Marvell：把根留在硅谷，到中国开花结果

美国无晶圆厂(fabless)公司美满电子 (Marvell Technology) 并没有完全迁移到中国的打算。不过，这家美国公司高层却表示，该公司的目标，是要成为“中国最大的半导体公司”。Marvell 在中国的员工人数已达1,900人，且仍在持续增加中，该公司明确承诺了在中国的发展目标，并表示，已经有一些目标已经陆续实现了。 今年稍早，《EE Times》曾专访Marvell的共同创办人戴伟立(Weili Dai)。出生于上海的戴伟立表示：“根留硅谷这一点对我们来说非常重要。因为这里有包括人才和技术在内的完整生态系统。” 然而，这些都无法阻挡Marvell跨向太平洋的那一端。

Marvell的共同创办人戴伟立(Weili Dai)
Ot7esmc

除了该公司在上海的设计/工程中心聘用的1,600名员工以外，Marvell也正准备在南京拓展第二个据点。Marvell并未透露太多细节，但承认正在与当地政府合作。新据点的规格预计会很庞大。 该公司解释了对中国的投资策略。“中国的技术发展非常快速，这个市场非常有活力，”Marvell中国区总经理李廷伟说。“大家都知道中国代表着巨大的机遇，但要真正将其转换成商机，你就必须亲自来到这里。” 这包括了Marvell与诸多本地领先的电信和电视服务供货商在系统开发和服务方面所进行的策略合作。Wei Li解释道，芯片公司必须与各地的生态系统密切合作，反应必须非常快速，而且必须要与当地服务结合。 他认为，Marvell目前正处于绝佳位置，同时具备美国硅谷公司的技术领先，以及在中国本地半导体公司能快速针对当地需求做出反应的优势。 李廷伟指出，在中国，一个很大的特色是他们缺乏应用文件的概念。“他们几乎不使用文件记录。”他表示，设计师和工程师必须参加许多会议去做演示文稿，并仔细聆听决策者说了些什么，这很重要，因为工程师和设计师必须要知道客户提出哪些新需求。“你必须随时与客户保持联系，”他说。 800工程师专攻移动应用 今天的Marvell在中国有着许多不同产品线。其中包括智能电视；机顶盒(STB)；通用被动光网络(UPON)技术──Marvell声称这是全球唯一支持EPON、GPON、主动式以太网络等三种光纤媒体协议的单芯片解决方案；以及行动芯片。 在所有产品线中，行动部门是Marvell目前在中国最庞大的业务，共有800名工程师致力于开发手机产品。 尽管Marvell的手机芯片业务曾依靠 Research In Mothion (RIM)的黑莓机(Blackberry)而大放共彩，但该公司的行动业务却也随着RIM在全球智能手机市场的失势而下滑。 然而，Mavell声称，该公司已经为再度争取行动市场占有率准备好足够的弹 药。而让该公司有信心做出此一承诺的关键，是Marvell在中国、美国和以色列的800名、200名以及500名工程师。 以下是Marvell移动产品副总裁李春潮(Ivan Lee)接受《EE Times》的专访记录。李春潮主要负责Marvell在中国和美国的开发业务。 问：为何Marvell在各个不同地点设置行动开发团队？各地团队之间如何分工？ 李春潮：这么做是有历史因素的。我们在以色列的500人主要来自2006年收购自英特尔的X-scale开发团队。这个团队的开发重点3G和WCDMA。Marvell之所以能赢得RIM的Blackberry设计订单，便是因为以色列团队的努力。 问：那么，中国和美国的移动开发团队扮演什么样的角色？ 李： 在中国，我们的重点一直在四年前开始的TD-SCDMA；现在则专注于TD-LTE的发展。我们在中国的行动产品开发团队是由硅芯片设计师、DSP设计 师、协议堆栈开发人员，以及负责开发Android和其它上层软件的设计师所组成。我们在美国有200人，包括在德州奥斯汀的SoC设计团队、加州的团 队、麻州的Android软件工程师，以及亚利桑那州的硅验证团队。 问：你说Marvell在中国有800人负责行动产品的研发。但我发现，许多人使用的名词定义有很大不同。你如何定义研发？ 李：这个团队负责从TD-SCDMA的概念到架构开发到量产的所有环节。在我所说的800人之中，有99.9%从事研发工作。 本文授权编译自EE Times，版权所有，谢绝转载 本文下一页：押注在LTE

相关阅读：
• Marvell、联想和中国移动共同推动中国3G TD 智能手机
• 三星GALAXY Xcover安卓智能手机采用Marvell WCDMA解决方案
• 三星Android手机S5820采用Marvell 3G TD解决方案Ot7esmc

