News HummingBoard, Banana Pi take on the Raspberry Pi

Pin-compatible boards offer better specs.
http://www.bit-tech.net/news/hardware/2014/04/23/hummingboard-banana-pi/1

 

The website's broken - or not finished yet - I'll grant you, but I'm pretty certain it's no hoax: mine's on the way from China as I write this.

Where are you getting that from? The A20 has two CSI channels as well as LVDS display support.

Because it does accept Raspberry Pi add-ons and accessories? The GPIO header is pin-for-pin compatible with the Pi. At least, it's supposed to be - I'll know more when mine arrives.

I'll give you that one - assuming it gets hit by Customs. Will be interesting to see what the price is when it hits the UK proper; I know CPC are investigating stocking the thing.

Don't hold your breath for that.

Interesting, I hadn't seen that - although why are those two entries in the feature list italicised, when nothing else is? Hmm. I've got a mate who's been testing a pair of HummingBoards for the last six months - I'll drop him an email, see if he can confirm or deny their presence. If they've managed to cram both on board, then that's a pretty stunning bit of engineering...

 

Spoken to my mate: the samples he's got don't have mSATA or PCIe anywhere on 'em, broken out or otherwise. Now, his samples are from the very first production run six months ago, so it's possible that things have changed since - but he also says Solid-Run hasn't mentioned such a change any time in the last couple of months. I'm wondering if the italics mean that the SoC supports 'em but they're not broken out - a sort of "would be nice to have if we can engineer it into a future revision" of sorts? Looking at the i.MX 6 block diagram, both SATA and PCIe are supported but only on "selected product lines"...

EDIT: Seriously, how did I miss this whackin' gert explanation at the top of the feature list? "Italic lines below are features not populated in the first batch for developers." Guess that explains it!

 

Should be possible with an interposer chip, although I have no idea if this is what they're doing; the website is absolutely atrocious. I'll know more when (if) my board arrives.

I'm not surprised. As far as I'm aware, the RPi Camera Module only works on the RPi; it's tied heavily into the BCM2835's VideoCore-IV processor. At least, I've not heard of anyone managing to get it running on any other CSI-enabled board.

Trialling 'em for import; he runs a low-power PC retail specialist and is a reseller for Solid-Run stuff among others. He's also said he's going to pop one of his samples in the post for me to have a play with, so yay!

 

Now I'm *really* confused, 'cos the very first post on that page says "CON2 is a DSI display connector."

The joy of Chinese SoCs!

Don't look at me, I've never tried a non-official camera with the Pi, nor the official Pi camera on another board. All's I know is that a journey to the tenth page of a Google search found absolutely no reference to anyone having got it working on anything but a Pi. (Mind you, that could be 'cos nobody's tried, but given I literally wrote the book on the Pi you'd think I'd have heard something to the contrary.)

Incidentally, despite the Pi's CSI connector being allegedly fully MIPI-compliant, no other CSI camera beyond the official Raspberry Pi Camera Board will work. As I said: the Camera Board is heavily tied into the VideoCore-IV processor, which remains locked down and proprietary. To get a camera to work over the CSI port, you need to write drivers that run on the VideoCore-IV; trouble is, Broadcom hasn't released the information required to write said drivers. So, I'm not seeing much difference betwix the Banana and Raspberry here: while both have what they claim is a CSI port, neither will really drive anything except their respective dedicated camera modules.

EDIT: As an aside, I'm actually testing out a Freescale i.MX 6 quad-core board right now: the TBS Matrix. Quick performance comparison to a Pi: compressing a 100MB file of randomness took 1m54.8s on a Raspberry Pi, and only 54.6s on the Matrix. Switching to the multi-threaded pigz boosted that to 14.4s. So, you can expect the HummingBoard to be around twice as fast as the Pi for single-threaded software, and eight times as fast for multi-threaded software. Not too shabby!

 

So, I've got a prototype HummingBoard. It's a six-month old design, back from when it was known as the Carrier-One or C1. Unlike the final design, it has a single-core iMX.6 Solo and just 512MB of RAM - but still manages to trounce the Pi in performance, running through my compression test in around a third the time(!)

​

Sure enough, the P1 header is present and correct. The board size, however, doesn't match the Pi - it's wider. As a result, it doesn't fit in any of the Pi cases I've got lying around. Whether that's something that's been resolved in a subsequent board respin I don't know.

​

(Apologies for the blurry shot - I'm trying to do this in a rush, 'cos I'm supposed to be doing something I'm actually getting paid for!)

The underside shows a bunch of unpopulated surface-mount solder pads. I'm guessing these are where the PCIe and SATA connections will go on the finished board. Also, you can see the micro-SD slot - a better choice, I reckon, than the full-size SD card slot of the Pi.

 

Right, I now have a Banana Pi. So clearly it *does* exist.

​

GPIO is currently shafted in software, with an updated image promised soon. Boots Raspbian fine, but I've got it running Linaro for performance reasons. Speaking of which: yes, it's considerably faster than a Raspberry Pi. I can get 140Mb/s over its Ethernet port without trouble, and 13MB/s file transfers via SSH if I use 128-bit RC4. That compares to the Pi's ~3.5MB/s throughput.

The board is physically larger than a Pi, which is a pain, and the GPIO header has been shifted towards the composite jack to make room for a mounting hole - which means that larger piggyback-board add-ons won't fit properly. Stupid decision, that. There are also considerable changes since the prototype board pictures that are doing the rounds: aside from PCB colour, note the straight rather than right-angle SATA connector and the shifted IrDA receiver.