Slashdot Mirror
How The IoT Will Change The Chip (techcrunch.com)
"Get ready for some big changes in the 'silicon' of Silicon Valley," writes tech CEO Narbeh Derhacobian who argues that the need to build tens of billions of connected sensor devices will change the way computers get built. "Just like smartphone owners like to pick and choose which apps they want, IoT manufacturers may want to shop for components individually without being locked into a single fab." An anonymous reader summarizes his article on TechCrunch:
Thousands of different hardware devices, each selling around one million units, "would suggest the need for a much greater diversity of chip configurations than we've seen to date." Currently smartphones are engineered using a "System on a Chip" design where all the components are "locked into a single manufacturing process," but Derhacobian predicts chip manufacturers will continue a trend of moving towards a "System in a Package" approach -- "packing components closely together, without the complete, end-to-end integration... In a smart, connected world, sensor requirements could vary greatly from factory to factory, not to mention between industries as varied as agriculture, urban planning and automotive."
"In some ways, the great trends of the PC and smartphone eras were toward standardization of devices. Apple's great vision was understanding that people prefer a beautiful, integrated package, and don't need many choices in hardware. But in software it's generally the opposite. People have different needs, and want to select the apps and programs that work best for them."
"In some ways, the great trends of the PC and smartphone eras were toward standardization of devices. Apple's great vision was understanding that people prefer a beautiful, integrated package, and don't need many choices in hardware. But in software it's generally the opposite. People have different needs, and want to select the apps and programs that work best for them."
Whenever spinning up silicon, either general purpose or purpose-built, one of the most pressing questions is 'what is the total addressable market value for the segment in which it will be used, and what share of the market is it likely to get?'. Until IoT finds its killer app its unlikely to get much love from silicon vendors.
Funny how the article is written by a CEO of a company currently using 45nm fab about why they should be used over a SOC in 14nm fab...
I think this only shows how isolated some people are from the rest of the world. Nothing new to see here.
I didn't realize until now that iOS apps run Android.
I hadn't the slightest objection to his spending his time planning massacres for the bourgeoisie... (P.G. Wodehouse)
Why not have small DIPed SoCs for at least SMT boards? With high levels of integration, you could have IoT chips replacing current MCUs. Lots of applications could be sufficiently MCU-like for that.
Ezekiel 23:20
I'm not saying the article is wrong, because I'm not really sure if it actually says anything.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
They'd grab whatever components were cheapest at the time and put them together to build a PC. You'd buy one Acer model PC, really like it and recommend it to a friend who would buy the exact same model, and his would come with completely different components and suck. It's how Acer earned their reputation as a low-rate PC manufacturer that they're still trying to shake today. Routers occasionally run into this problem as well, as hardware changes from revision 2 to revision 3 result in incompatibilities or different behaviors even though two routers are "the same model."
If you want to dumb down the software and use non-standardized hardware, be my guest. But you'd better make damn sure the user experience is consistent across all the different hardware versions you're using. Otherwise you're saving money on the front end just to end up paying more at the support end.
... the need to build tens of billions of connected sensor devices ...
There is no "need" for the IoT. We've been getting along fine w/o it and don't really see any point to 99% of it.
Thanks anyway Narbeh.
It must have been something you assimilated. . . .
...immediately identifies themselves as someone I need not take seriously.
Nobody rational GIVES A SHIT about connecting their coffeemaker to the internet.
Anyone reasonably skeptical will have nothing to do with the idea of connecting their front door locks to the internet.
Short version: it's the latest stupid internet fad, interesting only to the circle-jerk of people promoting such things.
-Styopa
I like limiting my attack surface. I have a smartphone, my computer, a few appliances (dedicated firewalling router behind the ISP's router to ensure that if someone blasts that thing wide open, my LAN is still secure), and a NAS.
I don't need a smart fridge. If I wanted to spend $3000 for a new fridge, I'd buy one that works on both electric and propane, so if there is a power failure, my stuff still stays cold.
