That's too bad, the Edison was a nice small developer board. The NUC is quite a few steps up and not quite for the same target market (I've got a few). I wonder if they're going to continue attempting to compete with ARM or they have just realized they lost the low end battle with x86
The problem with their boards is that they tried to create an ecosystem for makers, but it was way too expensive for hobby budgets. All the components they sold plug & play are also available in a plug&play format for Arduino, and they're much cheaper.
My friends have won a bunch of those boards at a hackaton, and they've been trying to sell them for the past year, to no avail. People just don't want those things.
I don't think x86/PC comparability is irrelevant. ARM is not an architecture. It's just a SoC where manufactures hook random crap to random pins and make patched to hell, non-upstreamable kernels:
Windows Mobile ARM at least required UEFI, but their bootloaders are locked. Most mobile phones don't support device tree. Even on ARM boards that support device tree, hardware support is still hit and miss:
My remark about it being irrelevant is due to the fact that for IoT applications, one doesn't care for backwards compatibility of existing applications.
IoT deployments are usually software developed for a special use case.
Then if the target platform is powerful enough to allow C, C++, Rust, Java, Lua, MicroPython, Pascal, Basic, <whatever language with rich library>, then the actual OS is also kind of irrelevant.
I am not thinking of boards to run GNU/Linux or Windows, mimicking a desktop experience.
Maybe it shows my 80's background, but for many use cases an Arduino like bare metal development is more than good enough, hence x86 being irrelevant when one has an high level language with a nice abstractions SDK.
If software compatibility across a broad range of devices becomes important, then the ARM ecosystem will just move to broader adoption of device trees. Switching architectures to x86 just to get a standardized platform doesn't make sense.
I'd love to have small, low power, easy to setup, x86 chips or modules-with-BIOS, in the 40-200 MHz range, with 486DX/DX4/early Pentiums instructions and similar or better performance; possibly with a few basic SoC functions (UARTs, simple VESA-like graphics (no unusable 3D acceleration), integrated memory). And not some Chinese NDA doc-less chips like Vortex.
The NUC is not really for the same space - it's a desktop/server thing, totally unsuitable for embedded.
It's really a shame. I don't think Intel needs to lose the low end - they have the technology, but lack the will. Their mobile parts would work fine on an RPi-class board and the architecture would be far more cohesive.
It may not be in exactly the same space, but it has the same problem. It's way overpriced for what it offers. I still don't understand why they cost more than a full tower PC with better specs. It's not like they're some crazy special hardware.
The real problem here is that embedded is a too broad term most of the time as it defines a specific way of using a computer, not its capabilities, size or power constraints, etc. You can put a desktop tower into a vending machine and still call it embedded (and if its a big machine it may actually be an OK solution).
I chose embedded systems as my specialization during my CS degree. We spent a week in one class just trying to define what exactly an embedded computer is.
Your definition is pretty much what we came up with--they're defined in terms of use. But at the end of the day it's one of those "I know it when I see it" things.
