Intel abandons future Atom line of processors


Android can run on X86. This have been Atom product shipping with Android. I don't think that news.

I am not sure however how usable Android on X86 is. I see no reason for Android / the OS itself to be fine. There could be some issue but I am sure this could be eventually be fixed as Android is open course.

More concerning is whether Android application will run (as most of them are closed sources). I am not at all familiar with Android app development... my understanding is that application that simply use Java should be able to run cross platform.
However some demanding application (games...) have additional optimized code which is compiled and architecture Dependant.

Going back to the Atom... I found the Atom server processor to be really really nice alternative. They are affordable (for the server /embeeded market), 64bit, ECC support and all other feature one might need (including Virtualization AES-NI...).
 
Android can run on X86. This have been Atom product shipping with Android. I don't think that news.

I am not sure however how usable Android on X86 is.
Probably comparable to WinXP-64bit...
Working in theory, market share too small for proper software support and no enthusiast community like here.
 
More concerning is whether Android application will run (as most of them are closed sources). I am not at all familiar with Android app development... my understanding is that application that simply use Java should be able to run cross platform.
However some demanding application (games...) have additional optimized code which is compiled and architecture Dependant.

i heard... that many Android games are not in fact coded in Java, but C++ with just a Java wrapper.
 
I don't see why devs would not compile with APP_ABI := all as long as they don't have any ASM code in there. One would think that reaching as many devices as possible would be an obvious goal.
I own an x86 Android tablet and its architecture honestly hasn't gotten in the way yet (not saying there aren't locked out apps, just not that many). I don't emulate on it, though. Got something else for that.
 
I don't see why devs would not compile with APP_ABI := all as long as they don't have any ASM code in there. One would think that reaching as many devices as possible would be an obvious goal.
I own an x86 Android tablet and its architecture honestly hasn't gotten in the way yet (not saying there aren't locked out apps, just not that many). I don't emulate on it, though. Got something else for that.

Middleware might have limited arch support, I know Unity only added x86 in late 2014. I bet that was the biggest reason why so many Android apps had x86-less NDK builds so recently.

But I can see why app developers are hesitant to add support for platforms that they're not willing to test on, especially for paid apps. Even if they compile fine and should work okay who knows what arch-specific bugs could be lurking. There's just enough undefined in languages like C/C++ to make this risk real.

On the other hand, Intel has foisted binary translation on x86-less NDKs and that has both a big performance hit and a probably higher risk for bugs. So there I'd rather take my chances with the x86 NDK build. Locking out every x86 device by hand is possible but it's a huge endeavor.
 
Not dead yet - just going embedded?
http://liliputing.com/intel-atom-isnt-dead-next-gen-chips-aimed-embedded-iot-systems/

Although it appears Intel may be abandoning the market for tablets and phones, they may still be wanting to compete for embedded systems using more or less the same capabilities?

"Intel says the processors will use the latest version of the company’s Gen 9 graphics and Intel also plans to offer new developer tools in a “Media SDK for embedded Linux for IoT.”"

Wouldn't that be about the same type of intention and support that TI had for the OMAP line? If they're going to concentrate the Atom X86 implemetations to embeded Linux in IoT devices - that may imply lower pricing per thousand too. Just speculating, but if they're going to envision this being used in microwave ovens, fridges, cars, etc - it would almost have to be, right?

But - that may also imply that they woiuld remove it from general availability so it may or may not be available for projects like this one. Still - interesting turn for the subject.
 
I don't really understand why you need 'Gen 9' graphics on a fridge or a microwave, but perhaps that's a lack of vision on my part. For automotive uses these days, I guess it makes sense though, but I don't really understand how those requirements differ from those for a tablet, given we're basically talking about in car tablets.
 