I don't need a smart TV. A HTPC does the job there.
I don't need smart deadbolts. I don't need burglars unlocking stuff with a press of a button, nor knowing when I lock and unlock my doors to build a casing schedule of when to rob or invade.
Want to know why IoT is so popular? Analytics. Every IoT device generates a lot of info and sends it all up constantly, because that info can be sold. Do I need any more devices monitoring my ass? Not really. If I have my phone with me, my location is tracked 24/7. If I buy a new VW or Chevy, the car tells where I am as well. Fuck lining people's wallets with surveillance products I have to buy.
So instead of connecting chips via a circuit board, connect chips directly. It sounds like a good idea, but I think the conclusion, that it would enable every IoT manufacturer to customize their chip configurations, is wrong. It would be expensive due to economies of scale, and I don't see why IoT devices can't use the plain old circuit board. Do IoT devices need to be super small and super power efficient? However, I do see that it might benefit smartphone manufacturers, where size and power consumption matter. However, it is likely that suppliers would offer chip configurations for narrower use cases.
Till security is fixed, the notion is not only preposterous but dangerous.
Something's fishy about this article. I'm not sure if my superpower to detect P.R. campaigns has developed to the point where I can get it in just one post, but my senses are tingling about this story. Let's see if there are a rash of IoT stories over the rest of the weekend.
You are welcome on my lawn.
Everything claimed as revolutionary changes have been happening over the past 5 years.
Welcome to 2008... oh and us hackers have been screwing around making our own fully open iOT devices for the past 2 years already on a "single chip" device. the ESP 8266 is already destroying the arduino and other platforms rapidly.
Do not look at laser with remaining good eye.
Just like all the other times, the home will be the last place where IoT is applied. It will rule the commercial and industrial world for a generation or two before hoi polloi trusts it in the home. We will see bridge beams that continually report their own stresses, freight containers that log their location, aircraft that log flight data to satellites instead of to onboard black boxes, and building load compensation weights that continually adjust themselves.
Yes and no. Security needs to be built in. People are in a rush to get on the bandwagon, and kids don't care about security, so it's left off of consumer IoT. But it's there in the professional stuff.
The current chips need three things for IoT, and the ball is only barely starting to roll from chip makers: Radio (sub gigahertz, not lame zigbee or wifi), Security, and Low Power. Those last two sort of fight with each other though. Hardware security features suck up power, though you don't have to use it all the time; software based security is even worse since it takes more time before you can drop back to low power mode. No one is really supporting PKI, very few people have secure key storage (ie so you can't sniff it out on the bus or with jtag), etc. Low power is an afterthought, a lot of consumer grade IoT assume you're recharging every day, dumb bluetooth things talking to your phone for example (not really IoT but people slap that label on), you want a a few microamp of sleep current if you want decent battery life, or less.
That's consumer IoT. I think it's a dumb idea, and I work in the IoT industry (non-consumer). Things like Nest are silly and ship with broken features (horrendous security), and Google is naive in trying to enter a field they know nothing about. Everyone's working at cross purposes everywhere, multiple competing standards that are not anywhere near what is required (networks too chatty for battery use, or bulky data like XML, etc).
For IoT you want a system-on-a-chip with a BLE / Bluetooth Smart radio. The big three there are Nordic, TI, and Dialog, and there are a couple others.
All three supply chips with non-BGA packaging options.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Yeah, IoT for business is all over security. Totally secure. It's all plug-n-play heaven with a big security bow on top. Because they've been doing it for so long, they can't do it wrong.
Sleep your way to a whiter smile...date a dentist!
I'm not saying industrial IoT is totally secure, but it is being worked on and the customers are starting to demand it. I'll agree that industry used to be nervous about security, angst that if they turn on all the security switches that they'll lose contact with the devices and waste money sending out trucks, but that attitude seems to be changing.
And there's no "plug and play". There are no widely accepted standards here, even with radio, and following a standard won't mean you can talk to a competitors device.