Not a big surprise, given the myriad of IoT ARM (and MIPS, an upcoming RISC-V) microcontrollers available on the market at every price point and capability already.
The lack of documentation is what really killed the Quark MCUs. The Maker and R&D communities were excited about these when they first came out but quickly realized the documentation and support was just non-existent. Hard to program them when you can't even figure out which registers you need.
Everyone needs to learn from the Raspberry Pi. Its not the fastest but its by far the most well documented and that trumps just about everything else. A Ferrari's not going anywhere without the key! Even a scooter will beat it off the line in that case.
This. Plenty of faster alternatives for the Pi but having an active development community around it makes Pi the obvious choice most times. And an active community needs active support from the manufacturer's side. An area in which Intel was always lacking.
Are you really serious about active support - maybe not from hacker community like Pi and Adreno, but all notebook manufactures for Windows PC and also Apple.
Pi is good to learn embedded system - but I know of no profession product that use Pi
I have seen several companies installing RPi boards inside of thin clients or zero clients for remote access. If you scrounge around the Internet you'll see them. Despite that, I agree that the Pi is generally not a commercial product nor was it initially intended for such uses.
My guess is that some one at Intel thought they could get into market of such devices and decided lack of use by people was not worth the effort. And decided to focus on commercial products.
What is interesting, is that Quark D2000 is listed on Intel Ark site as still available
Intel has active support for Core CPUs, not for Quark MCUs or any other side project of theirs. The point of the Pi is that it was widely adopted by its target audience because of the documentation. Whatever the target market the lack of docs killed the Quark. Broadcom sells lots of SoCs for products because the docs for the RPi still apply when used on a custom circuit board. That's millions of units moved due to good documentation.
A cautionary lesson, from both this and 2008: Don't trust Intel's commitment to embedded, ever. If you're going to build a product with long availability requirements, or want to move your code forward across multiple generations unmodified, Intel is not a safe bet. At least with ARM, you can move your assembly code and intrinsic-using C to another vendor, as there are dozens; as far as I know, the only deep-embedded x86 vendor remaining is DMP (and maybe RDC?)
Just my opinion. It would be a cold day in hell before I trusted an Intel embedded part for a design I knew was going to be around for a long time.
In my opinion, this whole market is not very trustful - Raspberry PI and especial Adreno ( which is primary started ) of home built electronics devices - but I think the big reason is for both Amazon Alexa and Google Home devices have pretty much taken over the market.
Besides Amazon and Google - parts most of this stuff - is for up and coming developers to experiment with - done some playing rom - no way it actually replace a setup and definitely not a computer. But it is cool to experiment with. I just think it too limited market for Intel to invest in.
For future comments, I'd focus on not using dashes at all... Makes your comment super hard to read and it still hardly makes sense beyond a general level. Sorry for criticism if English isnt your first language.
Don't be so critical, I am an Asperger and likely even a Savant in nature. A problem solver professional in development of software applications with 30 years experience in computers. English writing has never been one of my high ends. At Georgia Tech, I was required to take a class with football players in English. A lot of this is because my brain thinks faster than I write. Modern web sites have word processors to help with my thought process and usually I will come back and edit something. But unfortunately AnandTech forums do not allow you edit your posts and sorry if that confuses you. By the way, except for this "-", I did not use a dash.
The emerging open and free RISC-V ISA will gradually cater to an ARM business. Its openness and royalty-free nature helps to move from one design to another without changing compiled code, lets you compile your own RISC-V compliant (or not, or enhanced) cores into an FPGA. Founders of the ISA started to manufacture chips based on customers' requirements feature wise. So you can let them manufacture a chip for you uniquely suiting your design, including custom instructions and extensions.
The best of it is complete portability across the spectrum of RISC-V (compliant) chips AND manufacturers. In addition, you can manufacture your dream embedded CPU in the future without any failing company - you're not locked to any. The ability to test own design on an FPGA and then making a chip of it is truly a blessing. And the ISA and it's implementations are very efficient, because a lot of work went into optimizing the ISA for low power, low decode complexity and small code size.
I'm curious how long it will take to really conquer the world. nVidia and WD seem to be jumping on it for their embedded controllers now...
As long as I like the RISC-V idea of a "open source" microcontroller ISA, I would not say that portability is the best of its features when custom enhancements can be inserted into it. I mean, portability and customization aren't compatible with each other. If I create a library that exploits a particular custom instruction to shorten computation times, it won't be usable on anything else that does not have the same added instruction. If you count only on the "legacy" part of the ISA, then you are not exploiting its customization feature. RISC-V idea is very good and promising, but as said, customization may fragment the market, especially the embedded one when customization is mostly required.
