Another nice looking ARM board

Possible candidate for porting? I wonder how good the documentation for the SoC?

2Gb RAM, SATA, Gigabit, quad A9s at 1GHz, and under 100 quid.

http://www.buydvb.net/mini-pc-c-26/matrix-arm-mini-pc-p-97.html

2Gb RAM, SATA, Gigabit, quad A9s at 1GHz, and under 100 quid.

UK pricing: (£129)
http://www.redappledigital.co.uk/category-s/1513.htm

So over 100

Technical:
http://www.tbsdtv.com/launch/tbs-2910-matrix-arm-mini-pc-download-and-support.html

I presume you found it through one of the DVB related articles as it seems to be intended for use with USB connected TV tuners.

http://www.tbsdtv.com/download/document/tbs2910/TBS2910-Matrix-ARM-mini-PC-SCH_rev1.0.pdf
NB. Although the press release info suggests it needs the OS burning to eMMC the schematic seems to contradict this with a pair of tables for switch settings to boot from eMMC (onboard) SD and SATA.

What were you thinking? Chip in and buy one for Jeffrey to distract him from other work? Or having a go at a HAL yourself?

Just looked at the schematic info for the DDR3 memory – 2GB

So over 100

Possibly the site I linked to (£94.54) didn’t include VAT?

seems to contradict this with a pair of tables for switch settings to boot from eMMC (onboard) SD and SATA.

That’s good. So upgrading the firmware could be as easy as Pi.

What were you thinking?

I wasn’t thinking of anything specific. I recall our discussions on how to handle multicore processors, I recall that some people have asked for a board with SATA, and here’s one with oooooodles of memory.
The only downside is that until RISC OS can handle sharing out tasks among processors, it probably wouldn’t run much faster than a Beagle xM. Mind you, it may feel faster with a SATA drive instead of SD card…?

http://cache.freescale.com/files/32bit/doc/data_sheet/IMX6DQCEC.pdf

That’s good. So upgrading the firmware could be as easy as Pi.

It’s a good job you’re in France, closer and I might make an exception to my non-violence policy of 40+ years

I wasn’t thinking of anything specific. I recall our discussions on how to handle multicore processors, I recall that some people have asked for a board with SATA, and here’s one with oooooodles of memory.

clock speeds listed seem to be:
0: ARM – 792MHz, DDR – 532MHz, AXI -264MHz
1 :ARM – 396MHz, DDR – 352MHz, AXI -176MHz

setting 0 is less than the quoted 1GHz in the glossy bits

The only downside is that until RISC OS can handle sharing out tasks among processors, it probably wouldn’t run much faster than a Beagle xM. Mind you, it may feel faster with a SATA drive instead of SD card…?

Possibly less likely to have the storage naff up – spinny rust vs. exploding memory.

Of course it has SD in there as well.

PS. Big bonus, there’s a version of Ubuntu available ;) (Missile from France incoming?)

The only downside is that until RISC OS can handle sharing out tasks among processors, it probably wouldn’t run much faster than a Beagle xM. Mind you, it may feel faster with a SATA drive instead of SD card…?

It should feel a fair bit faster than a BB-xM, even running on just one core. Clock-for-clock Cortex-A9 outperforms Cortex-A8 by quite a bit (30%? 50%?) thanks to improvements like support for out-of-order execution.

clock speeds listed seem to be:
0: ARM – 792MHz, DDR – 532MHz, AXI -264MHz
1 :ARM – 396MHz, DDR – 352MHz, AXI -176MHz

setting 0 is less than the quoted 1GHz in the glossy bits

I’m not sure where you got those figures from, but the datasheet Rick links to shows that the specific SoC they’re using is capable of 1GHz (there are also other versions capable of 800MHz and 1.2GHz)

I’m not sure where you got those figures from, but the datasheet Rick links to shows that the specific SoC they’re using is capable of 1GHz

page 13 of the schematic. That was the third link I posted.

Even so the specification seems quite nice. What is needed is a HAL Howto that doesn’t involve much equipment then various people can play with their preferred toy and see how far they get.

Like this?

https://www.riscosopen.org/wiki/documentation/show/How%20to%20port%20RISC%20OS%20to%20new%20hardware

Like this?

Maybe a bit more fleshing out? I was thinking of something that halfway led people by the nose, then maybe you could sit back and supervise several efforts.

