Libre Computer Board Launches Another Allwinner/Mali ARM SBC

[coder]

Yeah, the XU4 is their most powerful one.

I wouldn't go past the C2. It's a high-volume part, inexpensive, and still a good improvement over the Pi 3.

Though Udoo x86 could also be an option.

That or an up-board.

[LoveRPi]

No it isn't. Raspberry prices are like that (can be bought from local market), but these come with shipping costs and VAT.


Not quite. RPi 3 is 64-bit ARMv8 and really efficient. The Allwinner chips run too hot with factory defaults. Recommend 1,2 GHz max or even less.


The most important thing on these boards is the GPU and desktop Linux support. Surprises me why nobody else but RPi wants to actually support their users.

If you compare apples to apples, the Raspberry Pi 3 is much more expensive after shipping and VAT.

Raspberry Pi 3 BCM2837 uses the Cortex A53 which is fabbed on the 40nm process just like Allwinner's H5. It is no faster per clock but the H5 supports crypto extensions.

Allwinner chips run just fine, it's just Chinese SBC makers never invested in proper software throttling mechanisms since the boards have different voltage regulators.

Le Potato (the last Kickstarter) has working desktop and OpenGL ES 2.0 support and Android 7.1.

Mali Utgard is ARM's fault for their licensing method with the fabless houses.

Raspberry Pi boards have terrible peripheral IO bandwidth. You can't get more than 25MB/s of aggregate IO bandwidth which really limits what you can do. If you stick an SSD via a USB bridge, the most you can get is 20MB/s compare to much more on every other board.

ODROID-C2 is out of stock until January per their last announcement. Don't know what happened.

[chithanh]

The Tritium boards are more expensive and/or less capable than the Orange Pi boards, while using the same Allwinner chips (so same GPL violations and Mali GPU). Why should we buy them again?

[LoveRPi]

The Tritium boards are more expensive and/or less capable than the Orange Pi boards, while using the same Allwinner chips (so same GPL violations and Mali GPU). Why should we buy them again?

Everybody's needs are different. Android 7, form factor, layout, eMMC daughterboard support, 4 independent bandwidth USB port, compatible GPIO layout, are all differentiating features that could make this board more or less interesting.

[chithanh]

Hardware layout is compatible with existing Raspberry Pi cases, that is indeed a plus for the Tritium. eMMC daughterboard is for the people who can't use microSD or USB storage, and don't want to solder eMMC onto the board I guess?

Everybody follows RPi standard for GPIO connectors.

[caligula]

Everybody follows RPi standard for GPIO connectors.

Some Allwinner boards support incompatible CAN bus if I recall correctly. They might also have some other extra features, extra I2S, audio, coax video, USB via the pins. Makes it incompatible.

[SystemCrasher]

Allwinner and Mali... that's like the opposite of open. Why do all these scumbags try to steal money from the people at Raspberry Pi who are doing the real work?

Broadcom is opposite of open too. Just in different ways.
- Proprietary boot loader full of blobs.
- Lack of CAD files for PCBs.
- No real documentations or ability to buy ICs on open markets.
- Just one PCB MFR which exclusively decides on available boards portfolio, timings and so on. So it happens to be vendor lock.

So whatever, Broadcom really deserves to see their margins cut, and why not in favor of Allwinner, if they've got cheap ICs one could at least buy? I could bet there're many ppl who are not okay with Broadcom approach to mentioned issues. Look, I could buy Allwinner ICs really cheap if I want to. Even single IC, should I want it. Without NDAs or weird contract terms. But if one wants to buy something from Broadcom, its worse than hell. Then there're PCBs with CAD files for e.g. Olimex boards, so I could even submit my own batch of PCBs to factory, possibly changing something. This is definitely new degree of freedom, right? Something that utterly unavailable when it comes to Broadcom and Raspberry Pi. Sure, not everyone needs this. But if we push this one step further, most ppl do not know how to write programs either, so next logical step would be to tell opensource is pointless, eh? At which point I would strongly disagree, after all this approach proven to thwart development. Just look how boring Windows is and how it utterly lacks diversity. You can create some desktop PC or mediocre server. Everything beyond that performs really bad with Windows. Bye-bye Windows Phone, I wouldn't miss you.

[SystemCrasher]

Some Allwinner boards support incompatible CAN bus if I recall correctly. They might also have some other extra features, extra I2S, audio, coax video, USB via the pins. Makes it incompatible.

Most SoCs around have thing called "pinmux", "alternate functions" or so. Names could be different, but idea is always the same: you could have different peripherals routed to particular pins/balls of IC. This means pins punctions aren't set in stone, software could eventually map something different on these pins if they want to and HW supports this remapping. Exact remapping abilities vary wildly across SoC, Linux has got at least some attempt to bring some order in this chaos, trying to do it more or less unified way. So when there're some pins on some header one can't really count on these being fixed function. Software could change meaning of most IOs. At this point compatibility happens to be quite vague. But, well, SBCs are usually meant to create more or less custom things. So if one plugs it to their car as some kind of on-board computer, capable of accessing state of internal car systems, they probably want CAN over whatever RasPi had at these pins, since this way it fits better their application. Of course one could just buy completely external CAN adapter but it would likely turn idea into mess of wires and devices, looking lame and incomplete, after all.