I don't really understand why you need 'Gen 9' graphics on a fridge or a microwave, but perhaps that's a lack of vision on my part.
Because everything is better with pretty graphics! ;)

They can slap the "smart" adjective to products and wow the unsuspecting consumer.
 
because ......
2hf7z93.png
 
I don't really understand why you need 'Gen 9' graphics on a fridge or a microwave, but perhaps that's a lack of vision on my part.
for the microwave: watch a short movie that is exactly as long as the time your food needs to warm up :D staring at the food gets old after ten seconds.
other stuff: watch an ad for the food you're about to consume, chance for a comeback for all those mini games on tiny cds that could be found in cereal packs, use presets for heating up food by displaying the actual food and not funny abbreviations (like "bccl" for broccoli), watch tutorials for heating up food correctly (no kidding, these exist), being able to read the manual on the device's screen, use the thing as a digital photo frame, use it as a video de/encoder, hack your neighbor's wifi with it, etc.
 
For cost reasons you want to keep the diversity in chips minimal.
Within the semiconductor industry it is all about volume. Variations are very costly (setup/change-over of production lines).

Keep in mind that within each generation there are variations from weak GPU circuitry to powerful as well.
Each generation drives changes in architecture and also the etching process.
Gen9 is etched with a critical dimension of 14nm, but that does not indicate the number of execution units, Ram etc...

So where cars will probably have high-end Gen9 GPUs on chip, your coffee maker will take the low-end GPU.
 
lack of vision on my part
Well, for the first generations the fridge display will be a gimmick, with very limited functionality. Once IoT starts, then you can picture this:

Open your fridge to get an egg. The amount of eggs is updated, an if-this-then-that rule triggers, and "eggs" is added to the grocery list. Which is updated to your parents phones, so they can buy it. Need something else? just swipe it into the grocery list.
stick an rfid to the bottom of your milk cardboard. put it to the display, identify it as milk. Now the fridge knows how much milk it has (due to digital scales that can identify the weight of things).
Synch all of your agenda's to the fridge. You can see what each person will do today, and when they estimate to get home.
Stream commercials 24-7 to your fridge, get some cents for each commercial streamed. Watch news headlines. put your phone nearby to load the news into it.
Going to be late? post a message from your phone, and it will display as a digital post-it note.
hack your neighbours fridge to display a giant d*ck on it.
hack your neighbours fridge to buy milk, and see a van come and bring them milk, which they already payed for, but do not need.
etc.
 
A barcode scanner, Raspberry Pi, some software and a bit of discipline is all you need to realize 90% of that scenario without all the negatives (commercials, hacking, privacy violations, etc...).

I realize that in a way this is the Blub Paradox in action.... I suppose I am getting old.
 
A barcode scanner, Raspberry Pi, some software and a bit of discipline is all you need to realize 90% of that scenario without all the negatives (commercials, hacking, privacy violations, etc...).

I realize that in a way this is the Blub Paradox in action.... I suppose I am getting old.
Anything that increases the cumulative cognitive load of mundane tasks is not a good direction, IMO. It should be as automatic as possible or not at all. Having to scan barcodes for the sake of keeping tabs on what's in a fridge sounds like more trouble than it's worth.
 
Nice. There is a "bzar's first law of automation" hidden in there :)
That said, "as automatic as possible or not at all" sounds like a partial contradiction.

I would say that the weakness in my alternative lies in the need for process discipline (i.e. scan when you put in, scan when you discard) to maintain proper inventory.
It is a stateful interaction with a high risk of loss of state and an expensive recovery procedure for that state. This is probably why barcode scanning fridges never took off.

The stateless process of "look in my fridge to see what is missing and needs to be bought" is thus way easier.

Hmmm... interesting observation.
 
Yes :)

No, that is where the software part comes in. For example, you can classify stuff as "beef", "pork", "chicken" or just throw them all together into a class "meat".
In terms of such categories, I suspect your fridge does not vary all that much.
 
Yes :)

No, that is where the software part comes in. For example, you can classify stuff as "beef", "pork", "chicken" or just throw them all together into a class "meat".
In terms of such categories, I suspect your fridge does not vary all that much.
I guess you've never seen a vegan's fridge from the inside. :D
It's usually ever changing and quite colourful.
 
I guess you've never seen a vegan's fridge from the inside. :D
I won't surprise you that I have not.

That said, the meat was just an example. I assume that as a vegan you have to pay even more care to the nutritional value of your food.
You could adjust the categories to those nutritional values and automation might actually give you insight into whether or not you are on a healthy diet or having shortages.

That said, productisation of such small and insignificant markets is unlikely, especially in the early stages of the IoT.
 
Back
Top