Hacker News new | past | comments | ask | show | jobs | submit login
Intel Edison Module (intel.com)
217 points by gao8a on Sept 9, 2014 | hide | past | web | favorite | 53 comments

Very interesting! The picture is kind of confusing though, it took me a minute to realize that the picture was of an "expansion board" with the module installed (in the lower left hand corner).

I like what I see though, they have packed quite a bit of processing power into the tiny package. Good amounts of memory too. And I really like the built in WiFi/bluetooth. Looks like it could be a good option for making connected devices with. The only problem is transitioning from prototype to production with Intels current business model.

This video does a good job to show off the Edison on it's own, for anyone interested. https://www.youtube.com/watch?feature=player_embedded&v=GY8k...

Its definitely a lot better than the Galileo - Intel was pretty liberal giving them away for free, yet still nobody used them since they sucked.

Not sure what the strategy is though - are they afraid to loose developer mindshare to Atmel / TI? Are they hoping that their chips will power the next kickstarter success?

My 5 cents: at least they are a player now, and the CC3200 has a competitor.


- for battery-powered projects, its a bit large

- 1.8V logic level is nice for energy consumption, but its really hard to source components for that, and level shifters are annoying

- price is a tad bit high, though adequate for its performance

You need to realize they are trying to compete in a market soon to be dominated by $4 a pop (RETAIL) Chinese modules. They will fail (and so will TI).


Galileo was a total failure. Intel gave up giving whole stock to MS, MS in turn gave them out to developers pretending they have viable IoT platform.

Thank you for this- I have a handful of completely unmarked boards that appear to be this module. Who knew they might be so capable? I hope someone translates the instruction set document linked. This is definitely the direction that things are going, at least more than an odd $50 Intel mongrel.

Intel wants us to pay for something that industry won't buy with the Edison. Contrast to Arduino and rPi, which are providing affordable access to something only available otherwise to industry:

AVR: High volume microcontroller- let's make it accessible and useful outside of heavy industry == Arduino.

ARM: Exploding in popularity due to mobile devices- let's make it accessible outside of device manufacturers == rPi.

Atom: No one is using these in industry- let's palm it off on hobbyists and see if they can drive demand == Edison.

Hard to find the point of the Edison. Too anemic and dull next to true next-gen SoC solutions like the Zynq [http://zedboard.org/], and too overpowered for most microcontroller applications. Weird price point. No GPU or vidout. There's many SoMs and SoCs to try before this one.

Intel has had to give away Atoms to get them used in any volume, this HN comment on a contemporaneous article has links on the subject: https://news.ycombinator.com/item?id=8294864

If Intel wants to hit the hobbyist embedded market running, they could sell Edisons for $10. By giving them away to heavy industry, yet selling them to hobbyists, they aren't garnering interest, they're patronizing.

I'd add STM32: Spark Core with built-in WiFi for $39.

Not sure if they will fail, depends on how big a niche there is for this kind of powerful hardware. Might have been different if it had H.264 accel. and HDMI out though.

You have to see that the ESP8266 is currently just a crappy wifi module with very low throughput (like below 100k) and limited range. I agree it will probably be very successful, never before has wifi been so cheap to add to a project, but I doubt you can transmit audio or video with it. I wonder if people will figure out how to run custom code on its MCU, that would be awesome.

I'm currently exploring the TI CC3200, it has a good price point (30$ via TI) and is quite capable (80Mhz M4 Cortex, 256k RAM) - for me its close to the sweet spot. If TI lowered the price even more, they'd own the market.

Its supposed to be >100MHz Xtensa. Should be enough for a lot of things if/when it gets hacked (as long as it has SPI for fast interconnect).

Then you have $10 modules based on 360MHz Ralink Mips soc, or $15 Qualcomm AR9331 400MHz (overclockable to ~500MHz) ones with 32/64MB ram. Those are very capable on their own.

Ah, I guess thats whats inside most Wifi routers. Are people using those for hobbyist projects though? I've never seen prototyping boards anywhere, and soldering those packages yourself requires a reflow oven by the looks of it.

Didn't know those SOCs are so high-powered, thanks for the info!

Various companies have been selling modules with these chips and all the appropriate RAM, Flash, power regulators, etc - for example, the VoCore, the Carambola, the HLK-RM04 (officially a WiFi-to-serial module but is supported by OpenWRT and breaks out most of the pins), etc.

This is like a more extreme case of saying RPi will fail because Arduino is cheaper...

