PCI backers push low-power features with an eye on Internet of Things
- 05 June, 2014 08:19
The PCI technology that's served PCs and servers well for decades is learning to get by on less power so it can play in mobile devices and the Internet of Things.
Even as it works to boost the speed of PCI Express in an upcoming version 4.0, the PCI Special Interest Group has a three-pronged strategy for serving markets that demand lower power consumption. That capability is already important in smartphones, tablets and thin and light laptops, and it will be critical in the billions of sensors and other embedded components expected to form the Internet of Things.
PCI is a family of technologies for connecting storage and other components to the host bus of a PC or server. Its current star, PCIe 3.0, can send 32GB of data per second over a typical link. PCIe 4.0 will be twice that fast when it's finished in 2015 or 2016. PCI's strong point is the fact that it's been around for more than 20 years and new PCI gear works with the old, stepping down the speed to match what's available on the other end.
But like the rest of the IT industry, the PCI-SIG sees that the big growth is coming from smartphones, tablets and small connected objects being installed around homes, streets and factories. To get adopted there, PCI standards need to sip energy so batteries last longer.
"We think we have good enough technology to enter those spaces," said Al Yanes, chairman and president of PCI-SIG, at the group's developer conference in Santa Clara, California, on Wednesday. PCI technologies have a built-in advantage for IoT because backward compatibility with software that's already in use makes it easier for devices in the field to get recognized, the group said.
PCI-SIG also highlighted three technologies for meeting low-power needs.
One of those is M-PCIe, an adaptation layer that puts PCI smarts on top of M-Phy, a specification from the MIPI Alliance that defines the physical aspect of interfaces inside devices. M-Phy is increasingly used in medical, automotive and wearable devices, according to the MIPI Alliance. M-PCIe lets system makers keep M-Phy and swap in PCI technology on top of it.
"You can use the PCI Express that you know and love very quickly, off the shelf," said Ramin Neshati, PCI-SIG's marketing workgroup chairman. Any existing software that was written for PCIe will automatically work with M-PCIe, he said.
Another specification, L1 Sub-States, lets a PCIe interface shut down more functions while in sleep mode and still wake up when needed. Finally, there's an option in PCIe called "half-swing" that cuts power consumption in half from 800 millivolts to 400 millivolts.
Adding low-power options is a good move but doesn't address PCIe's main problem in the world of phones and IoT, according to analyst Nathan Brookwood of Insight 64. The problem is that most of those systems pack all their components on a motherboard of an SoC (system on chip), so there's no need for an interface to attach other parts, he said.
"I'm not convinced it won't work. I'm just not convinced that it's going to add enough value to be a significant player," Brookwood said. PCI-SIG says it's studying how to make its specifications more attractive for SoC designs.
Where more power-efficient PCIe may turn heads is in thin and light laptops or the "two in one" PC-tablet systems Microsoft is pushing, Brookwood said. There, manufacturers still want to be able to configure products with various add-ons for networking, storage and graphics.