领域普遍缺乏议价能力,所以行业也是乐见ARM生态成长起来的。此外,ARM进入服务器生态,依旧携带着其最根本的功耗优势与成本优势。功耗能低,成本更低的ARM,相对更符合云时代的企业计算需求。还要看到的是,计算产业本身发生的迭代,也在让ARM似乎看起来更像是符合未来发展趋势的那个选择。比如在企业全云时代,对服务器计算需求将会大幅增加,产业普遍要求高并发、多线程和任务单一的计算模式,这些特性都非常适合ARM架构来处理。另一个趋势,是计算正在加强云边端的一体化与协调。而当更多计算发生在边端侧的时候,就要求服务器提供密度更高、功耗更低、迟延更低的算力,这其实是让服务器算力贴近移动设备的计算需求,因此ARM也将是个好的选择。所以说,AMR虽然屡次失败,但根本上来看并非一无是处。前人的不成功,也是为后来者馈赠的机会。在今天想要打赢ARM服务器芯片这一局,IT行业还有很多事要做。ARM想要一战成功,需要做好哪些事?总的来说,想要让ARM芯片真正在服务器市场站稳脚跟,需要的就是最大化发挥AMR优势,并且吸取前面那些巨头玩家的失败教训。有这样几个工作,属于今天的题中之义:1、依靠产业链内部完成垂直创新,构筑ARM的产业能用性。ARM服务器芯片的问题,很大一部分并不集中在性能与价格上,而是集中在是否真正能满足产业应用需求,达到替换标准线。这也就意味着,ARM服务器芯片不应该是一个单打独斗的产品,而是与生俱来应该就是一个底层应用序列,满足服务器场景中数据、存储、调用等基础应用,降低客户的兼容成本。2、发挥性价比是核心优势。On the ARM server chip, the vast majority of failure cases are considered to have not played the role of ARM chip low power, low cost, and performance is not weaker than x86. In the core cost-effective, energy-consuming ratio field, ARM server chips still have a long way to go, and technological breakthroughs are still something that can be expected. 3. The core task is to drive the industry and establish an ARM ecosystem. Today, when it comes to ARM server chips, everyone will mention the importance of ecology. How to deal with the unified hardware, development and infrastructure of the x86 platform, as well as the software development ecology accumulated over decades, is a real challenge for chip manufacturers and a problem that must be developed for a long time. On the one hand, chip companies create infrastructure clusters for the development of ecological services, such as storage networks and virtualization, to reduce the development costs of enterprises; at the same time, they also need to actively promote the software development industry to enter the ARM ecosystem and build the ARM application market and other basic platforms. In addition, the development of ARM in the new demand is also a variable, for example, to promote the artificial intelligence supply chain enterprises more ARM, to guide more application users to naturally transition to the AMR ecosystem along the AI ​​needs. 4. The supply chain safety card is a commercial bolus. Today, ARM server chips are related to industrial security and even international economic security. This is basically a matter of no dispute. Under this circumstance, facing the existence of risks and guiding the industry to actively seek for a safe and controllable industrial path is also an important channel to promote the ARM ecosystem. On the whole, the AMR server chip is an important card in the global semiconductor change today. But who is going to fight this card, how to play it, to what extent, or a very complicated test. In any case, the change has come, and it is not realistic to expect to maintain the status quo. Embracing and manipulating change is a new text that China’s semiconductor industry must learn today. .

Tags : Internet entrepreneurshipInternet entrepreneurship projectStart a business