Author Topic: The $1 Linux-capable ARM processor in TQFP...  (Read 22649 times)

0 Members and 1 Guest are viewing this topic.

Offline technixTopic starter

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
The $1 Linux-capable ARM processor in TQFP...
« on: September 12, 2018, 05:55:45 pm »
I have been evaluating multiple ARM-based processors, and one caught my eye again for a different reason: price. There are only a few ARM processors in TQFP package that is capable of running Linux, and one of which costs only $1 each: Allwinner A13. That is a 1GHz ARM Cortex-A8 processor with built-in Mali GPU, resistive touch panel support, and a maximum supported memory of 512MB DDR3.

I repeat, it is a $1 chip capable of running Linux.

It is pretty ironic when its memory chip costs $4 or more. Usually the processor is more expensive than the processor, but in this case things are the other way around somehow.
 

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 5974
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #1 on: September 12, 2018, 06:04:07 pm »
Despite the price is quite astonishing, to be honest I am not terribly impressed. Maybe they provide complete support for their customers, but to me the most expensive part of these projects is having a workable linux distro with decent support for a device's peripherals. Last time I tried to evaluate an AllWinner part, I found out they were quite closed and typically had one (maybe two) working linux distro. The terrible datasheet did not help wonders either.
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 

Offline technixTopic starter

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #2 on: September 12, 2018, 06:23:20 pm »
Despite the price is quite astonishing, to be honest I am not terribly impressed. Maybe they provide complete support for their customers, but to me the most expensive part of these projects is having a workable linux distro with decent support for a device's peripherals. Last time I tried to evaluate an AllWinner part, I found out they were quite closed and typically had one (maybe two) working linux distro. The terrible datasheet did not help wonders either.
AFAIK that is an older chip that has more community support and mainline drivers available. Allwinner is a difficult company to work with as a SME or hobbyists (as do most Chinese chip manufacturers, although UniIC and Pango seemed like quite an exception but they all have the same parent company that has a strong academic background) but they do have cheap enough chips to warrant some effort.

I have investigated other manufacturers:
* Hisilicon and Mediatek flat out refused to sell me anything, and they don't have publicly available documentation on anything they made.
* Allwinner don't provide docs either, but someone else have uploaded the docs so work can be done from those docs and independent resources.
* Microchip's processors are all fairly weak, be it SAM9, SAM A5 or PIC32MZ DA.
* NXP does have good chips with public documents, but they are all expensive.
 
The following users thanked this post: rsjsouza

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 5974
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #3 on: September 12, 2018, 06:47:32 pm »
Thanks for the other references; the unit price is usually moot when talking about hobby/academia/low production volume, therefore the strongest factor of AllWinner ends up being lost.

Since you are going to the more established manufacturers (and higher $$$ :) ), the most capable and open architecture I see in the 1GHz class is the BeagleBone ecosystem. You can buy a single part and the documentation, distros and SDK are quite up-to-date. It also has heaps of add-on HW.
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2752
  • Country: us
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #4 on: September 12, 2018, 07:01:44 pm »
I have been evaluating multiple ARM-based processors, and one caught my eye again for a different reason: price. There are only a few ARM processors in TQFP package that is capable of running Linux, and one of which costs only $1 each: Allwinner A13. That is a 1GHz ARM Cortex-A8 processor with built-in Mali GPU, resistive touch panel support, and a maximum supported memory of 512MB DDR3.

I repeat, it is a $1 chip capable of running Linux.

It is pretty ironic when its memory chip costs $4 or more. Usually the processor is more expensive than the processor, but in this case things are the other way around somehow.
Very interesting, but how much available I/O does it have?  I've been using the Beagle Bone, and love the features of that board.  I have to buy the board, as the processor is a BGA, and I'm not set up to deal with that.  Does somebody make a generic board for this?  I'd want to have at least 16, say, GPIOs plus network interface.

Jon
 

Offline stmdude

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: se
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #5 on: September 12, 2018, 07:05:13 pm »
The AllWinner A13 is basically an AllWinner A10, but with a bunch of integrated peripherals removed, such as HDMI, SATA, I2S, etc.

The only use I see for it is if you have an product that works with the reduced number of peripherals, _and_ can live in the integrated 64MB of RAM. The chip being TQFP is kind of moot if you need to put a BGA-96 DDR chip next to it.

It's certainly doable to do a product with the AW chips. The datasheets are out there, and the mainline kernel (through the sunxi linux efforts) supports a lot of functionality in the chips.