Not related because Microsoft never mentioned support for Quark for it's Windows 10 IoT initiatives. However, I predict we will see a hasty withdraw of Windows IoT Core this year. Almost a year on and we don't even have proper support for the Raspberry Pi 3+, a minor upgrade to the Raspberry Pi 3. Can't trust Microsoft's commitment to anything that doesn't have an 8" screen or bigger.
Windows IoT Core was really exciting news when it was on its way out the door, but yes, on-going support is lacking to say the least. The RPi is not exactly an uncommon platform and there are even a few commercial vendors putting them to use as thin/remote clients so they're out in sufficient quantities.
They just don't produce the chip, it is not like they are done with those IP. I think some form of it lives in the current 4G Modem that Apple is using.
And at the end this is another branch were Intel has failed to use its x86 architecture. No GPU (Larrabee), no HPC (Xeon Phi, recycle of the previous failure), no smartphone, being replaced also in other embedded devices like NAS, it becomes clear (also for those that have not understood this u to now) that anything else that is not desktop/server and has a high end OS on it doesn't fit x86 requirements as there are many other better solutions.
The question is, how long it will take for these other solutions (that are expanding in use and investments) to corrode its main core business. The (slow) process has already started. We'll see when it will become an issue for Intel.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
27 Comments
Back to Article
Thatguy97 - Tuesday, January 22, 2019 - link
is this suprising or no?deil - Tuesday, January 22, 2019 - link
nah.psychobriggsy - Tuesday, January 22, 2019 - link
Not a big surprise, given the myriad of IoT ARM (and MIPS, an upcoming RISC-V) microcontrollers available on the market at every price point and capability already.Ikefu - Tuesday, January 22, 2019 - link
The lack of documentation is what really killed the Quark MCUs. The Maker and R&D communities were excited about these when they first came out but quickly realized the documentation and support was just non-existent. Hard to program them when you can't even figure out which registers you need.Everyone needs to learn from the Raspberry Pi. Its not the fastest but its by far the most well documented and that trumps just about everything else. A Ferrari's not going anywhere without the key! Even a scooter will beat it off the line in that case.
close - Tuesday, January 22, 2019 - link
This. Plenty of faster alternatives for the Pi but having an active development community around it makes Pi the obvious choice most times. And an active community needs active support from the manufacturer's side. An area in which Intel was always lacking.HStewart - Tuesday, January 22, 2019 - link
Are you really serious about active support - maybe not from hacker community like Pi and Adreno, but all notebook manufactures for Windows PC and also Apple.Pi is good to learn embedded system - but I know of no profession product that use Pi
PeachNCream - Tuesday, January 22, 2019 - link
I have seen several companies installing RPi boards inside of thin clients or zero clients for remote access. If you scrounge around the Internet you'll see them. Despite that, I agree that the Pi is generally not a commercial product nor was it initially intended for such uses.HStewart - Tuesday, January 22, 2019 - link
My guess is that some one at Intel thought they could get into market of such devices and decided lack of use by people was not worth the effort. And decided to focus on commercial products.What is interesting, is that Quark D2000 is listed on Intel Ark site as still available
https://ark.intel.com/products/91947/Intel-Quark-M...
Ikefu - Wednesday, January 23, 2019 - link
Intel has active support for Core CPUs, not for Quark MCUs or any other side project of theirs. The point of the Pi is that it was widely adopted by its target audience because of the documentation. Whatever the target market the lack of docs killed the Quark. Broadcom sells lots of SoCs for products because the docs for the RPi still apply when used on a custom circuit board. That's millions of units moved due to good documentation.HStewart - Tuesday, January 22, 2019 - link
I think this is actually Intel getting rid of some of fat - possibly they a better solution in the future.SarahKerrigan - Tuesday, January 22, 2019 - link
A cautionary lesson, from both this and 2008: Don't trust Intel's commitment to embedded, ever. If you're going to build a product with long availability requirements, or want to move your code forward across multiple generations unmodified, Intel is not a safe bet. At least with ARM, you can move your assembly code and intrinsic-using C to another vendor, as there are dozens; as far as I know, the only deep-embedded x86 vendor remaining is DMP (and maybe RDC?)Just my opinion. It would be a cold day in hell before I trusted an Intel embedded part for a design I knew was going to be around for a long time.