Minor nitpick about the voltage: How many chips at this level (at the border of micrcontroller to microprocessor) actually use more than 1.8V for internal logic? Other than a few chips that have on board power switching from 3.3V, I can't find any other CPUs, DDR3 RAM, or even CMOS cameras (720p+ chips) that don't require a 1.8V rail.

For prototyping a divider with 1% resistors would do just fine (although with some instability with a bad breadboard) and if you can't find 1.8V power parts for production you have much bigger problems. As for level translators, I've found they are almost inevitable on any design that uses enough chips and sensors.

Impressive how much Intel can shrink their X86 processors and supporting hardware. However I think Intel is making a mistake trying to push Arduino 'compatibility' with this and the Galileo. Unless they actually put an AVR microcontroller on board it will have poor compatibility with Arduino shields (like the Galileo). You also won't have real-time control of the GPIO so making servos move smoothly, talking to 1-wire-like interfaces (like WS2811 LEDs), etc. will be problematic. Lots of folks were excited about the Galileo but unhappy after actually using it and realizing its limitations. Hopefully Intel will get the messaging a little better this time.

I think they've improved in this regard: they've added an independent microcontroller inside the core(which will probably run the arduino layer).This will solve compatibility and real time issues.

This is in general targeted to a bunch of chips that integrate a microprocessor and a microcontroller and are usually fabricated in 65nm-45nm. With it's process advantage(22nm finfet) , intel can offer orders of magnitude lower sleep currents, some nice decrease on power(depending on x86 vs arm details), and more power in general. All interesting to embedded guys.

But they still have to create a large library of peripherals and support it(which they might solve by creating a few speedy cores and let the crowd code and share/support peripherals), add analog blocks, and gain the trust of the embedded community as a reliable long term supplier(which might be the one thing intel couldn't solve - because of it's past as an unreliable embedded supplier).

The second option is that this is mainly to motivate other embedded chip companies guys to use it's fabs.

Anyway ,i'm grabbing the popcorn.

The one bit I can't find anywhere in the docs released so far. Is there a communication mechanism between the microcontroller and the main core?

It's not hard to dangle a micro off of i2c or some other bus, however, I can definitely see the the benefit of an easy prototyping solution combining an beefy (~100k RAM and 100Mhz CPU) real-time micro and a general purpose OS with wifi / bluetooth LE.

The Arduino block that SparkFun is selling[1] has an actual AVR on it that communicates with the Edison via a UART.


Fun thing, though obviously not quite the same: Transcend wifi SD cards are (or at least were) running Linux on ARM, hosting a Perl CGI script(!) that could be exploited to let you do your own thing with the card -- see http://haxit.blogspot.com/2013/08/hacking-transcend-wifi-sd-...

I went browsing around their site this morning when I got the email of the release. I cannot say that I understand Intel's strategy here. On the one hand they could release 'Quark' like Atmel ships AVR's or ST Micro ships STM32 chips with a data sheet and a demo board, on the other they seem to want to jump into some sort of "Intel at System supplier" kind of thing where its mosly proprietary interfaces and only works with their stuff.

When I look at something like the Nucleo boards[1] I see a chip company leveraging the energy around Arduino to push their own ISA, but that is just a form factor play AFAICT. What does Intel hope to achieve here? And can they do that without being a crapload more "open" than they have been in the past?

[1] http://www.mouser.com/new/stmicroelectronics/stm-nucleo-deve...

Intel's problem is that it's a lot harder to turn their chips into a complete system than it is AVRs or STM32, because Intel's solution is a full SoC running Linux that requires seperate RAM and Flash chips and external power supply circuitry that has to meet some nasty power sequencing requirements. This is basically Intel's version of the Raspberry Pi Compute Module which exists for much the same reason.

It's the SOC plus their quark microcontroller on a module. I think that's the niche they are going for - ease of programming of a general purpose OS plus real-time control via the quark.

Edison is the size of a postage stamp including DDR. The picture linked is of the module plugged into a board.

Thats great, but STM32 costs ~$.30, STM8 you can get down to ~$.10.

There is no room for x86 at the bottom. Nobody wants embedded chip that boots in legacy real addressing mode to maintain that sweet sweet compatibility with windows 3.11

Can anybody confirm whether the wireless chip used (BCM43340) is FullMAC or SoftMAC? The model is not listed on either brcmsmac nor brcmsmac. The wifi driver that comes with the source code downloaded from Intel is not in a Linux tree and it even has IP and TCP related stuff. I'm confused.

This is pretty close to something I've wanted for ages: the smallest possible Ceph OSD. This doesn't have the right set of devices on it but it's pretty close.

