Tiny quad-core system-on-module for makers and hardware startups | Hacker News Reader

Tiny quad-core system-on-module for makers and hardware startups

neutis.io

204 pointswilley119 comments

My flabber is gasted at people complaining about paying a meaty $49 for a quad-core, 64bit, cpu with 8gb of local storage. Twenty years ago I was writing docs for the SGI Challenge[1] with its four MIPS R4000s in a tidy cabinet the size of a two-drawer file (disk drives mounted externally) weighing about 250lbs.

This thing far exceeds the capacity of that machine, costs several orders of magnitude less, and FITS IN A FUCKING MATCHBOOK. Jeez. Get some perspective.

[1] https://en.wikipedia.org/wiki/SGI_Challenge

so? $49 is way too expensive for this. i can get a x86 from ebay way more cheaper than this.

who would even bother, the rpi zero is $5.

I've never seen a rpi zero for $5 in my country.

At least $11 and then there's shipping. and limits of one per customer.

(Anecdata) If you happen to be in Cambridge Massachusetts go to MicroCenter and ask: I have 5 that have all been bought at $5 ($10 for the Pi Zero W, which is still incredible to me).

I've bought multiple Pi Zero Ws for $5 at my local microcenter. The first one is $5 but the second one is $15. I just would pick one up on lunch every couple days. I have four in my house. $20

I wish they didn't pull that shit, though. I can't buy two of them because of that. I didn't even know you can buy more than one, every site limits to one per order. I wish they just priced them normally and let me buy them in bulk.

Right now my Pi Zero is collecting dust because I know I can't find a second one easily. Meanwhile, I buy ESP8266s by the tens and use them for every throwaway purpose.

Not an apples-to-apples comparison.

The pi zero has no wireless, a single-core ARM11 CPU, no on-board flash storage, less than half as many GPIO pins, requires a higher supply voltage, etc etc.

The pi 3 is much more comparable, and at a much more comparable price, especially after you add in the price of the SD card, which is necessary. Pi 3 is also much bigger, and, I'm guessing, draws more power.

Pi Zero W includes wlan and bluetooth, and the kit that includes PSU and SD card cost around ~$35.

Stil I agree, its not really comparable. The Pis is mostly learning/teaching tool. I can see using this in a product directly. Something I would grab when I needed something more than an ESP.

The closer comparison is with the Raspberry Pi Compute Modules. Notably the 3 and 3 Lite, with eMMC and without respectively, $30 and $25. No wireless though.

I'll PayPal you $5 if you can show me a receipt for a purchased Pi Zero for $5.

I've ordered multiple rpi zero's / zero w's at cost ($5/$10 respectively) + a few dollars for shipping. Purchased through adafruit, pi hut, and canakit. No need for a receipt, it's in stock for purchase right now [0][1][2]

What you don't get however, is the ability to purchase them in any bulk quantity at that price. However the bulk price is still much less than $49.

[0]: https://www.canakit.com/raspberry-pi-zero.html [1]: https://shop.pimoroni.com/products/raspberry-pi-zero [2]: https://www.adafruit.com/product/3400

https://imgur.com/gallery/myVny my PayPal is masle.com@gmail.com. Thanks!

That's more like ten dollars though.

Nobody claimed it was $5 including shipping. It's still £3.82 ($5.32) without VAT (deducting it since US prices never include sales tax, either) and shipping at Pimoroni UK right now.

It's not $5 plus $5 shipping unless I can buy three for $20. Otherwise it's just $10 including shipping.

Since you can't buy more than one, "shipping" is just a way to get the price up. It's like those eBay sellers who sell a $10 item for $1 and charge $12 for shipping, per item.

Its also a problem for warranty related reasons. Those eBay sellers who sell an item, will only have to reimburse $1 if the item becomes damaged. Same with $5 item + $5 S&H but different proportions of price and therefore severity.

It’s not like these shops only sell Pi Zeros, the shipping costs are the same for other Raspberry Pis and accessories. But you’re right, limiting saws to one per order raises the effective price to $10.

My local Microcenter has Pi zero[0] and Pi zero W[1] in stock for $5 each.

They will only sell you one of each per visit at that price.

Last time I bought a Pi Zero from them, I paid $0.99 for it.[2] Once again, they would only sell one per visit at that price when they were running that sale.

You aren't going to get enough for any kind of production run at that price, but they're not exactly unobtanium.

[0](http://www.microcenter.com/product/486575/Zero_W) [1](http://www.microcenter.com/product/486498/Zero_v13_Developme...) [2](https://i.imgur.com/uLf8bi8.png)

The RPi Zero at $5 is a (hardly verifiable) reality only for a few lucky purchasers, especially writers in some blogs/magazines, but an urban legend for everyone else. 10 people every now and then telling the world they got it at $5 mean nothing when there are thousands who can't get it without price inflating bundles. We already know why it is produced in so scarce quantities, so I won't comment further. Besides that, the RPi Zero has also way inferior capabilities compared to this one..

Regarding the Emlid board, yes, it's way overpriced, but it doesn't seem a product aimed at makers -though it also tries to sell itself as such-. Once you add the board to make it useable the price will grow a lot. To me it has its uses as an industrial module manufacturers can stick into their products. I personally would rather get a NanoPi NEO2 plus its dock at a total cost of $28, or 10 more for 1GB RAM, and the OrangePI also has some boards with similar characteristics, but that's a personal choice: Nano/OrangePI boards are wonderful boards for makers and experimenters but if you're building something critical then you should look elsewhere. If the Emlid board satisfies safety requirements then its price is totally justified (and selling it as a makers board could do bad PR), otherwise they should make a breadboard friendly version at a lower price.

THe RPi Zero is misleading. It has less performance than this.

The direct competitor is the RPi3 and costs $35 which is $15 less and it doesnt require a breakout board to access it's IO. The eMMC is the only interesting aspect about this board.

SD card manufacturers can get away with using the shittiest flash and controllers they can find because SD cards can be replaced. eMMC is not user serviceable and therefore actually needs to be... you know... reliable.

This board is clearly not for hobbyists. It's intended for hardware startups that will design their own custom breakout boards.

For the record, I have a few RPi Zeros but they were all sold as kits costing $20-$30.

Also, PIs aren't so solid. Definitely not good for industrial use.

I'm not sure where you live, but the Pi Zero isn't a myth. Just this week I bought two Pi Zero HW's and a standard Pi Zero. The standard Zero was just over €5 + shipping. I could only buy one zero, but I could have bought 50 or more of the HW's as they don't have the purchase restriction.

The Pi Zero is easily obtainable now Adafruit have it for $5+shipping in the US, Kiwi-electronics have it for €5+shipping in Europe and Pimoroni have it for £5+shipping in the UK. Hardly a myth.

Could you send me a link? I'm in the market for one of these boards and would be interested to see what the alternative secondhand option would be. Better to reuse & recycle than create more landfill.

Depends what you are going to do with it. If you are buying 1 for a hobby, you have a point. If you are buying 10000 for you business, you'll care about the cost down to the cent.

Designing your own board looks easy from the outside, but quickly turns out to take a lot of effort and time as you work through PCB design, first boards, initial bring-up, tweaks, new boards, EOL notice one of your components, slight redesign, and finally you're headed of to (pre-)certifications only to find you need to fix something there.

This assumes that your SoC vendor actually cares about you as a customer and provides you all documentation, erratas, and EOL notices in a timely manner and not considers you a liability that due to small volumes (<1M) will never earn someone in the SoC company any real bonuses.

I'm making a living developing big and small embedded systems for customers as a software developer, so I see this mostly from the outside, and depending on the experience of your team and how good your contacts to your component suppliers are, you may be able to speed up the process.

But for sure it takes a lot of effort to get to the stage where you have what this $49 SOM gives you. $49 times 10k units is quickly eaten into, and then have you have to consider what value you put on time to market.

It never ceases to amaze me how cheap our industry is, even about buying our own products.

It really depends on what you want to do with this product. What you're getting for $49 is hopefully a stable, tested, and well-supported platform which can really help when all you want to do (or can afford to do) is concentrate on the other aspects of your product. Built-in OTA upgrades and pre-built manufacturing tools are very helpful as well. The $49 price point however, assuming a retail product, probably puts your final retail price north of $100 by the time you add in all the other components you're going to need to make this work plus profit. If your product fits into that space then you could save a lot of time going with a solution like this (but I do see a cost-reduction round in your near future).

If you're trying to stay below the magic $99 retail price point though, you may be in for a challenge. Do you need a battery? Is that chip antenna good enough or do you need to have a better one made? Plastics? Power supply? Those DF40-80 connectors alone are going to account for north of $1 in manufacturing quantities. And after all that, you still have to pay for whatever it is your product actually does.

When I'm designing products, I sweat over every dollar, and big ticket items like this really need to be justified. If I don't need 8GB of flash, I don't buy it. Likewise quad core - how big a battery am I going to need to drive that thing for whatever my requierment is? Could I save $5 on the battery by dropping to single core (and save another $1 there too)?

As with anything, I think the use case for this amounts to "It Depends". I think the price point for this puts it out of range of the IOT space, and it is over-spec'd for a basic RF bridge type application (plus all the issues that go with trying to control noise when you don't own the chip layout). Maybe as the driver for a custom touchpad or something like that. Suffice it to say that I think it has a market, but that's probably not competing in the IOT space, and likely not the prototype space either.

$.02

> When I'm designing products, I sweat over every dollar, and big ticket items like this really need to be justified.

If you can spend (just) a month or two designing, debugging, and procuring a cost-optimized variant of this that costs $25 instead, your time spent on that is equivalent to just buying a thousand of these. One of the big things in the "it depends" question is how many of these things you plan to sell. If it's less than thousands, buy modules.

The problem is that there are a dozen Chinese companies that have already designed, debugged, and produced a cost-optimized variant of this for less than half the price.

The Orange Pi Zero Plus 2 for example has the exact same SoC, same 8GB MMC, Wifi/BT, and is $20. Not saying that this board necessarily isn't worth it but clearly there is stiff competition in the SBC space and a lot of the boards are extremely cheap. The size is an advantage. It has mezzanine connectors instead of rows of headers. They claim they'll actually keep the kernel up to date which is a major problem with many of these SBCs, though I am rather skeptical given that they're using Allwinner chips. They cannot magically produce a new kernel without sources from Allwinner.

It's a fair point. I deal in consumer electronics, and my company needs to make money, so I tend to think in terms of hundreds of thousands of units. I'm not sure how a company can even be profitable selling thousands of units unless they each cost thousands of dollars. When you get up into those prices, though, it feels (at least to me) like you're beyond embedded-class SOM devices like this. I guess if I could make a $100 device and sell it for $1000, I would. Maybe Apple would hire me... [ Disclaimer: I own an iPhone ]

> If you're trying to stay below the magic $99 retail price point though

What are you imagining using this for? That seems like a really low price.

Very sad to see so many negative comments. It seems a good product. As the owner of an emlid navio2 rasp pi hut I have to say that I am impressed with their work. Top hardware quality and great documentation. In the first hour of use, I used their C libraries and Python examples to manage to access some of the module's sensors via go. I hadn't even ever do C bindings with go before that.

This board seems to target people who want to create products with it. I wouldn't place a rasp pi in an end product and certainly wouldn't come near an orange pi even for a prototype.

No idea why anyone would want to use this.

There's dozens of SBCs with 2x better specs than a Pi, but the Pi is arguably the most dominant board out there. For makers (for anyone really), documentation is king. Working code, lots of examples, and good forums make the SBC, not the hardware.

As for the other group, "hardware startups": Why on earth would you lock yourself into another startup for the brains of your new project? It makes no sense. Remember what happened to Intel's SBC? If a company that big can't provide decent EOL support, what do you think will happen if Neutis folds? Pick a ARM something core with a certified BT/WiFi module, and choose from several manufacturers with real EOL schedules and a history of following them

Don't get me wrong, the 10 year guarantee is nice, but I don't think anyone can afford to believe it.

This isn't competing with the pi at all.

It's competing with the pi compute module, and it's significantly nicer (onboard eMMC, WiFi).

Also, you're really, REALLY underestimating the additional effort involved in a complex BGA layout for any modern device. For low-volume, specialist stuff, it's generally FAR more economical to buy-in a compute module then design in house. If you expect to sell a few thousand expensive specialty $INSTRUMENT_THINGs, the hundreds of hours of engineer time needed to bring up linux on a new board layout can buy a lot of compute modules.

After all, if at a later point you do want to move to in-house, you can always just design a footprint-compatible module at a later date.

Shit, I'd personally have been really excited about this if it were a few years ago. I was doing board layout for a project with a compute-module. The selection now is so much better then it was in 2012.

It's a niche, but I can see real value in this for small runs (<10,000 units) of space-constrained designs. Using a SoM is a huge time-saver for development, but most available are either expensive and have painful and partially closed BSPs (most SoMs on the market) or are bulky and weaker (Raspberry Pi Compute Module, C.H.I.P. Pro).

In the event Neutis is at risk of folding or you want to go to larger scale production, you can put the time and effort into dropping the Allwinner SoC and SoM peripherals straight onto your design.

I have a couple of projects in mind that could use this, where I was otherwise considering the C.H.I.P. Pro or going through the effort of integrating a SoC.

My thoughts exactly. I think many of the posts here are from people who are not designing for manufacture and are just looking at the specs like it's a small standalone computer like the rpi.

It looks pretty good to me. Havent had to use those high density connectors before but assume its straight forward enough.

Also, i can have a Pi from any number of stores.

Most of the other boards are either direct from manufacturer, or via some kind of specialized store that can set any markup they like.

That said, i would love to see some kind of mobile device built around this.

Well, Intel Edison was a different story. The elephant in the room is chip availability for the Allwinner CPU. Is that guaranteed to be in stock 5 years down the line?

It claims to run on mainline Linux but mainline Linux does not yet support CPUFreq, hardware crypto, HDMI, nor the GPU for the Allwinner H5.

http://linux-sunxi.org/Linux_mainlining_effort

It says "based on mainline linux", not "running on mainline linux". They probably have non-upstreamed patches.

But by that definition, what linux isn't 'based on mainline linux'? Unless they're offering some guarantee that they'll keep their patches up-to-date...

There some boards out there still running on 2.6.x series, and many running 3.0 - those would not qualify as “mainline linux״ at this time.

Some boards can boot a kernel.org kernel without patches; those run “mainline Linux”, not something “based” on it.

Do you know of any board that can run mainline Linux with support for all components?

The closest you're likely gonna get is an i.MX6 based board.

Everything is based on mainline Linux so that's a rather empty statement. Question is how far away is it, and what is the track record for getting changes upstream.

I could see this costing $30, but $49 is quite a stretch. What are your main advantages over Orange Pi Zero [1] which costs $8.49? I can only see the DDR3 rated to -25 and the secure enclave.

1. http://www.orangepi.org/orangepizero/

Apart from temperature range and secure chip which you already mentioned there is also a 8GB eMMC, different processor (H2->H5), more RAM, more peripherals available. Neutis is not competing with Orange Pi as the applications are different, it is meant to be a part of a product, so it has all peripherals available on board-to-board connectors.

But you are marketing this to "Makers". The Orange Pi, Raspberry Pi, and other single board computers, are maker-friendly, because of their adherence to the standard of 0.1" headers, USB or wall wart power, and a large community of users to drive support and development.

Makers largely don't care about import duties to China, or FCC certifications, or manufacturing tools, or barcodes, or Yocto, or secure enclave chips. Just give us the hardware in an easy to use, all in one form factor, with good drivers and support.

> But you are marketing this to "Makers".

I don't get that from their page at all. At minimum, it seems like they're marketing it to "Makers" who are trying to make an actual product, at which point they do have to care.

Orange Pi is also one and a half times larger

There is also the Nano Pi Neo 2 [1]. $18 for Allwinner H5, 512MB RAM, Gigabit Ethernet.

Haven't used the Neo2 yet. But its predecessor, Nano Pi Neo is $10, has excellent Armbian support, and is on par with a RPi3.

[1] http://www.friendlyarm.com/index.php?route=product/product&p...

I have a Neo2 seedbox. It runs great. Rock solid. Pulls 900mb/second over gbe. Barely hits 5% cpu running dozens of torrents, over a VPN it handles, saving everything over NFS. Size of a damn oreo. The OLED kit is great. Easy library and you can write text or 1bpp images to it. Refresh rate 0.3Hz is a bit annoying but it looks cool as hell. Nice undocumented feature: the microusb and full-size usb ports have the +5v and ground wires bridged, so you can power it from an A-A cable. I like this because the A cable is the same side as the ethernet and looks cleaner. [1]

1: http://www.friendlyarm.com/index.php?route=product/product&p...

The Orange and Nano Pis are not FCC-certified, so they're not likely to be integrated into any product sold in the United States, at least legally-speaking, assuming the manufacturer doesn't want to be fined. Orange pretends to be certified[1] (the PDF is named FCCCERTIFICATION), but it's mere verification (now called Supplier's Declaration of Conformity, or SDoC), and SDoC is not the same as certification[2].

If a manufacturer were to use the on-board wireless they could be heavily fined. Certification for these boards would cost around $15k, significantly increasing the price of an integrated product.

At least FriendlyElec doesn't pretend to have gotten their boards certified.

[1] http://www.orangepi.org/download/FCCCERTIFICATION.pdf

[2] https://www.fcc.gov/general/equipment-authorization-procedur...

I can't speak to pretending or not, but at the bottom of this NanoPi/FriendlyElec page, there are very similar looking pdfs for fcc, rohs, and ce.

http://www.friendlyarm.com/index.php?route=product/product&p...

Thank you. I couldn't find that with a Google site search[1]. But yes, that's not full certification.

[1] https://www.google.com/search?q=fcc+site%3Afriendlyarm.com&r...

I've always wondered where the line is on a product's FCC certification. Obviously US retailers such as Adafruit and Sparkfun have sold fully assembled unlicensed radio equipment for over a decade. What subtle changes could these companies make with their products that would suddenly gain the interest of the FCC?

Do you have an example? SparkFun even has a blog entry about certification that's a useful read, although not entirely accurate. https://www.sparkfun.com/tutorials/398

thanks, thats an interesting link.

That said, a manufacturer could still use an Orange or Nano Pi to integrate. Simply disable the onboard Wi-Fi and include a certified USB Wi-Fi adapter.

A new product like this seems to come out every month! Lots of fun for embedded folks. This is a higher-end one, and the power-per-price keeps going down. You can get small dual-thread processors for a few bucks (only 1Mb RAM etc) on up to things like this for $49. Seems like the ecosystem is filling out nicely.

Some background--emlid, the company that is producing this module, relied on Intel Edison modules for the compute power behind their RTK GPS products. The Edison modules were recently EOL'ed.

I would guess that this module is primarily being developed as a replacement for the Edison emlid's RTK products, and they are pitching it as a separate product as well to recoup some of the development costs.

Eh, I'll stick with the PocketBeagle board. It is half the price, very easy to source, doesn't require any special connectors and has those extremely useful PRUs.

Slightly larger, slower and no wireless connectivity but that is a worthwhile tradeoff.

Are you actually using the PRUs? They sound fantastic on paper, but I can't shake the feeling they'd be extremely difficult to actually use.

In my experience the PRU code is fairly straightforward and fun to write, but wrangling Linux drivers/remoteproc/TI SDKs/devicetree etc is a bit of a hassle.

https://github.com/mkj/pru-mic/blob/master/pru/asmpr0.asm is the guts of some code to read from MEMS microphones, for example.

PocketBeagle unfortunately isn't sufficient for integration into a commercial product. It doesn't breakout most of the AM3357's pins. If all you need is USB/I2C/SPI/GPIOs you're OK but PocketBeagle omits things like ethernet, LCD, CAN. If I'm not mistaken PB also has no onboard wireless.

I'd imagine it'd be competing directly with the likes of Raspberry Pi 3 Compute Modules. And the RPi3 CMs don't seem to be as popular as the RPi 0 so far. And that's given the RPi was super popular, and was known for their amazing community and commitment to connection standard, support and production.

This is even more hardcore than the RPi 3 CMs, and more expensive. Seems like you have a brave team to tackle on this pie.

True! I think that the reason why we do not perceive CM3 as popular is that it is not something that you can use right away. Unlike Pi Zero it requires a custom carrier board, which narrows customers to those who are trying to build a product with it. They normally do not share much online :)

Neutis has advantages over CM3, it is much smaller, comes with Wi-Fi and BT and the connectors take much less space on the PCB!

Furthermore, the Pi Zero is single core and very hard to source in bulk for product integration.

What secure element chipset are you using? This could be very interesting for a project I am working on. Does it run JavaCard / GlobalPlatform and are the access keys available?

Not sure why he didn't reply to your question as well, but it's a https://www.microchip.com/wwwproducts/en/ATECC508A

I think this looks good! I also think the price is reasonable, and the linked page also mentions volume discounts.

Extended temp range, long-term supply, CE & FCC certification, BSP with OTA upgrade support, manufacturing tool, and best of all: "we rely on Neutis ourselves in our line of Reach M+ and Reach RS+ RTK GNSS receivers".

What WiFi+BT chip is used?

Webpage mentions Debian briefly. How good is that support today?

Thanks! Wireless module is AP6212A, which is based on BCM43430. It has very mature software support.

I will ask team about Debian tomorrow, can you get in touch with us directly so that we can keep this conversation going? You can find the contact email on Neutis website.

Crazy question, does monitor mode work on the wifi? That chipset reminds me of my HTC Wildfire S - aka Mr Robot Phone [1]

1: https://bcmon.blogspot.com/?m=1

I don't know firsthand, but "Debian" is primarily responsible for userspace binaries, which just need to be built for aarch64. Just about everything specific to the chip & peripherals is handled in the kernel. (Yes Debian has a linux-image deb but I don't think anyone uses that on ARM. You build your own kernel.)

I am oversimplifying a bit, but you should be able to take these instructions [1], combine with a kernel and bootloader built for this board, and write the resulting image to eMMC/flash. You need to configure a few things correctly, boot cmdline & bootloader options but that's it.

Source: have used debootstrap to build custom distros for a couple ARM modules as well as x86.

[1] https://www.olimex.com/wiki/AM335x#File_system

I am new to hardware things and just wanted to check. Is Neutis N5 suitable for making something like PocketCHIP but in more compact form (smaller device)? If so I would definitely prefer 49$ over 5$ rpi0, just because of better performance and much smaller form factor.

I like how they threw “makers” (or “pro makers”, whatever that means, on the actual page) into the title. Uhhh... Yeah I’m going to spend $50 and then probably at least another $50-$100 for the breakout board, which I’ll need as a (pro|) maker because they use high-density FMC connectors on the actual module itself... At this price point and level of integration difficulty why not just use a (much more standard and better-documented) Pi Compute Module?

$49 seems very expensive for something like this. There are similar packages available in the $5-30. What makes this special?

Marketing & a very pretty website</cynical>

It's a 64-bit 4-core ARM. So I expect it to cost a little more than a Raspberry Pi.

This is the identical CPU of a Raspberry Pi 3. 512 MB of RAM with 64 bit is a pain, and that Mali GPU is a proprietary piece of shit. What's the market for a SoM for "makers and hardware startups" that has negligible upstream support.

That's before you consider that they chose a SoC with considerable GPU/VPU capabilities that can only do HDMI or composite, the first you would never use in an embedded application where you have a fixed display, the latter you would never use, ever.

It's not identical to the RPi 3. Raspberry Pi is using a Broadcom chip, this one is an Allwinner H5. They are similar, but at least the H5 has far better support in mainline Linux.

Still Mali, though. Which I totally agree is crap. And no drivers for other display types.

Mali 450 just won't die. Atleast it is not the Mali 400.

We were talking about the CPUs, and these are pretty identical. Both use quad ARM Cortex A53, the H5 has 100 MHz more top clock rate and is probably manufactured in a more modern process, but real-world performance isn't going to be much different.

Performance wise they are very similar, but you mentioned upstream support. That is actually a major point of differentiation between the two.

Well, yes, the Raspberry Pi has much better upstream support. The biggest obvious issue being the Mali.

I mean upstream support in the Linux kernel. The Allwinner chip is supported by the mainline kernel, the one in the Raspberry Pi is not. To me that's a big plus in its favor.

Also in terms of support by the creators, if they're serious about the 10 year commitment then I'd say actually that is better than Raspberry Pi. Mindshare in the maker community isn't that important to someone doing serious embedded hardware development for production as long as the company behind it is solid. Which kinda sucks, I wish there was more incentive for chipmakers like Broadcom to be more open but...

No offence, but I think you haven't been keeping up-to-date. The RPi is fully supported by the mainline kernel, most significantly, there is a full atomic DRM kernel driver for the VC4 GPU and corresponding Mesa userland, maintained by someone with transistor-level access to hardware documentation.

I have great appreciation for the work Maxime Ripard et al. have been doing on mainline support for the various Allwinner chips, hence why I kicked in a few bucks for their kickstarter on reverse-engineering the Allwinner VPU:

https://www.kickstarter.com/projects/bootlin/allwinner-vpu-s...

But lets not kid ourselves, none of this was Allwinners doing and they could not give a fuck about mainline Linux.

Wow you're right, not sure how I missed that. Guess I just got too used to all those years of Broadcom chips being giant clusterfucks of patches and old kernels.

> but at least the H5 has far better support in mainline Linux

Could you elaborate on this?

My experience has been that the BCM2710 (and the RPi 3 board, in general) has had better (and earlier) support in mainline than the Allwinner H5. A client of mine is shipping a product based on the RPi3 after evaluating both (including building their own H5 based board), because support for H5 on mainline was poor and lagging compared to the RPi3.

> 512 MB of RAM with 64 bit is a pain

Hence why all extant Raspberry Pi 3s have a full GiB of RAM.

> 512 MB of RAM with 64 bit is a pain

That's an odd statement. How are those two things related?

You can still run 32-bit binaries (including thumb-2) under 64-bit OS in 32-bit execution state -- as long as kernel is configured to use 4kB pages. So your usermode binaries don't need to be bigger.

Or you can take advantage of higher performing 64-bit mode.

(Unrelated, but my pet peeve, one thing a lot of people get wrong: 32-bit ARM (Aarch32 and most ARMv7) supports 1 TB of physical RAM and 64-bit IIRC 16 TB. Not sure where the idea 32-bit ARM could only support 4 GB comes from.)

It is all about the details. Industrial DDR3 with -25C temperature is quite expensive, but it is a requirement for any company building a product that could be used outdoors. Quad-core GHz A53 in such tiny footprint, but really with everything needed onboard. Very motherboard and embedding friendly and with proper software support.

I think I will give some backstory which might be relevant. This product was born out of a need to migrate from Intel Edison in our high precision GPS equipment. We looked long and hard for Edison alternatives but there were none. They were either too big, or too expensive or did not have enough computing power. So we decided to build our own.

Disclosure: I am co-founder of Emlid, the company behind Neutis. Happy to answer any questions.

"Neutis runs modern Linux kernel based on the mainline version. Unlike other modules that are stuck with old legacy kernels without updates, Neutis will get security fixes and new features on time."

You must clarify that. You are on a technical forum and that simply is not acceptable. "Mainline" is 4.16-rc2, there is only one mainline, so you are saying it's shipping with an rc quality kernel that is still in active development?

Or do you mean it comes with an LTS? "Yocto and Debian". Debian does not use mainline kernel, in fact, no linux distro does, ever. What version of debian?

In the context of ARM devices its understood that it isn't based off a custom fork maintained only by the SOC manufacturer.

Currently our main focus is on Yocto, since this is what most people would use in production. Right now BSP runs on 4.15 and the processor is well supported in mainline. Since our own product is based on Neutis we put a lot of effort in testing of stability of the system and peripherals.

There is so much work done by the great linux-sunxi community, you can see the mainlining progress for H5 processor here: http://linux-sunxi.org/Linux_mainlining_effort

Sorry, it is the middle of the night here and I am not ready to confirm on the exact Debian version, but I believe that it has not been decided yet. You can contact us directly and we will get back to you with more info on this.

true but its smaller than say a pi3 and has extended temp ranges

Would you kindly shed more light on what kind of Crypto chip are you using? What are its capabilities?

Sure! It is ATECC508A from Microchip and it does a whole lot of things. Please check out the chip description from Microchip for full list of capabilities.

That's good enough for U2F (the same chip is used in u2fzero). But not good enough for RSA.

I wonder if it can be made to work for attestation like TPM

I just want a cheap one that has bluetooth LE so it can be controlled from a phone app. Does such a thing exist?

ESP32 sounds like it meets your needs, but it won't run Linux. It will run MicroPython though.

I will have a look, thanks!

I will second the ESP32. It's sub $4 in most places, has a dual core processor, BLE, WiFi, hardware PWM and I2C, and much more.

Also their C SDK environment is quite pleasant to work in.

Any linux board that has a USB A port, runs the bluez stack and node js, can do this with a $5 bluetooth dongle.

There's a really nice set of node.js libraries, noble and bleno, which allow you to communicate in both directions. Noble allows your linux box to be a central device, and bleno allows your linux box to be a peripheral.

Yup, that is too big. I want it to create wearable tech, or other small devices.

Except when you find you bought a fake CSR Bluetooth dongle that doesn't actually support BTLE.

what would the current draw on this be for average use ?. Would it be suitable for a handheld mobile device ?

Sure, we use it in a battery powered device ourselves. Consumption really depends on how you use it, but roughly it will be around 1W for average use. You can limit frequency if you need to.

For mobile devices the standby current is very important. Is any kind of sleep mode implemented? The low cost SoCs always seem to omit this.

With a secure element this has a huge potential, though obviously it depends on the details.

Unless it gets the ecosystem the Pi and the Beagle have, this gets old rather quickly.

I think "ecosystem" doesn't actually matter much (if "ecosystem" means maker/ tinkerer mindshare) for commercial applications. All you need is (1) kernel support for the chip including drivers e.g. for video/ peripherals, and (2) good userland support for aarch64. You can debootstrap a Debian rootfs that will work on any aarch64 chip. As mentioned elsewhere, it sounds like the H5 chip has good kernel support which covers (1.) Actually there's a (3), which is bootloader support. I don't know if the H3 typically uses u-boot or some other bootloader.

What differs is what each chip actually supports. E.g the Pi/BCM2835 doesn't have onboard ethernet MAC/Phy. Neither Pi nor H3 have onboard CAN but maybe neither of those things matter. Point is these days most of the details of the chip are abstracted away; if you have good kernel support there's not a lot to worry about.

There is no real demand for alternatives to the RPi.

People who buy RPis don't care about specs and price. They usually buy only one for $35 and use it to get linux experience or they already have a specific usecase in mind like retropi, pihole, etc.

I once wanted to work on the mali open source driver but now I'm thinking: What's the point? Nobody (including me) gives a shit about it anyway. Just spend $15 more on that RPi 3.

> There is no real demand for alternatives to the RPi.

Only if you quality that with "among consumers." There is tons of demand for ARM-based modules that are designed to be embedded in a commercial product. Comparing this product to any SBC such as RPi3 or Orange Pi completely misses the point. Just look at the headers on pretty much every SBC. If you see 0.1" pitch headers, it's meant for hobbyists, not commercial application.

As mentioned elsewhere, you can compare this to CM3, but there are tons of other players in this space (and there have been for longer than the CM has been around.) To name a few: Anything from Variscite, Myir, Toradex: note many of these cost well north of $50 because they're meant for commercial/ industrial. Samsung Artik 053, Chip Pro, Olimex SOMs and SOPINE are some newer players that have been pushing the price down a bit in this space. So this product is price competitive, esp since there aren't many aarch64 modules out there.

Specifically: if you were designing a product that needs Ethernet (esp GbE) you almost certainly would not choose the RPi CM because it's not on the BCM chip. i.MX and AM335x support 10/100 MAC/Phy, H5 supports GbE. If you were doing an automotive product, RPi and H5 do not have CAN, so you might choose an AM335x SOM. There is definitely a real demand for alternatives to the RPi in the space that this product is targeting.

Ecosystem does not mean maker/thinkerer mindshare in this case. It means hardware, I/O, software, firmware and available knowledge from the board perspective, and active distro integration from a software perspective. There are enough people and there is enough momentum to build and maintain multiple software distributions for the Pi (and it's BCM SoC), as well as upstream patches and maintain those. Even in the commercial space you will not find such well supported boards.

This does not mean that other boards won't work or won't have a place, but are highly unlikely to cover the same global space the Pi does.

Does this require any of the binary blobs that are in the mainline Linux kernel?

> the binary blobs that are in the mainline Linux kernel?

There are binary blobs in mainline?! Or do you mean the linux-firmware project?

This looks like a lot of fun.

Would have been perfect for an automotive project years ago.

incredibly boring piece of kit. years behind...no thx.

Reply on news.ycombinator.com