Intel loves the 'internet of things' – if the 'things' have Quarks and Atoms in them, anyway
Chipzilla wants Intel Inside EVERYTHING
Still licking its wounds from the beating it took in the mobile market, Intel is bound and determined to not get mugged when the next computing big boy comes tromping into town – the Internet of Things, or "IoT" to the digiscenti – and has introduced two chip lines and one interoperability effort to shore up its defenses.
"I'll admit that we have been somewhat quiet about this over the last year or so," the general manager of Intel's Intelligent System Group, Ton Steenman, told attendees at the company's IoT rollout event in San Francisco on Tuesday.
"But there have been some good reasons for that," he added, noting that he and his team have been busy identifying what he called the "pain points" in the industry by working with customers, then determining what Intel could bring to the IoT party to reduce that pain.
"So you can ask yourself the question, 'Why now? Why is this happening at this point'?"
According to Steenman, the answer is that there is now a "critical mass" of IoT devices being connected to – where else – the cloud, a development that he identified as a "precursor for IoT to become real."
IoT is the inevitable next step in comprehensive computing, Steenman said, and backed up that assertion with some hefty numbers.
"If you look at the industries that are largely affected in the short-term horizon by the advantages and the potential of the Internet of Things," he said, "those industries spend about thirty-six trillion dollars on operating costs."
Even if the IoT can improve those industries' efficiencies by only a few percentage points, well – as has often been misattributed to former US senator Everett Dirksen – "A billion here, a billion there, pretty soon, you're talking real money."
The number of IoT devices on the horizon are also staggering. Intel has often trotted out the figure of 15 billion such devices in place by 2015, but Steenman upped the ante on Tuesday, citing research by IDC, Intel, and the United Nations that balloon that number to 200 billion by 2020.
That horde of devices will need a lot of chippery, and Steenman offered two lines of IoT-centric chips to feed that need, one based on the new E3800 "Bay Trail-I" Atom processors and a second based on the ultra low-power Quark processor cores that debuted at this year's Intel Developer Forum, and which will first appear in the X1000 product line unveiled last week by Intel CEO Brian Brian Krzanich at the Maker Faire in Rome.
The IoT versions of both processor families will include error-correction code (ECC), industrial-level temperature certification, and integrated security. The current 64-bit Atom E3800 line includes parts with both two and four cores – one thread per core – but Steenman didn't say whether the IoT E3800s will be available in both versions.
The E3800 line comes with graphics and media processing improved over previous Atom parts, but the X1000 family is designed for less-demanding workloads – they're single-core, single-thread 32-bit parts based on the Pentium ISA, and will be available at speeds up to 400MHz.
Chips for future IoT devices are all well and good, but as Steenman pointed out, 85 per cent of eventual IoT devices will be comprised of existing infrastructure, updated to communicate with one another and with the cloud. "As an industry," he said, "we can't wait until a full refresh of the infrastructure in order to enable the benefit of business transformation and the great things that the Internet of Things can bring."
And so Intel is working with its wholly owned partners Wind River and McAfee to build what Steenman calls a "systems-to-systems view," an open-standards platform built with off-the-shelf components that allows legacy hardware to acquire IoT data, share analytics among devices and the cloud, and create middleware to tie it all together.
"All of these components, including software and transation security, have to come together in a full solution," he said, and to accomplish this goal, "Intel will bring a scalable roadmap of gateway products to market."
Intel won't produce actual hardware products, however. Rather, Steenman said, "We are building an integrated, validated software stack on our hardware with specific interfaces, APIs, northbound to the cloud and southbound to devices, and a certain amount of functionality in the middleware, and we will offer that to our customers as integrated, validated building blocks."
Intel's customers will take this and build their own end-product gateways. "At this point," said Steenman, "Intel has no plans to bring a full, branded gateway product to market."
These gateways will be designed for many levels of use models. For example, all the HVAC (heating, ventilation, and air conditioning) units in a building could each have a gateway that would communicate with the building's central gateway, that gateway could then communicate with a central gateway in that building's industrial or academic campus, which could in turn communicate with a main coordinating gateway that could manage all interoperability of the entire enterprise.
As you might have guessed by now, gateways built using Atom or Quark processors wouldn't have enough oomph to manage an enterprise's IoT needs. Steenman says that there will eventually be gateways powered by Core and Xeon processors, as well – although the first two gateway designs will be based on Atom and Quark, with products expected to appear in the first quarter of next year.
As the IoT evolves, Steenman believes, customers will begin to look for higher and higher-end gateways, and Intel will be quite happy to sell them the Core and Xeon chips needed to power them. ®
Sponsored: Benefits from the lessons learned in HPC