Actions
Forum Thread Details
  • Replies 20 replies
  • Subscribers 59 subscribers
  • Views 2385 views
  • Users 0 members are here
Related

Minnowboard

Former Member

Just received the minnowboard today, so here's a couple of first impressions.

 

Two and a bit times the size of a BeagleBone Black, but still tiny for a x86

 

LED's are green, not blue image

 

Angstrom. image

 

5v input takes 1.5A sitting at the desktop with just monitor, keyboard/mouse and 1G network connection.

 

It runs warm.

 

There's a mac address on a sticky label, but the board seems to have a different one, looks like it could be a default or test one: DE:AD:01:FF:xx:xx

some stuff in the yocto mailing list suggests there were firmware problems leading to the mac being stored in an EFI variable at one point, so I'll have to dig into that.

 

'reboot' doesn't, probably an angstrom bug.

 

No rtc battery included.  It's annoying that other boards (RPi, BBB, Sabre-Lite etc) don't have useable rtc's at all, but to have gone to the bother of having the battery holder and not include a battery is just annoying.

 

Supplied with psu that has a US 2-pin plug on it.  While I understand these are made in the US, that's not really a good enough excuse when they're being sold elsewhere.  The SL did the same even though it's built in China.  e14 if you're listening, stop doing this, these useless psu's are going straight to landfill.

A good example to follow would be the freescale i.MX5-QSB which came with a PSU that had a swappable plug and enough adapters to cover most of the world. Otherwise, rather than a wall-wart, just put in a PSU that has a standard IEC inlet. Letting the user source a local power lead would make lots more sense.

 

Seemingly no Lures (their version of cape/shield) actually available at this point.  When you follow the links through the wiki to get a part number for the required connector you get a digikey part number, you then have to translate that to the manufacturers (TE / AMP) part number, then find the version of that part number that isn't a reel of 1000.  Nice one CircuitCo.  For anyone who might be interested it seems that uk.farnell.com part numbers are 10983481098348 & 10983531098353 for 5mm and 9mm high versions.

Usefully, there are 8 gpios on a normal 0.1" header (duplicated on the high density connector too) so some out of the box experimentation is possible.

The one Lure you're likely to think essential is the BoB (or Break out Board),  http://www.elinux.org/Minnowboard:Bob_Lure luckily it's quite simple and the schematics & gerbers are available, so reasonably possible to build your own. I have the feeling that I'll modify it slightly to bring out the second sata port at the very least.

 

They have the same 'dual-use' warning on the despatch note as Beaglebone Black. This will likely mean extra paperwork when you order one, depending on where you live, see http://www.element14.com/community/thread/26383?tstart=30

 

Some of you will no doubt be thinking I've gone mad, given the price of this. You may well be right image 

So here's a couple of the reasons I think it's interesting:

Being x86 opens up certain choices in OS, you don't need to wait for someone to port Raspbian for example.

Having an intel gpu, means driver support is likely to be better. (How good, or bad, the gpu is remains to be seen.)

Intel have been pushing support for chipsets like this into the mainline linux kernel for a number of years, so less likelyhood of being stuck on some ancient buggy kernel while a vendor procrastinates.

 

The downsides are higher power consumption, some x86 baggage and UEFI

 

Some photos:

image

 

image

image

Yes, it comes with the standoffs and feet..

image

Parents Reply Children
  • in reply to morgaine

    Morgaine Dinova wrote:

     

    When you start playing with the software (preferably before moving away from the default Angstrom), it would be cool if you could add a line for Minnow to our simple benchmark data thread.

    Ah yes, I was looking for that earlier image

     

    I'd like to run it, but the version of bc in angstrom seems to be broken.. When I try to it just segfaults, doesn't seem to be specific to the calculation as all calculations segfault as soon as mathlib is asked for, even 1+1

    I'll try building bc from source later. First thing is to get some form of sshd running so I can access it without a monitor and a walk into another room...

  • in reply to morgaine

    Morgaine Dinova wrote:

     

    preferably before moving away from the default Angstrom

    how did you guess image

  • in reply to Former Member

    selsinork wrote:

     

    I'd like to run it, but the version of bc in angstrom seems to be broken.. When I try to it just segfaults

    Something broken in Angstrom?  Never!!! image image

  • in reply to morgaine

    Morgaine Dinova wrote:

     

    Something broken in Angstrom?  Never!!! image image

    Well you've probably guessed by now that there is actually a dropbear ssh server configured, however it's broken too, the log shows

     

    dropbear[568]: Premature exit: Bad buf_getptr

     

    I think I'll go download a new copy of angstrom and put it on a known working sd card then try again.  If that doesn't work then I'm installing some known working software to rule out a hardware problem.

    Not a good start for angstrom though.

  • in reply to Former Member

    writing a newly downloaded copy of seemingly the same image to a new SD card seems to have solved the dropbear sshd problem, bc however still seems broken.

     

    So it would appear that the supplied kingston sd card either wasn't written properly or is just a low quality card

  • in reply to morgaine

    Morgaine Dinova wrote:

     

    When you start playing with the software (preferably before moving away from the default Angstrom), it would be cool if you could add a line for Minnow to our lightweight benchmark data thread.

    approx 21 seconds.

     

    This is with bc 1.06.95 rebuilt from source on the minnowboard with -O2 flag since the angstrom default one segfaults whenever the -l switch is used.  Otherwise the OS is production-Angstrom-development-GNOME-image-eglibc-ipk-v2012.12-minnow-2013.07.10.img

    Interestingly, the same version of bc built on a 32 bit x86 desktop also segfaults on the minnow, so it makes me wonder if this is a problem caused by angstrom being cross compiled but having carried over something from the host.

     

    The additional observation is that the system feels sluggish, not really sure if this is simply due to running off sd card or something else.  Looks like there are some hdd images available, so I'm going to try one of those later as well.