{pagination} Marvell的LTE赌注 问：直到2008年以前，Marvell都还未进入中国行动市场。你为何不去购买一些已经开发出TD-SCDMA芯片，或是去并购握有相关IP的公司？ 李：事实上，我们观察过所有能提供IP的公司以及潜在的收购目标。但最终我们认为，如果自己来，我们可以做得更好。 当然，当时很多人说我们太晚进入市场。他们还说我们应该去买一些IP，我们应该使用成熟的协议堆栈。 问：回到2008年，当你决定加入市场时，当时有哪些公司已经能提供TD-SCDMA解决方案？ 李：展讯(Spreadtrum)、联芯(Leadcore)、联发科(MediaTek, 该公司当时收购了ADI的调制解调器业务)等。但这些解决方案的成本很高，而且我们也发现他们的整合度并不好。我们认为，我们应该针对未来发发展制定出一个架构。 问：未来的架构──你指的是TD-LTE吗？ 李：目前我们正在开发TD-LTE调制解调器。在TD-LTE领域，事实证明，不仅需要多模，还需要能应对多频段的应用。 中国移动(China Mobile)近日公布了该公司对TD-LTE的需求。必须能在中国境内的TD-SCDMA和GSM频带上操作，而且还需要TD-LTE调制解调器来提供在4G和FDD上的漫游能力，更别说是3G和WCDMA了。 2012年底前，我们将开发出能符合中国移动所有要求的TD-LTE调制解调器。 问：我的理解是，要设计一个能将每一种基频──包括TD-SCDMA──都包含在内的真正多模、多频LTE，并不是容易的事，甚至对高通(Qulacomm)来说也是如此。这是真的吗？ 李：没错。因为并不是每个人都拥有完整的解决方案。但在Marvell，我们所开发LTE解决方案将会是普遍的──不仅可用于中国移动，而且能用在全球市场。 问：我看到Marvell押注在LTE的行动策略。但在中阶和入门级智能手机市场呢？有人说Marvell并不看重这块市场。而在此同时，像联发科这类公司却正大举进军此一领域。 李： 我们不会对这块市场视而不见。事实上，我们的 PXA920 已经在中国的TD-SCDMA智能手机市场居主导地位。我们了解到多核心解决方案的重要性正日渐提升，有关这点，我们的竞争对手做得很好。因此，我们计划 在未来两到三季内推出整合双核心、四核心和GPU的TD-SCDMA芯片。 问：你说你的团队并不像许多跨国公司一般仅专注在客户支持层面，即所谓的“轻产品开发”(light product development)？ 李：没错。我们拥有相当程度的自主权。我们在中国进行研发和产品开发等工作。 问：你担心IP失窃或员工流动问题吗？ 李：我知道，过去几年来IP盗窃问题频传。但在Marvell，我必须说我们很幸运。我们很重视知识产权的保护，我们已经有完善的IP政策。如果贯彻我们的保护机制，我想很多相关问题都能缓解的。 问：你们的员工流动率高吗？ 李：我们在中国的员工都是年轻人，年轻人往往更常换工作。但我们的员工一直很稳定。一部分原因在于我们是少数真正在做硬核心开发的企业。这在中国并不普遍。而我们的工程团队训练有素，非常专业。 编译： Joy Teng 本文授权编译自EE Times，版权所有，谢绝转载 本文下一页：参考英文原文：Marvell aims to be China chip leader ，by Junko Yoshida

相关阅读：
• Marvell、联想和中国移动共同推动中国3G TD 智能手机
• 三星GALAXY Xcover安卓智能手机采用Marvell WCDMA解决方案
• 三星Android手机S5820采用Marvell 3G TD解决方案Ot7esmc