Ditch the wifi for gigabit ethernet, slap a sata controller on the pci-e and mount it in a stackable frame with a power bus. Buy a drive, one of these nodes and stack it on top of your existing ones. Bam your own tiny cloud you can expand in increments of one drive (and commodity ethernet). It's probably a commercial non-starter now that 8TB drives exist but I still love the idea.

Or now that I think about it skip the sata and just sell it with some NVMe flash storage built in. Stackable storage bricks.

http://www.minnowboard.org/meet-minnowboard-max/ might be closer to what you're looking for.

Hey so it is, that's awesome. Thanks!

Have you seen those Seagate HDDs with built in ethernet? Might be better to just plug all the drives in to a switch and co-ordinate a cluster of them from a controller node

So you want a small PC, not an embedded platform.

How much power does it use? In documents I read:

Standby (No radios): 13 mW Standby (Bluetooth 4.0): 21.5 mW (BTLE in Q4-14) Standby (Wi-Fi): 35 mW

It this turned ON and waiting on standby? Or is this turned OFF standby? And how much does it use when CPU is working?

That's almost certainly standby as in turned off, given the figures I've seen for previous Intel IoT boards with much lower end chips than this one.

I've looked around at all their docs and still can't find this number. Almost like they are hiding from something... I wouldn't imagine a full Atom can get anywhere close to running off a standard USB port 500 mA, like the Beaglebone Black or RPi.

Seems weird that they wouldn't at least drop it somewhere. Kinda important if you have to make a carrier board for it.

Presumably its shipping now, but I can't seem to find anywhere to buy it.

WLAN has been somewhat sorely lacking for these mini computers, your only option was pretty much the rather terrible MIPS platforms made for routers.

You have to scroll to below the list of links that look like they're a footer, and you'll see a big blue bar that shows the 'where to buy' links. It sells for 50 bucks at Sparkfun, you can also buy the kit at makershed.

Hopefully they'll send some to Farnell so we Europeans can get a taste of it as well.

Personally, I love the idea of low-spec 86 chips being available at rock-bottom prices. It's an architecture that while far from perfect is well known, and most of us here have a boatload of experience with it I'll wager. I'd really like to see something similar to this but for 2 or 3x the price (it's only $50) that includes a framebuffer, hdmi and a fully documented hardware blitter or bargain-basement GPU.

There have been low end x86s available for many years, eg the Vortex SoCs. They are still around, the company even put out an Arduino compatible board recently: http://linuxgizmos.com/arduino-compatible-boardset-runs-linu...

And SATA! Please someone give me SATA, gigabit ethernet, and a cheap/easy way to do DMA!

Wouldn't that more or less be the already available Tegra K1?

It's my understanding you need to be a multi-million-dollar company to even get NVidia to answer your calls to talk about showing you the NDA you need to sign to get access to register-level Tegra programming docs? If I'm wrong I'd be very happy. Intel has been really good at documenting their GPUs though. I want to play on the metal, running under Linux with a binary blob framebuffer driver doesn't interest me.

It's my understanding that all you have to do is develop an interesting and popular GPU application and they will bend over backwards to help you make it run efficiently on their hardware.

Compare and contrast doing so with giving them the finger on video because they won't follow your marching orders.

experience? you mean you use assembler? ever touch registers directly? no? thought so. x86 has zero advantages over other architectures in embedded world.

For people like me who are too stupid to make things with it, can I use this to replace my Raspberry Pi+RaspBMC setup? How's the performance?

There's no video core, so no HDMI or even VGA output. Edison is designed for things like robotics and home automation. It's not a general-purpose x86 computer.

Seems like a prime candidate for a unikernel like mirage (http://openmirage.org/blog/announcing-mirage-20-release)

Give it PIC-Express lanes with SATA Raid or Gigabit Ethernet Controller, it will be perfect for NAS / Router.

There are plenty of boards for that kind of application already (PC Engines APU[1], for example). That's not really what Edison is intended for.

[1]: http://www.pcengines.ch/apu.htm

I couldn't find any info, but I guess one will have to wait for adapter so Edison can output to http://en.wikipedia.org/wiki/FPD-Link that is connect Edison to most LCD monitors.

How about adding Erlang support ? It would be great to be able to program multiple devices with OTP-level of support for mesh network topologies.

Erlang is not supported on x86?

What is the price?

Scroll to the bottom of the "More for Makers" page:


at "Where to Buy"

I'd love to see Tesla Inc sue Intel saying that they stole all the tech for their Edison Module :)

And what exactly are you going to do with an "x86" expensive dev board with 500 Mhz Atom CPUs and 1GB of RAM? Run XP on it?

This is why we can't have nice things.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact