Interesting "Competitors" for the Raspberry Pi

Hello

Have you heard about the Sabre-Lite board based on a Freescale i.MX6Q (Cortex-A9, Quad-Core, 1GHz) ?

i.MX6 is really a powerfull processor considering its 4x Cores @1GHz, its Graphical Unit Vivante for 2/3D, its Video Unit allowing 1080p multi-Encoding/Decoding and its Image Unit for multi-camera input/display outputs.

The IC was announced few months before but the launch is imminent (November 15th during Electronica) with samples available.

Sabre-Lite will be sold (with an aggressive resale price) and supported by Element14.

GregC, is the price of the forthcoming Freescale i.MX6Q SoC openly available as a ballpark figure?

I'm wondering what the cheapest possible evaluation board for it might cost, essentially just a basic Pi-type board containing your device and a PMIC in place of the Pi's BCM2835 and its linear regulators.

Morgaine.

GregC, is the price of the forthcoming Freescale i.MX6Q SoC openly available as a ballpark figure?

I'm wondering what the cheapest possible evaluation board for it might cost, essentially just a basic Pi-type board containing your device and a PMIC in place of the Pi's BCM2835 and its linear regulators.

Morgaine.

Freescale i.MX6 SoC pricing will be openly available after the launch, November 13th (during Electronica).

IC will be available in Single-Core, Single-Core-Lite, Dual-Core, Dual-Core-lite and Quad versions with different frequencies and qualifications (automotive, industrial and consumer).

Main differences between the versions are detailled in the table below (extracted from the iMX6 family fact sheet available on Freescale website).

For your information, Farnell/Element14 product page for the Sabre-Lite board is now active herehere

Your reflexion concerning a cheaper board, with close specifications to Rasberry-Pi, is interesting (I will check internally if there is some partners working on it).

Don't hesitate to share with me which features would you expect on such a board based on an iMX6 processor ?

For example, considering the efficiency of the ARM Cortex-A9 core (compare to the ARM11), would you recommend a Single/Dual/Quad-core version ?

How much RAM, how many USB port, SD-Card Socket, which display output (only HDMI or LVDS and Parallel), do you need the battery charger ? 

Depending from the connectors populated, I believe that it could be possible to improve a little bit the cost of the Wandboard which is already pretty agressive.

Oh Greg, you're cruel inviting us to spec a board, knowing full well that that is irresistible.

Greg C wrote:

 

For example, considering the efficiency of the ARM Cortex-A9 core (compare to the ARM11), would you recommend a Single/Dual/Quad-core version ?

Price is king when first introducing a new product and desiring lots of mindshare to develop around it.  There is absolutely no question in my mind, use the cheapest version and you'll sell the largest number which makes the BOM cost even smaller, a joyous vicious circle for everyone concerned.  Sure, i.MX6SoloLite is not as sexy as the top one, but that's what future models of the board can feature.  You have a perfect path for upward evolution based on those 5 SoCs.  So ...

• The cheapest, i.MX6SoloLite for Model A.

How much RAM, how many USB port, SD-Card Socket, which display output (only HDMI or LVDS and Parallel), do you need the battery charger ? 

How much RAM depends on the current RAM pricing sweet spot.  Since Pi moved to 512MB, I assume that this size must be at the sweet spot or just above it, because Pi is very price-sensitive.  The other items get "Use the cheapest" answer from me:

• 512MB RAM
• Just one USB.  Desktop users with old k/b and mouse can add a hub.
• SD card socket probably, but worth investigating root-only embedded flash instead.
• Regarding display, I'd prefer to listen to the rationale that others give for their choice.
• No USB charger input, that was silly design by RPF.  Barrel connector into your PMIC.
• If the board is connected to a USB host then of course your PMIC can power from there.

Depending from the connectors populated, I believe that it could be possible to improve a little bit the cost of the Wandboard which is already pretty agressive.

Regarding GPIO connectors, your engineers have already done all the investigation needed there and designed it into your Cortex-M0+ Freedom board, so that sounds like an excellent path to take here too.

Morgaine.