{pagination} Marvell aims to be China chip leader Junko Yoshida SHANGHAI, China – No, Marvell Technology is not moving to China. However, the U.S. fabless chip company based in Santa Clara, Calif., has a goal to become “the largest semiconductor company in China,” according to a Marvell executive here. With 1,900 people already employed in China and the number climbing, the California- based company is clearly committed to its China aspirations, and already sees some of its goals within reach. Earlier this year during an interview with EE Times, Weili Dai, Marvell’s Shanghai-born co-founder, went on record saying: “It’s important for us to stay in Silicon Valley. That’s where the entire eco-system of great talent and technologies resides.” But that isn’t stopping Marvell from stretching a tentacle across the Pacific. In addition to the company’s main design/engineering site in Shanghai – which employs some 1,600 people, Marvell is in the process of adding its second campus in Nanjing. The company is not disclosing details, but it acknowledges that it’s working with the local government. The new site is intended to be of a substantial size. Marvell’s Chinese investments require little explanation. “Technology is moving very fast and China is a very dynamic market,” said Ting Wei Li, vice president and general manager, responsible for Marvell’s China business. “Everyone knows that there is a huge opportunity here, but to convert that opportunity to a real business, you need to be here.” That includes Marvell getting engaged with leading telecom and TV operators and strategic partners in the system and service businesses in the local market. The ecosystem the chip company needs to work with is becoming increasingly local, moving very fast, and integrated with services, Wei Li explained. He believes Marvell is in a good position to leverage the best of the both worlds: As a Silicon Valley-company where U.S. technology still leads, while being a local semiconductor company in China capable of responding to local needs quickly. Wei Li noted that one area where the Chinese are deficient is documentation. “They’re not used to it.” He said designers and engineers need to be present at more meetings, listening carefully to what decision makers are saying, and need to be aware of new requirements as they emerge. “You need to stay connected,” said Wei Li. 800 engineers focused on mobile Marvell in China is working on a number of different product segments today. They include smart TVs, set-tops; Universal Passive Optical Network (UPON) technology, which Marvell claims is the world's only single-chip solution that supports all three fiber media protocols, including EPON, Gigabit Passive Optical Network (GPON) and Active Ethernet; and mobile chips. Of all these product lines, the mobile group commands the biggest presence in Marvell’s China operation, with 800 engineers now devoted to R&D in mobile products. Although Marvell’s mobile chip business has had a great run and a slew of design-wins with Research In Motion’s Blackberry products, the chip company’s mobile momentum seems to have fizzled a little as RIM began getting squeezed out of the global smartphone market. Mavell, however, claims that the company has a enough amunitions in reserve to regain its share. Driving the company’s renewed commitment in the mobile business are 800 engineers in China, 200 in the United States and 500 in Israel. Following is an excerpt from a one-on-one interview EE Times conducted with Ivan Lee, vice president of mobile products at Marvell. He oversees development work both in China and in the United States. EE Times: Why is Marvell’s mobile product team so spread out? And what’s the division of labor among those teams in different locations? Ivan Lee: It’s largely for historical reasons. Our 500 people in Israel are largely coming from an Intel X-scale team that Marvell acquired in 2006. Their focus is on 3G, WCDMA development. Marvell’s success at getting design wins in RIM’s Blackberry was the work of our Israel team. EE Times: What roles do your China and U.S. mobile teams play, then? Lee: In China, our focus has been more on the TD-SCDMA work we started four years ago; and now more on TD-LTE development. Our mobile products team in China consists of: silicon designers; DSP designers; those who develop protocol stacks; software developers working on Android and other upper layers. Our 200 people in the U.S. include a SoC design team in Austin, Texas and a team in Calif.; software engineers and Android development in Mass.; and a silicon validation team in Chandler, Arizona. EE Times: You said Marvell’s 800 people responsible for mobile products in China are engaged in R&D. I find some people use the term very loosely. How do you define R&D? Lee: The team here did everything from conceptualization of TD-SCDMA to its architectural development and production. I describe 99.9 percent of these 800 people as engaged in R&D. Marvell bets on LTE for mobile strategy EE Times: With all due respect, Marvell didn’t get into the mobile business in China until 2008. Why did you decide not to buy a company who had already developed TD-SCDMA or acquire anyone who had IPs? Lee: We actually did look at all the IP options and potential acquisition targets. And we decided that we can do better if we do it ourselves. Of course, at that time, many people told us that we were already too late to the market. They also advised us that we should purchase some IPs, and we should be using protocol stacks that have matured. EE Times: Who already had TD-SCDMA solutions on the market in 2008 -- when you decided to join the baseband fray? Lee: Spreadtrum, Leadcore, MediaTek (after their acquisition of ADI’s modem team), etc. The cost for these solutions was high; and we found them integrated poorly. As we were coming from nowhere, we thought we should develop an architecture that scales for the future. EE Times: Speaking of architecture that scales for the future, are you referring to TD-LTE? Lee: We are currently working on a TD-LTE modem. The TD-LTE, as it turns out, needs to be not only multi-mode but also it must respond to multi-frequency bands. China Mobile recently announced the requirement for their TD-LTE. While it needs to be able to operate on TD-SCDMA and GSM inside China, it requires TD-LTE modem to offer roaming capability on 4G and FDD, not to mention 3G, WCDMA. We’ll have a TD-LTE modem by the end of 2012, which meets all the requirements set forth by China Mobile. EE Times: My understanding is that to design a truly multi-mode, multi-frequency LTE with every baseband flavor in it – including TD-SCDMA – is not easy to do, even for Qulacomm. Is that right? Lee: True, because not everyone has a complete solution. But at Marvell, we do. Our coming LTE solution will be universal – not only applicable for China Mobile but for the worldwide market. EE Times: I see Marvell betting on LTE for the company’s mobile strategy. But what about the growing mid- to entry-level smart phone market? Some say Marvell took its eyes off on that segment. Meanwhile, companies like MediaTek are eating your lunch. Lee: I wouldn’t call it us taking our eyes off. We’ve done very well with our PXA920, dominating China’s TD-SCDMA-based smartphone market. But we do understand that the multi-core solutions are becoming very important, and our competitors have done very well. Our plan is to roll out over the next two to three quarters a family of TD-SCDMA-based chips – integrated with dual core, quad core and GPU. EE Times: You say your group is not about “light product development” focused more on customer support – which a lot of multinationals have been doing. Lee: Ye. We have quite a bit of autonomy here. We execute R&D, product development here in China. EE Times: Are you at all concerned at all about IP theft or employee retention? Lee: I understand that IP theft happened in the past. But at Marvell, I must say we’ve been lucky. We take IP protection very seriously, and we have put our IP policies in place. I think a lot of problems can be alleviated if we check our IP mechanism carefully, train our employees right. EE Times: What about employee retention? Lee: Our employees in China are young, and young people tend to move more often. But our work force has been pretty stable. Part of the reason is that we are one of the very few companies doing hardcore development work. That’s not very common in China. Our engineering team is well-trained and very specialized.