PS. I notice from a comment in one of the news articles the GigE port is limited to 480Mb/s by the internal bus. How would we cope with such limitations? :)

To be honest if someone reading that guide finds it insufficient then they’re probably not the right person for the job!

When creating a hardware port you’ll need to build up a lot of domain-specific knowledge (RISC OS internals, ARM architecture, target platform, etc.). There’s simply too much to teach in a simple how-to guide, you need to be someone who’s capable of researching and working things out on your own. You also need to be capable of debugging crashes on your own – you won’t be any good to anyone if you come crying to the forums for help every five minutes, especially if you’re the only person with access to your source code and the hardware to run it on.

Also don’t forget that creating a hardware port requires a lot of dedication. An initial port might take only a few weeks, but to finish it could easily take a year or more. E.g. it’s now 5 years since I started work on the OMAP3 port, and yet I’m still working on major features for it. With the limited developer manpower that we have available, we want people who’ll stick with projects until the end, not get bored halfway through and wander off and do something else.

Whatever happened to Michael Grunditz? He started an i.MX5 port for the Efika MX devices.

To be honest if someone reading that guide finds it insufficient then they’re probably not the right person for the job!

In my case, spot on. I did read it as a skim and I’ve corrected a typo in the last of the early bullet points.

I’m sure this has been mentioned but wouldn’t it be better to invest this amount of work in an A-15 port? Thinking of TI’s OMAP6432 EVM @ USD 329

@ patric – spec looks good, it also does SATA (don’t think you listed that). However, at $329 it may be out of reach for a number of people? I could consider saving up to buy one of the Matrix thingies. Given the usual $→£ rate, I simply could not justify paying that much for an ARM board when my eeePC is ages old and needs to be replaced. For that kind of money I could get a lower-end desktop PC which, really, I’ll need soon. The only reason I have not gone for one sooner is I played with the demo of Windows 8 in a local supermarket and hated it.

I might try to pick up a second hand box that dates from this decade (!) and put Ubuntu on it. [any offers? ;-)]

The processor is a dual core. I wonder if a well designed multitasker would get better performance from a fast dual core (it doesn’t appear to say what speed this clocks at) or from a slower quad core?

Might be worth reading the comments on the cnx page. :-(

I’m sure this has been mentioned but wouldn’t it be better to invest this amount of work in an A-15 port? Thinking of TI’s OMAP6432 EVM @ USD 329

Available from which UK source? and price?

It had a rather quiet release almost as though they had lost interest in the eval boards but already had that in the pipeline.

It had a rather quiet release almost as though they had lost interest in the eval boards but already had that in the pipeline.

Hasn’t TI sort of jacked in the OMAP? It looks like they have abandoned the mobile/tablet sector to concentrate on embedded. I don’t imagine that standard embedded devices need anything like the multimedia capabilities of the OMAP.

http://www.zdnet.com/how-tis-move-out-of-smartphones-into-servers-highlights-chip-issues-7000007439/

Hasn’t TI sort of jacked in the OMAP?

Yes, I think the EVM the Patric noted (5432 rather than 6432 I think) was in the pipeline when they made the announcement.
As earlier the choice of target for a port need to be readily available, preferably relatively cheap, and likely to continue in production for some while.

The Freescale in this instance looks like an item they’ve made for their multi-media use so may have a longer availability.

That’s true, TI seems to have dropped out in favour of the embedded market which currently leaves the Arndale board as the only A-15 alternative. Unfortunately Samsung aren’t very forecoming with documentation for their Exynos line.
It’s not exactly cheap either. I’d be happy to offer a £50 contribution towards a “no strings attached” hardware bounty though.
As Steffen Hubert said Michael Grunditz is already working on a Freescale port. The amount of work is massive however and there’s no huge advantage over the current Pandaboard. All IMHO of course.

I might try to pick up a second hand box that dates from this decade (!) and put Ubuntu on it. [any offers? ;-)]

I think there are still two ex WinXP boxes I need to get rid off. Not sure about specs, I put Ubuntu 10.10 on one and Puppy on the other but they’re fairly old, more like 2006 or so I’d guess.

