搞不清PC外市场规则，霸主英特尔坠在云里雾里

在近日的2011年度分析师会议上，英特尔宣称已经在其Medfield处理器中展示了一款32纳米内核，该处理器用于智能手机和平板电脑，英特尔CEO Otellini承诺英特尔将在2012年上半年为Medfield推出一款主要智能手机设计。Medfield是Atom的32纳米版本(参阅电子工程专辑报道：英特尔公布芯片开发新策略，Atom华丽转身)。

架构总经理Dadi Perimutter展示了Medfield智能手机设计和一个七英寸平板设计，运行谷歌Android的Gingerbread版本。英特尔正在转向x86谷歌的Honeycomb，这是平板电脑版Android，并预期在年底前推出一款10英寸平板电脑参考设计和开发工具。

Perlmutter表示，Medfield的功率将是毫瓦级别，可与目前的40nm智能手机芯片匹敌。他说，该芯片为单核，而竞争对手则正在推出双核芯片，但Atom内核将提供比对手更好的性能。

Perimutter展示了Medfield智能手机

但是英特尔在年度分析师会议上所说的东西与他们所做的很值得斟酌。分析师认为英特尔在PC领域的霸主地位已经过气，ARM智能手机与平板才是潮物。英特尔的高管们在很多场合下反驳着这个潮流，但他们的声音盖不过自己的表现。

我听着英特尔的高管们一再宣称他们能设计出功耗可与ARM世界产品相比的处理器。结果呢？这次会议上最大的新闻仍然是英特尔打算如何加强低功耗设计的努力，说明高管们认为公司需要做得更好才行。

Dadi Perimutter确实通过展示一张图表——但没有任何确凿的数字——表示英特尔的32纳米Medfield智能手机处理器在功耗方面几乎和目前的任何一款40纳米芯片一样好。但这款芯片被诺基亚拒绝了，英特尔也没有勇气公布有哪些公司将采用该处理器。

一家OEM厂商的人告诉我说Medfield耗电太严重。他呼吁英特尔从头设计一款x86处理器，而不是在Atom的基础上缩减能耗。

英特尔的高管们还宣称公司打造了强劲的软件开发实力，面对ARM平台竞争对手占据优势。但英特尔的智能手机与平板参考设计只能运行Android 2.3——去年的版本。Android 3.0的移植工作仍在进行之中，而MeeGo——英特尔的移动版Linux——仍然处于开发阶段。

英特尔的高管们说，拥有9000名软件工程师的英特尔是全球第三大软件公司。我怀疑它至少得排第四——列在Google、微软、IBM之后。不管怎么算，英特尔三分之一的软件工程师正在新的McAfee子公司开发杀毒软件和安全软件。

英特尔还在宣传PC市场是如何的充满活力，在委内瑞拉等相对较小的新兴市场取得了大增长，在墨西哥、俄国等中等规模市场发展也不错。

英特尔展示的图表没有提供PC在西欧、日本、美国等主要市场的黯淡数字。英特尔仍然预测全球PC市场在未来几年的涨幅依然能够达到14%，但他们也承认如果平板的发展有部分预测的那么好，这个涨幅会削减为11%。

英特尔展示的营收趋势

我已经在这里嗅到了一些拒绝承认现实的意味。先别谈英特尔对ARM版Windows的攻击。左侧的图表展示了更准确的数字。英特尔PC客户端业务季度销售额已经快要到顶，如果图表统计时间更长的话这个趋势会更明显。

下一页：烟雾背后

{pagination}

PC业务增长的复杂运算背后掖着重振PC设计的理念。这可不是什么新东西。

前英特尔高管Pat Gelsinger十年前就开始唱这调子，鼓动台湾厂商去和苹果iMac比拼工业设计。两年前英特尔说自己将会通过桌面级功能、十小时续航、立即开机的超薄平板引领市场——听起来像是2011版iPad。

有些话弄得我不明所以。英特尔说Thunderbolt技术将会在今年进入主流市场，带来一亿美元的营收。首先，我以为这个接口是不需要授权费的；其次目前除了苹果公司以外还没有一家PC厂商采纳该技术。

英特尔确实拿出了一些让人印象深刻的东西。英特尔嵌入式系统营收达到15亿美元，到2013年会增长25%，这得益于各种高端控制器广泛采用x86。

英特尔收购McAfee也值得赞许，英特尔得以从今年秋天开始为所有操作系统和管理程序开启硬件可信根(Root of Trust)。这是x86架构的一个重大胜利。

英特尔在数据中心方面也是凯歌高奏，Web 2.0公司仍然成集装箱地购买x86服务器。多亏了Facebook和Youtube。

英特尔拥有极端出色的工艺制程、无数聪明的工程师和管理人员，试图进行大的尝试、更灵活、跑得更快。有时候他们只是搞不清楚充满烟雾的市场规则。

嵌入式X86的销售有很大的成长空间

下一页：两个困惑

{pagination}

困惑之一：英特尔在软件业务方面

英特尔表示它现在是全球第三大软件公司，拥有9000名软件工程师。但是其中三分之一是致力于安全业务，主要来自英特尔新收购的子公司McAfee。其他很大一部分是风河子公司的嵌入式团队。

这确实是一个大型的软件公司，但仍未建立任何移动Linux方案，能在今天火热的智能手机和平板市场上产生任何影响力。在这一点上，英特尔应该做得更好，将更多的注意力放在已经准备出货的多核处理器的编程问题上。

英特尔软件工程师分布情况

困惑之二：Medfield的功耗

英特尔软件工程师分布情况

为了消除其Atom芯片的功耗传闻，英特尔展示了Medfield 32纳米与目前40纳米工艺的对比。注意英特尔隐去了纵轴数字。

点击参考原文：Intel: Under the smoke and mirrors

《电子工程专辑》网站版权所有，谢绝转载

{pagination}

Intel: Under the smoke and mirrors

Rick Merritt

The contrast between what Intel was saying and doing at its annual analyst meeting here yesterday was pretty revealing.