HStewart - Tuesday, January 22, 2019 - link
In my opinion, this whole market is not very trustful - Raspberry PI and especial Adreno ( which is primary started ) of home built electronics devices - but I think the big reason is for both Amazon Alexa and Google Home devices have pretty much taken over the market.Besides Amazon and Google - parts most of this stuff - is for up and coming developers to experiment with - done some playing rom - no way it actually replace a setup and definitely not a computer. But it is cool to experiment with. I just think it too limited market for Intel to invest in.
thesavvymage - Tuesday, January 22, 2019 - link
For future comments, I'd focus on not using dashes at all... Makes your comment super hard to read and it still hardly makes sense beyond a general level. Sorry for criticism if English isnt your first language.HStewart - Tuesday, January 22, 2019 - link
Don't be so critical, I am an Asperger and likely even a Savant in nature. A problem solver professional in development of software applications with 30 years experience in computers. English writing has never been one of my high ends. At Georgia Tech, I was required to take a class with football players in English. A lot of this is because my brain thinks faster than I write. Modern web sites have word processors to help with my thought process and usually I will come back and edit something. But unfortunately AnandTech forums do not allow you edit your posts and sorry if that confuses you. By the way, except for this "-", I did not use a dash.mode_13h - Tuesday, January 22, 2019 - link
@the (questionably) savvy mage didn't say not to use paragraphs. No one likes a wall of text.mode_13h - Tuesday, January 22, 2019 - link
It's sure better than the same text sans dashes.IMO, there's nothing wrong with dashes. Just don't go crazy with them.
edzieba - Tuesday, January 22, 2019 - link
2013-2022 is not a bad lifecycle for Lakemont. And don't be so quick to assume ARM code is easily portable between vendors, for That Way Madness Lies.SarahKerrigan - Tuesday, January 22, 2019 - link
I'm not assuming. I've done it. Writing an optimized lib for Cortex-M3 Vendor A is something that carries over near-directly to Cortex-M3 Vendor B.lada - Tuesday, January 22, 2019 - link
The emerging open and free RISC-V ISA will gradually cater to an ARM business. Its openness and royalty-free nature helps to move from one design to another without changing compiled code, lets you compile your own RISC-V compliant (or not, or enhanced) cores into an FPGA. Founders of the ISA started to manufacture chips based on customers' requirements feature wise. So you can let them manufacture a chip for you uniquely suiting your design, including custom instructions and extensions.The best of it is complete portability across the spectrum of RISC-V (compliant) chips AND manufacturers. In addition, you can manufacture your dream embedded CPU in the future without any failing company - you're not locked to any. The ability to test own design on an FPGA and then making a chip of it is truly a blessing. And the ISA and it's implementations are very efficient, because a lot of work went into optimizing the ISA for low power, low decode complexity and small code size.
I'm curious how long it will take to really conquer the world. nVidia and WD seem to be jumping on it for their embedded controllers now...
CiccioB - Wednesday, January 23, 2019 - link
As long as I like the RISC-V idea of a "open source" microcontroller ISA, I would not say that portability is the best of its features when custom enhancements can be inserted into it.I mean, portability and customization aren't compatible with each other.
If I create a library that exploits a particular custom instruction to shorten computation times, it won't be usable on anything else that does not have the same added instruction. If you count only on the "legacy" part of the ISA, then you are not exploiting its customization feature.
RISC-V idea is very good and promising, but as said, customization may fragment the market, especially the embedded one when customization is mostly required.
Silma - Tuesday, January 22, 2019 - link
Source ?Silma - Tuesday, January 22, 2019 - link
Sorry, you can delete this comment and the parent comment.bji - Tuesday, January 22, 2019 - link
Ha ha! As if Anandtech would ever curate their comments or even provide a reasonable comment system allowing you to do this yourself.HardwareDufus - Tuesday, January 22, 2019 - link
Not related because Microsoft never mentioned support for Quark for it's Windows 10 IoT initiatives. However, I predict we will see a hasty withdraw of Windows IoT Core this year. Almost a year on and we don't even have proper support for the Raspberry Pi 3+, a minor upgrade to the Raspberry Pi 3. Can't trust Microsoft's commitment to anything that doesn't have an 8" screen or bigger.PeachNCream - Tuesday, January 22, 2019 - link
Windows IoT Core was really exciting news when it was on its way out the door, but yes, on-going support is lacking to say the least. The RPi is not exactly an uncommon platform and there are even a few commercial vendors putting them to use as thin/remote clients so they're out in sufficient quantities.iwod - Wednesday, January 23, 2019 - link
They just don't produce the chip, it is not like they are done with those IP. I think some form of it lives in the current 4G Modem that Apple is using.CiccioB - Wednesday, January 23, 2019 - link
And at the end this is another branch were Intel has failed to use its x86 architecture.No GPU (Larrabee), no HPC (Xeon Phi, recycle of the previous failure), no smartphone, being replaced also in other embedded devices like NAS, it becomes clear (also for those that have not understood this u to now) that anything else that is not desktop/server and has a high end OS on it doesn't fit x86 requirements as there are many other better solutions.
The question is, how long it will take for these other solutions (that are expanding in use and investments) to corrode its main core business. The (slow) process has already started. We'll see when it will become an issue for Intel.