An alternative design approach that doesn't start with the Wandboard (because that might be hard to cost-reduce):

• Take a Freescale ZL25 Freedom Board, extend it a bit lengthways, add i.MX6, PMIC and RAM, and that's pretty much it.
• Leave the Cortex-M0+ on the board, it can do all the I/O.  Linux kernels are really bad at realtime response, that's what microntrollers are best at.  And you've already designed and routed everything we'll ever need for basic I/O on that board.
• All models in the i.MX6 family have on-SoC Ethernet (yay!), but leave Model A without RJ45+magnetics, again for cheapness.
• Launch Models A and B (which has networking) simultaneously, and base marketting on the Model A's price, as RPF did.

Morgaine.

• Don't think I agree with Morgaine on only 1 USB. The SOC usually supports at least 2 so why not bring them out to a connector?
• Prefer the option of having NAND on board to boot from
• For display having access to parallel RGB and/or LVDS is nice so a direct connection to a LCD panel is possible although these should be in addition to HDMI

Other than that I agree with Morgaine!

George Ioakimedes wrote:

 

• Don't think I agree with Morgaine on only 1 USB. The SOC usually supports at least 2 so why not bring them out to a connector?

For the sole reason that it's cheaper.  Left to choose based on what I personally would like, I'd just keep on adding more and more goodies.  The trouble is, pennies add up, and very soon you end up with a price that will not create a large community behind it.

If it turns out that a dual USB connector is actually no more expensive than a single (because they're made in exceedingly high volumes), then of course I'd opt for dual too.  Price is the only thing that I believe must be kept at the top of the requirements.

After all we have lots of expensive boards around already.  They don't make the cut.

dual USB is fairly standard, more so than single but it may still add $0.20 but I think it's worth the extra cost, more so than say a composite video jack

Hi,

It does look like a very powerful processor. And it has GigEth!

(Maybe have the correct integrated jacks to allow someone to externally develop a PoE addon to run the board. That's low priority). But maybe those magnetics cost the same (I have no idea).

I had these other suggestions. I was thinking, enough RxD/TxD pins brought out to have two serial interfaces at least, and then that way if anyone wants to use one of them as a "terminal" connection, they could connect an external RS232 chip or USB chip. The second interface would be good for interfacing other devices (e.g. GPS modules etc).

And some space for people to solder a clock CPU and XTAL and a couple of pins for external non-rechargeable cell : ) NTP is not always possible.

And some safe way to power down, e.g. some interrupt pin that will allow the OS to begin shutdown (just the main CPU, not the ancillary

one which we will be able to make sleep easily if needed), and send a signal on halt. Then we can implement wake-on-timer for example, with an external counter, e.g. to wake up hourly, do some heavy duty processing and then conserve batteries.

The ancillary and main CPUs could communicate using some low speed links.

Remember that the Wandboard already has all the niceties of lots of connections and great specs.  The only point in designing a bottom-end i.MX6 evaluation board is to bring the price down to a "so cheap you don't think twice" level (below the Wandboard Solo's $69) so that the boards are purchased by huge numbers of people which results in the creation of a good community base.  To achieve that requires chucking some nice things out, which is painful but without it the goal won't be achieved.

There's also an engineering reason rather than a marketting / social rationale for a cut-down board.  Not all applications require a fully featured board, and the extra features are wasted on them.  Just like the Pi Model A, a less capable board serves a useful purpose.

shabaz wrote:

 

It does look like a very powerful processor. And it has GigEth!

The gigabit Ethernet is a unique selling point for this product in ARM boards under $100, and will undoubtedly make this board the leading candidate soon for applications in which the 100Mbps of other cheap boards is a bottleneck.  i.MX6 seems to be leading the way there.

It's  worth pointing out that the i.MX6 SoC implements the Ethernet controller natively on the chip, so the Wandboard is going to be signiticantly more efficient than many other ARM boards which, like Raspberry Pi, implement Ethernet over USB with the help of another device like the LAN9512.  Networking performance can be expected to have lower latency and higher throughput even when operating at 100Mbps, as well as imposing a lower load on the CPU during operation.

I'm  looking forward to Wandboard with special interest because of this specific feature.

Morgaine.

OK, now I get it. If the goal is to get the price down closer to the RPi then I'm with you. Sorry if I missed it, but who is looking at building this i.MX6 low cost board? I too am eagerly waiting the Wandaboard but a lower cost version could be equally as attractive.