The context was clear: Analysts think Intel's PC franchise is starting to look like a pair of worn disco pants at a Web 2.0 party where ARM smartphones and tablets are the latest fashion. Intel execs fought the characterization on many fronts. But their actions spoke louder than their words.

I heard Intel execs repeatedly claim they can design a processor as low in power consumption as anything in the ARM world. Yet the big news from the event was about how Intel was going to increase its focus on low power design, implying execs thought the company needs to do better.

Intel's Dadi Perlmutter did show a chart that—without any hard numbers—suggested Intel's 32nm Atom-based Medfield smartphone processor is about as good as any 40nm chip out today in milliwatt-level power consumption. But so far the chip has been rejected by Nokia, and Intel lacks the confidence to share the names of anyone who might be committed to using it.

One OEM told me Medfield is too power hungry. He called for a new ground-up x86 design beyond the scope of an Atom shrink.

Intel execs also crowed about how Intel has become a software powerhouse, giving it an edge over ARM-based competitors. Yet the reference design for smartphones and tablets it trotted out ran Gingerbread, last year's version of Google's Android environment. A port to the superhot Honeycomb software is still in the works, and MeeGo—Intel's version of mobile Linux—was a footnote in the proceedings.

With 9,000 software engineers, Intel is now the world's third largest software company, execs said. I suspect it is at least fourth behind Google, Microsoft and IBM. In any case a third of those Intel engineers are working on antivirus and other security software in the new McAfee subsidiary, a big factor in how you view the numbers.

I also heard plenty of chest beating from Intel about how vibrant the PC market is with big growth in relatively small emerging markets like Venezuela and moderate growth in mid-sized markets like Mexico and Russia.

The slides Intel put up did not show the lackluster numbers in major markets like Western Europe, Japan and the U.S. Overall the company still estimates the PC TAM will grow 14 percent worldwide over the next several years. But it did admit its estimate will get a haircut down to 11 percent if tablets are as big as some predict.

I am beginning to sense some denial here. Let's not even talk about the FUD Intel is starting to spread on Windows on ARM.

The chart on the left told a more accurate tail. Quarterly sales of Intel's PC client division appears pretty clearly to be peaking. The image would be even more clear if Intel showed a longer swath of the group's financial history.

Tomorrow's PC looks like today's tablet

The complex calculations about PC TAM growth were buttressed with pledges to reinvigorate the design of PC. Boy, that's an old one.

Former Intel exec Pat Gelsinger sang that tune a decade ago, staging a PC fashion show to get Taiwan's beige box makers to compete in industrial design with the Apple iMac. Now Intel says in two years it will lead the market in creating ultrathin tablets that have all the power of a desktop PC, a ten-hour battery life and instant-on software—sounds like the 2011 iPad to me.

Some things just left me scratching my head. Intel said Thunderbolt will go mainstream this year and generate $100 million in revenues. First of all, I thought the interface was royalty free, and secondly I have yet to hear a PC maker other than Apple adopt it.

The company did show some credible highlights. Revenue in the catch-all category of embedded systems is up to $1.5 billion and should grow a heady 25 percent through 2013, thanks in art to the broad adoption of the x86 as a high-end controller in everything from routers and switches to intelligent everythings.

I also give Intel kudos for the McAfee deal which will give it the clout to enable hardware root of trust to any operating system or hypervisor starting this fall. That's a big win for the x86 architecture.

The company also seems to be firing on all cylinders in the data center where Web 2.0 companies are still buying x86 boxes literally by the shipping container. Thank you, Facebook and YouTube.

At the end of the day, Intel has amazingly good process technology, tons of smart engineers and managers with the cojones to try big things, be flexible and move fast. Sometimes they are just a little too ineffective with that whole smoke-and-mirrors marketing routine.

Embedded x86 sales have plenty of room to grow.

Obfuscation by Powerpoint: Intel in software

Intel said it is now the world's third largest software company with 9,000 software engineers. But a third of them a security people, mainly dedicated to its new McAfee subsidiary. Another big group are embedded folks at the Wind River division.

This really big software company has yet to establish a version of mobile Linux that has any clout in today's super-heated smartphone and tablet markets. At this point, it might do well to turn more of its attention to the problem of programming the many core processors the company is preparing to ship.

On the left, a breakdown of Intel's software engineers by function. The upper left sector in green represents 34 percent of staff dedicated to security in the current configuration of the group run by Renee James.

Obfuscation by Powerpoint: Medfield power consumption

To dispel rumors about power use in its Atom chip, Intel showed this chart comparing its 32nm Medfield to today's 40nm apps processors. Note the lack of figures on the vertical axis.