ARM uses a unique survival model to cope with the melee of giants such as Intel

　　In 2012, the world's Internet-connected devices were mainly smart terminals, with shipments reaching 1.6 billion. Most of them were ARM-based smart mobile terminals, with ARM chips accounting for about 75%.

　　By 2017, in addition to traditional PCs and mobile terminals, a large number of digital set-top boxes, in-car entertainment systems, and even other smart wearable devices will become indispensable Internet-connected devices. According to Gartner data, global shipments will reach 4 billion, but the specific share of ARM chips is not emphasized. However, according to the chart, the market share of PC platforms will further decline.

　　With the rise of smart mobile terminals, ARM has grown from a small company with only one foundry to a world-renowned IT company supported by as many as 300 companies around the world (including Samsung, Qualcomm, MediaTek, etc.). In the future, with the popularization of 4G and 5G networks, low-power embedded system platforms will once again usher in explosive growth, and ARM's market share will undoubtedly expand further.

　　Since 2008, Intel has been trying to enter the mobile and portable field, but has never been able to gain a foothold. With the release of the new 22nm SoC process and the new Atom architecture, Intel finally began to show its power, and ARM also came out of its low-key backstage to talk about its development strategy. Intel often does this, but this is the first time for ARM.

　　The performance improvement and power reduction of Intel's new Atom processor will help Intel increase its share in the mobile market. A good example is the recent cooperation between Intel and Samsung, where the new Samsung Galaxy Tab 3 will use Intel's Clover Trail chip.

　　Intel's Atom processor is superior to ARM's equivalent chips in terms of performance, but it is at a clear disadvantage in terms of power consumption, life span and management. Compared with the low power consumption of ARM architecture processors, Intel processors have higher power consumption.

　　Although ARM architecture processors have great advantages in terms of energy consumption, they are far behind Intel or AMD's X86 processors in terms of performance. Mobile terminals require portability and battery life more than performance, while desktop platforms are just the opposite. This is exactly the bottleneck of ARM processors.

　　Intel plans to ship a new version of its Atom processor, Silvermont, later this year, followed by Airmont in 2014. Intel will eat into the mobile chip market dominated by Qualcomm and other ARM licensees. Global smartphone sales are expected to grow 26% this year to 870 million units. As the smartphone market becomes saturated, the growth rate is expected to slow, but Intel still has a chance.

　　Imagination launches MIPS architecture CPU to compete with ARM

　　Imagination, which owns the PowerVR series of graphics processors that occupy more than 50% of the mobile GPU market share, acquired MIPS, a traditional RISC processor design company, last year. Now Imagination has officially launched a CPU Core based on the MIPS architecture, named Warrior (MIPS Series 5). This obvious naming is undoubtedly intended to challenge ARM's long-standing position.

　　The CPU Core can support Android system. Warrior 32-bit and 64-bit versions will provide the best performance and efficiency among similar products and will change the entire CPU industry. Of course, whether this is real strength or marketing exaggeration, we still have to wait for further tests to verify. Warrior is scheduled to be launched at the end of 2013.

　　Intel CEO: It is not impossible to manufacture ARM processors

　　So far, Intel's foundry business can only be described as a small-scale operation. Although it has a number of customers and is even preparing to use the 14nm process that has not yet been mass-produced, it is still far from being able to form a scale and has no truly heavyweight orders. On the other hand, Intel has enough factories and production capacity around the world, but now the demand in the PC industry is declining. Will a large number of production lines be idle and wasted?

　　Obviously, this is not the case. Intel CEO Krzanich recently accepted an in-depth interview with Reuters, in which he talked about the foundry issue. From his statement, it can be seen that Intel is very serious about foundry, and will become bigger and bigger in the future, and even does not rule out the possibility of foundry for ARM processors.

　　In fact, former CEO Paul Otellini has made it clear that Intel can manufacture different types of chips and will not be limited to x86. Rumors of Intel and Apple's foundry have been circulating for a long time, but in the end, Apple and TSMC signed a long-term contract to use the latter's 20/16/10nm process to manufacture future A-series processors.

　　Krzanich's attitude is clearer and more open, claiming that he is ready to manufacture for more customers. He said: "I think you will see some of our chips around the end of this year or early next year. We are trying to sell some chips to them (foundry customers), and we are also making our own chips, and we are working hard to understand the uses and create an ecosystem."

　　Obviously, we can't expect Intel to reveal the details now. It looks like we'll have to wait another half a year before we see Intel's big move in foundry. Interestingly, the end of this year and the beginning of next year are when new Atom mobile phones and platform processors will be released. Is Intel going to provide power to its competitors at the same time?

　　AMD doubles down on ARM-based servers

　　AMD is optimistic about the prospects of the low-energy server market and is working hard to expand this business and tap the potential of this market using ARM architecture as the main platform.

　　Andrew Feldman now runs AMD's server chip business, so it's no surprise that Intel's old rival is betting on ultra-low-power processors, which AMD believes will gradually permeate the server world. Feldman

　　joined AMD when AMD acquired chipmaker SeaMicro. From the beginning, his Silicon Valley startup wanted to join the low-power server revolution. SeaMicro originally combined hundreds of Intel mobile phone chips into a server, but now that Feldman has joined AMD, he is starting to integrate another low-power chip into the machines that are spread all over the Internet.

　　AMD has already announced that it is developing server chips based on the ARM architecture, which is already used in most smartphones and tablets around the world. On Wednesday, in order to fill the gap before the launch of ARM chips, Feldman and AMD also developed similar low-power server chips based on the company's existing x86 architecture.

ARM enters the server market, and LG becomes the lead partner of Cortex A50

　　The ARM Cortex A50 chip is expected to be initially used in tablets and servers, and the 64-bit support will allow developers to configure more than 4GB of memory without having to use physical address extension (PAE) technology. Companies licensed from ARM have already begun developing products based on the Cortex A50.

　　The ARM Cortex A50 series chips are probably the company's most important designs to date, using the ARMv8 architecture that supports 64-bit technology. The A50 series chips will include the Cortex A57 and the next-generation Mali GPU (graphics processing unit).

　　ARM said the agreement between the company and LG means that LG will be able to provide a range of system-on-chip (SoC) chips, while also being able to better take advantage of the GPU computing capabilities of ARM's next-generation Mali GPU.

　　LG becoming the chief partner of Cortex A50 may mean that the company is trying to challenge Samsung in the server chip market, and it is relying on ARM Cortex A50. In fact, ARM is also relying on LG to enter the server market.

Unique survival mode

In the traditional PC field, semiconductor manufacturers generally have two options. The first is like Intel, which takes care of everything from beginning to end, and does not rely on anyone for the design and production of architecture and chips. In fact, Intel also provided x86 instruction set licenses to the outside world in the early days, but soon took them back.

　　To do so requires extremely strong and comprehensive strength as a guarantee, including money, people and technology. Especially in today's increasingly complex semiconductor technology, there are only a handful of people who can do so.

　　The benefits are of course obvious. Not only can Intel completely control its own lifeline, but the profits are also extremely considerable. Almost any product of Intel can enjoy very high profits, and it can sell it for any price it wants.

　　However, in the mobile field, this "black-hearted" model does not work, because the processor accounts for a very high proportion of the cost of the PC system, but it is often less than 10% or even less than 5% in smartphones and tablets. No matter how hard you try, you can't make much profit. This is also one of the key reasons why Intel has difficulty in promoting in the mobile field.

　　The other is the fabless model. NVIDIA is like this, and AMD also became like this because it couldn't outstrip Intel. This type of company only designs its own chips and outsources the manufacturing to foundries, such as TSMC, UMC, GlobalFoundries, and Samsung Electronics.

　　The advantages are obvious. The burden is very light and you only need to design it yourself. There is no need to spend huge amounts of money to build a wafer fab or develop new processes. But the disadvantages are also prominent: once you have designed it, you have no say on whether it can be manufactured or what it will look like even if it is manufactured. It all depends on the ability of the foundry partner.

　　There are of course many lessons to be learned in this regard: TSMC's 40/28nm processes were initially very immature, and production capacity was slow to increase, which dragged down the entire industry; GlobalFoundries' 32nm process did not meet AMD's expectations, and the frequency and voltage of the first-generation FX/APU processors were far from the design. The 28nm process has been hyped for so long and is only now being put into use, forcing AMD to abandon an entire generation of low-power APUs and have to redesign them and look for TSMC again.

　　ARM is completely different from them. It does not manufacture or sell any chips, but only designs its own IP, including instruction set architecture, microprocessor, graphics core, and interconnect architecture, and then sells the license to whoever likes it. Customers can do whatever they want with ARM IP.

        Analysis:

        The fierce competition between ARM and Intel seems to be mixed with mutual assistance. Both are marching into each other's territory, and they also have to join other giants in the melee. The outcome is still determined by product technology.