That’s true, TI seems to have dropped out in favour of the embedded market which currently leaves the Arndale board as the only A-15 alternative. Unfortunately Samsung aren’t very forecoming with documentation for their Exynos line.
It’s not exactly cheap either. I’d be happy to offer a £50 contribution towards a “no strings attached” hardware bounty though.

There is an alternative A-15 board, or at least there will be once it’s released. It’s the IGEPv5 OMAP5432. Available in europe so should be easier to get hold of and hopefully cheaper. I think pricing is supposed to be announced some time in January. It’s very similar to the TI EVM board but has a higher spec. 2GHz cpu and 4GBytes of ram.
There’s also a lite version so should be more affordable. Same Cpu speed but 1GByte ram and no wireless.
All the details are here.

There is an alternative A-15 board, or at least there will be once it’s released. It’s the IGEPv5 OMAP5432.

The use of the OMAP may be a bit of a stumbling block – if TI have killed the embedded development (and their press releases say as much) the core processor could be in short supply

The use of the OMAP may be a bit of a stumbling block – if TI have killed the embedded development (and their press releases say as much) the core processor could be in short supply

The way I saw it was they can no longer compete with Qualcomm, NVIDIA and Samsung in the mobile processor business so are focusing on other markets.
There’s no sign of OMAP3, 4 or 5 going end of life.
From the point of view of enhancing and porting RISC OS they seem to be the best documented.
There are already ports using OMAP3 and 4 so it makes sense for the next port to be to the OMAP5. The IGEPv5 looks the best candidate so far. The main stumbling block is finding someone to actually do the port.

There’s no sign of OMAP3, 4 or 5 going end of life.

…but there have been “rumours” of shortages of parts. Perhaps the best bet at the moment is the OMAP4 (one, 4430?, is used in the Kindle Fire).

From the point of view of enhancing and porting RISC OS they seem to be the best documented.

Indeed. It is sad to see TI pull out on this fact alone. The Pi’s documentation is embarrassing in comparison.

There are already ports using OMAP3 and 4 so it makes sense for the next port to be to the OMAP5.

…and where did the Pi fit into this? I’d say it makes sense for the next port to be to something that people are likely to want to buy – perhaps when they need a little more oomph than the Pi can offer, but not at a price that is hard to justify. Of course, a lot of this also depends upon the quality of the documentation…

I just wish they would have coupled a good ARM processor to a useful GPU; underwhelming display capabilities seems to be something of a trend… http://xkcd.com/768/

but there have been “rumours” of shortages of parts. Perhaps the best bet at the moment is the OMAP4 (one, 4430?, is used in the Kindle Fire).

Doesn’t this just apply to the mobile phone industry where orders for processors are measured in the tens of millions. Dev boards are usually produced in the tens of thousands.

and where did the Pi fit into this? I’d say it makes sense for the next port to be to something that people are likely to want to buy – perhaps when they need a little more oomph than the Pi can offer, but not at a price that is hard to justify. Of course, a lot of this also depends upon the quality of the documentation

The RPi sits at the low end entry level starting at £28.

Next up from this is the Pandaboard ES starting at £145 which caters for the medium end.

What’s missing is a high end machine. This is where the TI EVM/IGEPv5 comes in. Starting at around the £250 mark. Higher processor speeds, more memory and much higher memory bandwidth. Also the potential of using USB3, SATA 2.0, eMCC and Gigabit Ethernet.

Dev boards are usually produced in the tens of thousands.

True. However if a part sees popular use, it is more likely to be in regular fab; because I don’t imagine TI would gear up a process to run off just a few thousand, when something else can be made in the millions.

Next up from this is the Pandaboard ES starting at £145 which caters for the medium end.

How about a medium range solution with SATA support for proper spinning rust?

Starting at around the £250 mark.

Where did you get that price? The isee site describes the options of the IGEPv5 with a lot of “up to” but it isn’t part of their active sales yet. It offers USB3 and an A15 at “up to” 2GHz, but there are too many variables to make a direct comparison.
As for the Arndale – the Octa doesn’t offer SATA, WiFi, or Gigabit. The other board does (except Gigabit) and the GPS/BT/FM/WiFi chip looks quite interesting. Though, a price to match.

That said – if RISC OS was available for the Matrix board, I would be interested. It looks like it offers quite a lot for the price.

I’d like one. I like the way you can just plug in a msata ssd underneath it – very neat. The box with the screen shown here is what I’d like.