For the first time in the long history of Intel’s attempts to break into the mobile processor market, there is a really positive perception building of its chances. Samsung’s recent decision to use the Atom Cloverfield system-on-chip for some of its Galaxy Tab models was widely seen as a breakthrough, since the Korean vendor has its own ARM-based processors too. And just as the shift from the traditional PC/smartphone divide, towards hybrid post-PC devices, opens up opportunities for Intel, so it will also hope to exploit the changes in the rules of wireless infrastructure, staking its claim to lead the nascent moves towards Cloud-RAN.
And a recent ABI Research report, claiming the Atom Z2580 outperformed rivals from Nvidia, Qualcomm and Samsung, generated a large number of column inches. In fact, ABI did not specify its benchmark, and similar results have generally related only to a single benchtest, AnTuTu. As Jim McGregor pointed out in a great EETimes blog: ‘The only clear conclusions that should be drawn are that the leading ARM-based processors still have a performance lead over the latest Intel processor; all benchmarks should be questioned and none used exclusively; and recent headlines were more sensational than truthful.’
But these headlines indicate one important reality ‘ that the industry is starting to take Intel’s Atom seriously in the mobile arena. While its success will remain to some extent tied to the uncertain fortunes of Windows 8 in the post-PC space, it needs to push further into ARM territory, and Samsung will help it there, in credibility as well as sales. The Korean firm is the most credible challenger to Apple in tablets and hybrids, and is using Intel chips in smart TVs and other gadgets apart from the new Tabs.
New Intel CEO Brian Krzanich has made it very clear that he will reverse the old situation where the Core processors received more resources and attention than Atom. He recently told a meeting: ‘We see that Atom is now at the same importance [as Core] – it’s launching on the same leading edge technology, sometimes even before Core.’
However, even a successful Intel will have huge adjustments to make. It has addressed some of the technology challenges, notably in power consumption, at last, but it needs to change more radically to adapt to a market where there are multiple fierce competitors, and where unit prices and margins are so much lower than it is used to. Qualcomm shipped 173m chips last quarter, generating about $3.9bn in revenue, implying an average selling price of around $22 ‘ compared to Intel’s reported ASP of $107 in PCs.
As Motley Fool points out: ‘At those prices, even if Intel sold every mobile computing processor expected to ship this year, it wouldn’t match the annual revenue it generates from PCs. Hypothetically, if Intel could fetch the same $22 Qualcomm commands and it could power all of the roughly 1.15bn smartphones and tablets expected to ship this year, it would translate to $25.3bn in revenue, $9 billion less than what Intel’s PC client group brought in last year. For mobile to surpass its PC client group revenue, Intel would need to command an ASP of $30, a 36% premium over Qualcomm’s ASP.’
Of course, Intel will not achieve 100% of the market, and the share enjoyed by Qualcomm would be a very tough challenge even in a five-year timescale. It will have to accept revenue declines in devices and seek to balance that elsewhere. Unlike most of its competitors, it has many other businesses, notably the servers ‘ though in the cloud/blade area that is being assaulted by ARM ‘ and, it hopes, the embedded ‘internet of things’ arena.
There are also other parts of the mobile landscape where Intel has more original technology and the chance of leading, not following. These are on the infrastructure side, in particular Cloud-RAN, a platform to which the chip giant has attached its colors decisively. Since C-RAN relies on virtualizing baseband functions on large servers and in the cloud, it fits perfectly with Intel’s incumbent status in servers, and its aggressive moves to bring virtualization from the data center to the network via carrier SDN (software defined networking). C-RAN and carrier SDN will move key network functions ‘ despite the best efforts of the telecoms equipment vendors to keep things at least semi-proprietary ‘ away from specialized network hardware and onto generic routers and servers, based on, so far at least, x86 platforms.
So Intel has a chance to lead here, though it will be a slow burning fuse. Mass adoption of C-RAN and SDN is years away, with many issues to confront first ‘ the trade-offs of economical standardized hardware versus tuned-up, dedicated systems; the issue that C-RAN, with its very low latency requirements, is only currently suited to operators with massive fiber availability; the emergence of full standards and therefore a wide choice of chips and kit.
At this early stage, C-RAN has been driven by China Mobile, a massive owner of fiber, and Intel has been one of the key partners in its development projects, along with others such as Alcatel-Lucent and ZTE. NTT Docomo, KT and SK Telecom are also engaged in the beginnings of C-RAN projects, and Intel has been involved to some degree with all of these, most prominently with KT’s Cloud Communications Center (CCC) platform, codeveloped with Samsung, which spans LTE, 3G, Wi-Fi and even WiMAX.
These early movers are using largely proprietary, customized C-RAN platforms to gain advantage in extremely pressurized markets, in the case of Japan and Korea, or to exploit a virtually clean slate in network design terms, in the case of China Mobile. While European majors like Deutsche Telekom and Orange are evaluating the approach, most operators will wait for greater standardization and the appearance of commodity servers and reliable software, while others ‘ including several US giants ‘ are actively sceptical.
To gain the richness and breadth to meet the needs of a large number of carriers, not just the atypical Asian giants, C-RAN also needs a wider ecosystem and plenty of innovators. One of these is Asocs, supplier of silicon cores for modems, which has worked on the China Mobile project, but is also backing a would-be standard modem programming language, MPL, to simplify and harmonize C-RAN efforts.
MPL is being used by China Mobile, which will start its second round of C-RAN trials later this year using servers with PCI Express cards for the baseband processing. Each card will run on four FPGA (field programmable gate array) chips with Asocs C2100 cores, and each FPGA will be able to handle 12 LTE sectors, according to Asocs CEO Gilad Garon, in an interview with EETimes. The cellco will also test MPL, which allows developers to call C libraries to run jobs on processors. The approach reduces the cost per LTE sector to under $30 in a large roll-out, says Asocs, compared to $10,000 two years ago (of course, however cheap the digital signal processing, C-RAN has many other costs such as servers, software and fiber).
In this trial, Asocs has worked with Intel, with its cores linking to Xeon processors, but it also points to a near future in which C-RAN, and SDN in general, will be less x86-specific. It expects the emerging breed of low power ARM-based servers to have an important role in telecoms networks, and C-RAN in particular, in future, an outcome which puts the pressure on Intel to establish itself firmly now, even at the difficult stage where there are only actually a few target customers.
It also faces the dilemma inherent in any emerging technology ‘ while a headstart is valuable, there will be no wide operator acceptance of C-RAN until there is a broad ecosystem and plenty of silicon choices. That means the market needs the traditional infrastructure chip providers, like Broadcom, Freescale and Texas Instruments, to expand their efforts, even though this could, once again, see non-x86 platforms eating into Intel’s carefully carved-out niche. TI has already extended its KeyStone architecture into key elements of C-RAN, such as low energy servers, and looks poised to be an important challenger once carrier support becomes more assured outside of the three east Asian mobile powerhouses.
Gordon Mansfield, chairman of the Small Cell Forum and executive director of small cell solutions at AT&T, told EETimes: “Having general purpose hardware and software is the ultimate goal, and all the big, heavy processing silicon providers will end up there I believe, but right now it’s too early to see who will come out as leaders.’