For the more "established" (read: western) manufacturers, I've evaluated TIs Sitara (aka, beaglebone, aka, Octavo), but didn't like their codebase. Quite old and hacky. Also, only single-core, and no upgrade route.

I've settled for the NXP iMX6ULL chips (fairly cheap) for more price sensitive designs, and iMX6Dual/Quad for more powerful designs that require 3D acceleration. It allows me to be a bit more scalable than I would have been with TI, and allows me to re-use my know-how and software base.
 

Offline stmdude

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: se
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #6 on: September 12, 2018, 07:09:57 pm »
I've been using the Beagle Bone, and love the features of that board.  I have to buy the board, as the processor is a BGA, and I'm not set up to deal with that.  Does somebody make a generic board for this?  I'd want to have at least 16, say, GPIOs plus network interface.

You should probably look at the pocketbeagle first. If that doesn't fit you, have a look at the Octavo Systems SiPs. It's basically a beaglebone in a chip. IIRC, they require 3.3V+5V, a crystal, and a few passives to give you a booting linux system. All the highspeed black magic is done within the chip itself.

It's BGA though, but with a massive 1mm pitch, so it's no issue with using $2 chinese PCBs with it. However, I'd really recommend getting a reflow oven to solder it. The chip is so large that a hot-air gun doesn't seem to be able to reflow all the balls at once (I've tried).
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13679
  • Country: gb
    • Mike's Electric Stuff
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #7 on: September 12, 2018, 08:51:58 pm »
It's BGA though, but with a massive 1mm pitch, so it's no issue with using $2 chinese PCBs with it. However, I'd really recommend getting a reflow oven to solder it. The chip is so large that a hot-air gun doesn't seem to be able to reflow all the balls at once (I've tried).
A hotplate (/skillet)  may be another option
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline technixTopic starter

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #8 on: September 13, 2018, 03:34:41 am »
I have been evaluating multiple ARM-based processors, and one caught my eye again for a different reason: price. There are only a few ARM processors in TQFP package that is capable of running Linux, and one of which costs only $1 each: Allwinner A13. That is a 1GHz ARM Cortex-A8 processor with built-in Mali GPU, resistive touch panel support, and a maximum supported memory of 512MB DDR3.

I repeat, it is a $1 chip capable of running Linux.

It is pretty ironic when its memory chip costs $4 or more. Usually the processor is more expensive than the processor, but in this case things are the other way around somehow.
Very interesting, but how much available I/O does it have?  I've been using the Beagle Bone, and love the features of that board.  I have to buy the board, as the processor is a BGA, and I'm not set up to deal with that.  Does somebody make a generic board for this?  I'd want to have at least 16, say, GPIOs plus network interface.

Jon
It has about 40 or so GPIO multiplexed with various other functionalities after you has a basic booting board, you get 21 if you don't use the display output.

The only use I see for it is if you have an product that works with the reduced number of peripherals, _and_ can live in the integrated 64MB of RAM. The chip being TQFP is kind of moot if you need to put a BGA-96 DDR chip next to it.
The DDR SDRAM chip is the only BGA required, and if you are okay with wasting 3/4 of the capacity you can go with a SODIMM memory module and use only 1/4 of its byte lanes. And that chip maxes out at 512MB not 64.

For the more "established" (read: western) manufacturers, I've evaluated TIs Sitara (aka, beaglebone, aka, Octavo), but didn't like their codebase. Quite old and hacky. Also, only single-core, and no upgrade route.
TI suffers from the same problem as Microchip: weak processors.
 

Offline stmdude

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: se
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #9 on: September 13, 2018, 06:57:36 am »
The DDR SDRAM chip is the only BGA required, and if you are okay with wasting 3/4 of the capacity you can go with a SODIMM memory module and use only 1/4 of its byte lanes. And that chip maxes out at 512MB not 64.

Yes, it maxes out with 512MB, but it has 64MB _built in_, so you can get away without the BGA DDR if you can live with 64MB.
 

Offline chickenHeadKnob

  • Super Contributor
  • ***
  • Posts: 1054
  • Country: ca
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #10 on: September 13, 2018, 07:06:09 am »
The AllWinner A13 is basically an AllWinner A10, but with a bunch of integrated peripherals removed, such as HDMI, SATA, I2S, etc.

The only use I see for it is if you have an product that works with the reduced number of peripherals, _and_ can live in the integrated 64MB of RAM. The chip being TQFP is kind of moot if you need to put a BGA-96 DDR chip next to it.

I think you have conflated A13 with V3 and V3s. The V3 has the internal 64MB ram, small for linux but huge for a roll your own RTOS, which I considered once apon a time.
 
A13 has 48K static ram, external 16 bit sdram bus, (as well as the typical ARM 32k i-cache, 32k d-cache, and 256k L2 cache)
I agree with the sentiment that the need for  external memory and all of the other draw backs you get with AllWinner quenches the interest. :-X
 
The following users thanked this post: rsjsouza

Offline rasteri

  • Contributor
  • Posts: 17
  • Country: gb
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #11 on: September 13, 2018, 10:23:19 am »
Olimex sells an A13-based board with 256/512MB DDR3, €13.50 in 100-off quantity...

https://www.olimex.com/Products/SOM/A13/A13-SOM-256/

Assuming software support isn't completely terrible, this is a VERY good alternative to the Raspberry Pi compute module in a project I have coming up.
 

Offline technixTopic starter

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #12 on: September 13, 2018, 04:19:57 pm »
Olimex sells an A13-based board with 256/512MB DDR3, €13.50 in 100-off quantity...

https://www.olimex.com/Products/SOM/A13/A13-SOM-256/

Assuming software support isn't completely terrible, this is a VERY good alternative to the Raspberry Pi compute module in a project I have coming up.
I am actually rolling my own A13 module with 512MB SDRAM. I am actually trying to build this board with exclusively made in China chips: Allwinner V3s SoC, XPower AXP209 PMIC, UniIC DDR3 SDRAM, etc. (It is just within this year that a Chinese chip fab, UniIC, backed by Tsinghua University, finally managed to make DDR3 SDRAM chips. They are now trying DDR4.)
 
The following users thanked this post: kehribar

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 14230
  • Country: fr
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #13 on: September 18, 2018, 06:53:25 pm »
Despite the price is quite astonishing, to be honest I am not terribly impressed. Maybe they provide complete support for their customers, but to me the most expensive part of these projects is having a workable linux distro with decent support for a device's peripherals. Last time I tried to evaluate an AllWinner part, I found out they were quite closed and typically had one (maybe two) working linux distro. The terrible datasheet did not help wonders either.

I agree with that point. Which makes those chips attractive ONLY if you target very high volumes.

Software is actually also where all those SBCs (some of which are very cheap) fall short. There are dozens of such small boards which look great spec-wise but turn out unusable in a lot of cases due to the lack of proper software support.

 
The following users thanked this post: rsjsouza

Offline technixTopic starter

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #14 on: September 18, 2018, 08:00:20 pm »
Despite the price is quite astonishing, to be honest I am not terribly impressed. Maybe they provide complete support for their customers, but to me the most expensive part of these projects is having a workable linux distro with decent support for a device's peripherals. Last time I tried to evaluate an AllWinner part, I found out they were quite closed and typically had one (maybe two) working linux distro. The terrible datasheet did not help wonders either.

I agree with that point. Which makes those chips attractive ONLY if you target very high volumes.

Software is actually also where all those SBCs (some of which are very cheap) fall short. There are dozens of such small boards which look great spec-wise but turn out unusable in a lot of cases due to the lack of proper software support.
I has a fairly nice ARM Cortex-A53 board collecting dust, simply because of the poor software support. One point: while the board comes with an unpopulated eMMC and SPI Flash footprints, there is no support of using those in the accompanying software at all, let alone boot from them, although the underlying chipset is capable of doing so. I have hacked my board with both 16MB SPI Flash and 16GB eMMC, none of them work at all, and the chip maker flat out refuses to respond to my documentation and open source code requests (that means I was asking for GPL-licensed software Linux kernel and u-boot.)
 
The following users thanked this post: rsjsouza

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 5974
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #15 on: September 18, 2018, 08:50:05 pm »
Despite the price is quite astonishing, to be honest I am not terribly impressed. Maybe they provide complete support for their customers, but to me the most expensive part of these projects is having a workable linux distro with decent support for a device's peripherals. Last time I tried to evaluate an AllWinner part, I found out they were quite closed and typically had one (maybe two) working linux distro. The terrible datasheet did not help wonders either.

I agree with that point. Which makes those chips attractive ONLY if you target very high volumes.

Software is actually also where all those SBCs (some of which are very cheap) fall short. There are dozens of such small boards which look great spec-wise but turn out unusable in a lot of cases due to the lack of proper software support.
I has a fairly nice ARM Cortex-A53 board collecting dust, simply because of the poor software support. One point: while the board comes with an unpopulated eMMC and SPI Flash footprints, there is no support of using those in the accompanying software at all, let alone boot from them, although the underlying chipset is capable of doing so. I have hacked my board with both 16MB SPI Flash and 16GB eMMC, none of them work at all, and the chip maker flat out refuses to respond to my documentation and open source code requests (that means I was asking for GPL-licensed software Linux kernel and u-boot.)
That was my observation as well; sometimes the basic distro works well with whatever is put on the board but not with what the device itself is capable. To add to your list, I have seen inoperative or terribly non-optimized QSPI, USB OTG, SATA, etc.
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 

Offline hermit

  • Frequent Contributor
  • **
  • Posts: 482
  • Country: us
 

Offline LapTop006

  • Supporter
  • ****
  • Posts: 467
  • Country: au
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #17 on: September 19, 2018, 02:28:30 am »
I've been working with the packaged up Octavo SiP used in the PocketBeagle, it's as hand-solderable as a BGA ever gets, but is rather easier to use from the built-in components.

Rather more expensive though.
 

Offline technixTopic starter

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #18 on: September 19, 2018, 03:00:45 am »
:palm:
https://hardware.slashdot.org/story/18/09/18/183237/a-1-linux-capable-hand-solderable-processor#comments
Why is this deemed Slashdot- and Hackaday-worthy?

Rather more expensive though.
That is the reason why I never considered Octavo: can't quite afford it.
 

Offline donotdespisethesnake

  • Super Contributor
  • ***
  • Posts: 1093
  • Country: gb
  • Embedded stuff
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #19 on: September 19, 2018, 06:19:09 am »
Why is this deemed Slashdot- and Hackaday-worthy?

Every site needs some clickbait, it keeps the lights on.

It was worth reading for the reply from Olimex though.
Bob
"All you said is just a bunch of opinions."
 

Offline technixTopic starter

  • Super Contributor
  • ***
  • Posts: 3507
  • Country: cn
  • From Shanghai With Love
    • My Untitled Blog
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #20 on: September 19, 2018, 09:22:53 am »
Why is this deemed Slashdot- and Hackaday-worthy?

Every site needs some clickbait, it keeps the lights on.

It was worth reading for the reply from Olimex though.
Sad to see they even failed some research on the chip though. A13 is R8, which is used in the $9 CHIP board. My version of A13 base module isn’t too dissimilar from CHIP either (although I do have 512MB DDR3 and 32GB eMMC, a step up from the 4GB/8GB NAND.)
 

Offline MT

  • Super Contributor
  • ***
  • Posts: 1616
  • Country: aq
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #21 on: September 19, 2018, 12:05:43 pm »
It always puzzled me why all those 0.99cent Linux SOC manufacturers so closed and secret of the SOC since lifetime is so short and they basically do the same thing, namely run Linux on a small plattform?! I hardly imagine there is so much IP
in there thats making them so much more "profiled" then next in line SOC manufacturer.
« Last Edit: September 19, 2018, 12:12:57 pm by MT »
 

Offline rsjsouza

  • Super Contributor
  • ***
  • Posts: 5974
  • Country: us
  • Eternally curious
    • Vbe - vídeo blog eletrônico
Re: The $1 Linux-capable ARM processor in TQFP...
« Reply #22 on: September 19, 2018, 03:22:55 pm »
Why is this deemed Slashdot- and Hackaday-worthy?

Every site needs some clickbait, it keeps the lights on.

It was worth reading for the reply from Olimex though.
Sad to see they even failed some research on the chip though. A13 is R8, which is used in the $9 CHIP board. My version of A13 base module isn’t too dissimilar from CHIP either (although I do have 512MB DDR3 and 32GB eMMC, a step up from the 4GB/8GB NAND.)
I loved Olimex' response as well, and their only mistake was to say they were part of refurbished tablets. Otherwise, a lot of what was said is quite relevant and can be used as warning to the ill-advised.

If Allwinner is selling these devices at $1 unit price directly or from one of their distributors, then I can say Olimex' post is BS. Otherwise, it is spot on.
Vbe - vídeo blog eletrônico http://videos.vbeletronico.com

Oh, the "whys" of the datasheets... The information is there not to be an axiomatic truth, but instead each speck of data must be slowly inhaled while carefully performing a deep search inside oneself to find the true metaphysical sense...
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf