32- and 64-bit ARM Open Hardware Boards | Hacker News Reader

32- and 64-bit ARM Open Hardware Boards

96boards.org

91 pointstimthorn39 comments

This wasn't completely clear to me from the get-go, but this organization and it's boards are an effort put forward by the Linaro organization. That, to me, makes the product much more relevant and reliable. I wouldn't buy a board from a random startup with no track records of providing support and updates for their devices.

Linaro is an engineering non-profit, support by many of the major players, to improve support for the ARM ecosystem. This includes improvements to GCC, LLVM, the Linux kernel, various libc's, AOSP, ... They also provide Ubuntu, Debian and Android RFS images for supported reference boards.

I understand that they went with a Mali GPU, since it is the reference GPU design from ARM. However, it is unfortunate that there is no open source driver for this platform. Maybe this will change in the future.

> However, it is unfortunate that there is no open source driver for this platform.

What about this one:

http://malideveloper.arm.com/develop-for-mali/drivers/open-s... ?

The closest to Open Source driver is lima, http://limadriver.org/ but libv has stopped working on it for now. see http://libv.livejournal.com/tag/lima

Sadly, the kernel driver is of limited use since the user-space driver is what does everything.

Some more info: http://www.phoronix.com/scan.php?page=news_item&px=MTY3OTM

Exactly - the open source kernel driver is just a gateway for the user space driver to access the hardware. The real magic is in the closed source, user space driver.

I believe Linus has stopped accepting these kinds of Trojan horse drivers into the mainline kernel, unless an open source user space component that does at least handle 2D graphics through it is also available.

They use a HiSilicon SoC. Does HiSilicon publish docs?

They're also near impossible to buy outside of China or NDAs. So it may be open, but I couldn't build one.

No, it doesn't appear that they do publicly.

George Grey and Bob Booth ride again.

(I worked for gcg @ Tadpole in the early-mid 90s.)

I'm not convinced. Despite the cpu the specs looks like some 2011ish TI PandaBoard. Slow usb2 i/o and no SATA ports won't be able to compete against low-end development boards (SolidRun HummingBoard or Radxa Rock Pro). While 64bit 8core ARM is something interesting, the low i/o speeds and ports rather limit any "computational" usage.

Finally, the main reason not to build products out of the variety of ARM development boards is IMHO the missing CE/FCC certification for retail products. Afaik every board including the Raspberry PIs are only for demo/lab purposes. So if you build some embedded server product you'll do have the bureaucracy and costs. At least they should offer 2-3 FCC/CE ready combinations of boards+cases ready to be deployed in the wild.

In my eyes those boards are just an replacement for the Texas Instruments PandaBoards which are discontinued because TI gave up their OMAP business a while ago and the typical low-cost board makers are using cheap dual/quad-core 32bit A8 and A9 IP.

A lot of companies, like where I work, purchase boards like this as a reference implementation. Ultimately at least for us, FCC and CE become conformance studies against our own product in this case.

But I could see this being a problem for companies reselling these boards inside some product they put out.

The Rpi and beaglebone and the like are most definitely CE/FCC certified.

It's probably hard to get a CE mark or FCC certification for a PCB-style product, right? Those organizations probably want to know about emissions but it doesn't make sense to measure a PCB for radiated emissions since it has so few protections against it.

Arguably, that's your job -- when you buy the PCB you are planning to integrate it with an enclosure that will solve those problems and now that you've designed this new product you should get the certifications.

You're basically right with the second paragraph; a Pi in a box is a different "product" from a Pi under CE rules, and the person putting the Pi in the box and selling it is a "manufacturer" with all the CE and WEEE liabilities thereof.

The Pi is certified for radiated emissions - as "EN 55022 Class A product". http://docs-europe.electrocomponents.com/webdocs/109e/090076... See discussion of class A vs class B: http://www.ecma-international.org/publications/files/ECMA-TR... Effectively, this puts some of the interference mitigation responsibility onto the customer.

It's entirely missing the ESD certifications which might normally be applied to cased products. If it had wifi that would trigger a whole other class of certification requirements.

The whole thing is tedious and hateful. I wish there was some way to get rid of it without people immediately flooding the market with cheap high emission SMPS.

The thing is, people are flooding the market with cheap noisy crap anyway, distributed as any of hundreds of non-brands and shipped directly to the customer from outside the regulatory area. Technically it falls upon customs to stop these from entering, but customs is not qualified to do compatibility testing. Personally I think the correct response would be a customer protection law that requires that manufacturers and vendors take back/replace problematic products, rather than forcing pre-certification and membership in predatory organizations on everyone. The current regulatory environment (particularly WEEE) is toxic to any kind of small-scale manufacture, not because of the specific requirements imposed, but because of the gigantic costs of complying with said requirements in a regulation-approved way.

How about just changing it to be more sensible, as an interim solution?

AFAIK the big problem without a case is Wifi/BT (and maybe an RTC) - that said if the vendors would at least get an class B certification for one pcb+case combination they could probably improve OEM shipments.

Some commenters say the Raspberry PI got a Class B certification without having a case - but it also has no Wifi/BT/RTC - so this could be possible.

nope

"This evaluation board/kit is intended for use for ENGINEERING DEVELOPMENT, DEMONSTRATION, OR EVALUATION PURPOSES ONLY"

Seems that the RasPi is: http://www.raspberrypi.org/tag/ce-compliance/ . Apparently they were forced into it by their distributors.

Interesting, the pictures show it being tested without a case. At the frequencies it uses, I would expect it not to pass without some kind of shielding.

Do you know if the Raspberry Pi 2 has been tested yet?

I don't know: I tried to search for an answer but found nothing. I assume it has, on the basis that the distributors would probably have the same objections to selling it uncertified.

How open is this hardware? Are there open specs for Mali? Are there open specs for CSI and DSI? All this is on the board. Available specs are needed to write open source software or at least make it much simpler. I have difficulties to consider a board with missing specification for important parts or connectors to be "open".

Well, then you must have difficulties considering almost any embedded board, because they all have a combination of chips with closed specs, or chips that require blobs to do important things, e.g. boot.

Is there a truly open board except the Lemote stuff?

The Beaglebone Black is the most open board, so far as I can tell.

Here are the PCB sources for the latest revision: http://elinux.org/Beagleboard:BeagleBoneBlack#Hardware_Files

The datasheet for the AM3558 processor is here, no NDA required: http://www.ti.com/product/AM3358/technicaldocuments

The bootloader is U-boot, which is open source, based here: http://www.denx.de/wiki/U-Boot

The GPU in the AM3558 is a PowerVR SGX530, which is not open source in any sense-- I believe that the Linux driver uses a binary blob (but I'm not 100% sure of that).

For comparison, the Raspberry Pi has open schematics (http://www.raspberrypi.org/wp-content/uploads/2012/04/Raspbe...) but not open PCB layout source files. There's a partial datasheet for the processor here: http://www.raspberrypi.org/wp-content/uploads/2012/02/BCM283..., but you can't buy the processor from Broadcom unless you're willing to buy 100,000 or so.

> For comparison, the Raspberry Pi has open schematics...

For the Model B rev 1.0 only. The B+ only has very partial schematics available (link below), and there's nothing at all out there for the 2.0 yet.

http://www.raspberrypi.org/documentation/hardware/raspberryp...

Last time I looked, the proprietary bootloader for the Pi also had a license forbidding you from running it on boards not supplied by the Raspberry Pi Foundation.

That is true, for Broadcom's proprietary BOOTCODE.BIN, the (ThreadX-based) START.ELF VPU-side microkernel and their user-space graphics driver components which talk to it via the Linux kernel shims. (Although, note: you won't get a BCM2836 anywhere else anyway.) Broadcom do, however, have someone writing an open firmware I understand - definitely an unfamiliar world for them, so let's give them time and see what happens.

I am also aware of a complete reverse-engineering effort (for the purposes of interoperability), and that is all I'm going to say about that right now. :)

Olimex has products that stack up similarly.

Minnowboard Max is also open.

I do not have difficulties considering the existing boards. I would just not call them "open", because specs are missing, blobs for some functions, etc. This board is advertised as open and my critique pointed towards that. Advertising non-open things as open does in my opinion hinder the progress to more open hardware. That's a bad thing.

I think the OLinuXino series is the closest to fully-open ARM boards: https://www.olimex.com/Products/OLinuXino/A20/A20-OLinuXino-...

Unfortunately the GPU is a Mali, so software-wise it's not completely open, but the hardware is fully open-source and the SoC datasheets are available.

Two questions:

- ARM licenses its designs so that would make their SoC design not open-source (both CPU and GPU)? (a sub question: they didn't specify in the main page who made the Mali. ARM doesn't make chips itself).

- By "software-wise" you mean the GPU driver? the way nVidia and AMD drivers are not open-source. But since people are trying to reverse-engineer those GPUs to make open-source drivers (nouveau and xf86-video-ati respectively), do you think this could be done for Mali too?

Complete noob (to hardware/ARM stuff) question: Why would I buy this over the much cheaper Raspberry Pi 2? Does it just have a (much?) faster processor?

It's a good question. It has more CPU power, and likely more GPU power as well, but not an order of magnitude more. They main selling point (from my point of view) is for people who want to get their hands on a dev board that has 64-bit ARM.

This board also has software support from Linaro, which is very good, but the Raspberry Pi community should offset this advantage.

In addition to sspiff's good info, take a look at the I/O options:

https://www.96boards.org/wp-content/uploads/2015/02/HiKey_Us...

This will likely limit how useful the board can be, because the I/O choices will prevent a good ecosystem from growing up around it.

Correct me if I'm wrong, but isn't this true for all popular ARM boards today? Decent CPU, low power consumption, and terrible I/O performance?

I remember the PandaBoard as exceptionally bad, most are better but still pretty slow.

Take a look at the BeagleBone Black. It has pretty decent I/O. The big gaping missing piece, of course, is better video output, but that's sort of in the same bucket as open source GPUs: missing in action.

I actually have one on my desk, waiting to be used. Hoping to play around with OpenBSD on it.

Still they (Linaro) are calling it "Consumer Edition" - so IMHO no market fit.

Reply on news.ycombinator.com