OlinuXino LIME vs BBB ?

Excellent board and excellent price!

By the way, I don't think you should list "100Mb/s network" as a Downside, because you titled your thread "LIME vs BBB", and it's not a downside vs BBB since BBB also has the same nominal link speed.

I agree with your intended sentiment of course, that all networking should be gigabit or faster.

selsinork wrote:

 

 

element14Dave any chance of passing along to your procurement people that we'd really love it it e14 were to offer more of the OlinuXino range ?

Seconded -- I want to buy several of these, as I did BBBs.  I don't think that there's much hope of it though.  Farnell UK isn't even listing the now-established A20-OLinuXino boards yet (let alone stocking), so LIME probably won't appear until Q2 2014, if then.  Farnell procurement seems to be moving at snails' pace for some reason.  It's so bad that Farnell is beginning to look silly, and I'll have to start looking elsewhere.

Morgaine.

Yes, 100Mb/s network is just a general thing.  I've been compiling a lot of software on some of these boards recently and the bits that stand out are that none of them have the memory for me to be able to setup an 8GB ram based tmpfs on /tmp and use that to speed up compiles the way I would on x86, so I'm looking for either decent SATA performance, or gigabit network so that I can nfs mount /tmp to a large ram backed store on another system.

With the 3.12 kernel, the SL has just become my new favourite build machine, even though it's taken a full year to get there. I'm aiming for a common kernel & userspace across BBB, SL & Cubie2.  It's taken a while, but BBB & SL seem to be getting their acts together and are reasonably close to mainline, the A10/A20 stuff is still way behind with 3.4 kernels even though there's a good community around sunxi.org, hopefully they'll catch up soon.

Like Ubuntu/Fedora etc, I'm rapidly reaching the conclusion that the critical mass around Cortex A8 class onwards is a good thing and that dropping the RPi and it's ARM11 is the right decision. Final nail in that coffin springs to mind...

If you're still interested in more BBB's, CPC have a modest offer on them from last weeks mailer. Only a couple of quid off, but still... valid to the end of January IIRC.

Hi,

just want to make small update: today we assembled LIME with A20 and the result is: A20 Dual Core Cortex-A7 @1Ghz need just 0.14A@5V vs. Single Core Cortex-A8 @1Ghz needs 0.18A i.e. A20 is 30% less power hungry and runs about x1.7 times faster than A10

A20-OLinuXino-LIME update | olimex

when comparing to BBB also note that A10/A20 HDMI is native and supports any resolution up to 2160p while BBB uses converter IC and have limited resolutions, LIME have LiPo charger and can run 30 hours on our 6.6Ah LiPo battery

and of course the native SATA

Tsvetan

Morgaine Dinova wrote:

 

I just need the Sabre-Lite to be cheaper!

Agreed. However it's a bit of a compromise for me... The cost of one (quad core) SL is equivalent to four-and-a-bit BBB's.  So does the Cortex A9 + now decently performing SATA outweigh setting up distcc + getting everything on nfs?  It's a hard decision..  Now if the cost of one SL got me 5x BBB it'd be different.

The LIME changes things again, that 30 euro price tag suggests I'll get five-and-a-bit of them for the cost of one SL, plus the LIME has SATA while the BBB doesn't.  So for compilation duties the LIME could end up being better than both.

Don't get me wrong, for all sorts of other things the BBB with onboard eMMC can be a better choice. We need to see what the LIME with onboard 4Gb NAND comes in at to pick a winner.  Like you say, the PRUs will decide for certain applications..  For things like my RPi GPS NTP boxes, all I need is one GPIO for PPS, the ability to run ntpd and for it to run reliably forever.  The recent nonesense with an RPi kernel update making a few of mine unbootable due to no longer liking the previously working SD cards is the sort of thing that makes you think that onboard eMMC or NAND in the case of LIME are good solutions for reliability - they're bound to get tested!

Been a while since I posted any pics of this, but the collection is growing, I'll be needing to build the expansion soon

Each carrier looks something like this:

Hi Tsvetan,

Your A20 LIME looks really interesting. I have an A20 based Cubieboard2 and have some issues with the kernel locking up randomly after multiple days. I suspect it's simply a case of the sunxi guys needing to get a bit more of the A20 kernel sorted out, but once that happens I'm certainly looking forward to more A20 based boards.

Since you're working for Olimex, I don't suppose you can tell us if e14 will become distributors for your A20 boards as they have for some of the other OlinuXino boards ?

Tsvetan Usunov wrote:

 

Hi,

 

just want to make small update: today we assembled LIME with A20 and the result is: A20 Dual Core Cortex-A7 @1Ghz need just 0.14A@5V vs. Single Core Cortex-A8 @1Ghz needs 0.18A i.e. A20 is 30% less power hungry and runs about x1.7 times faster than A10

 

A20-OLinuXino-LIME update | olimex

 

when comparing to BBB also note that A10/A20 HDMI is native and supports any resolution up to 2160p while BBB uses converter IC and have limited resolutions, LIME have LiPo charger and can run 30 hours on our 6.6Ah LiPo battery

and of course the native SATA

That's excellent news, Tsvetan!

You're going to have a big seller on your hands with LIME, and I'll definitely be helping to exhaust your stocks.

The lower power consumed by A20 attracts me a lot, because I'll be using the LiPo facility as a 2nd stage UPS and so operating time under backup power will improve too.  It probably runs cooler as well, which is always a good thing.  And dual cores can't hurt even if the application doesn't need them.

But I'll be getting a few A10 LIME anyway, since that's coming out first.  Can't wait!

Morgaine.

selsinork wrote:

 

Since you're working for Olimex, I don't suppose you can tell us if e14 will become distributors for your A20 boards as they have for some of the other OlinuXino boards ?

I second your question to Tsvetan.

I'll also add that olimex.co.uk has not updated the Amazon UK offering either, so our domestic availability of A20-OLinuXino-MICRO[-4GB] (both versions) is currently non-existent from any major retailer, AFAIAA.

Morgaine.

@selsinork we have lot of A20-OLinuXino-MICRO sold and while doing extensive tests we found that about 20% of the processors fail after some hours of work at higher than 400Mhz DDR3 settings, this is why we set our memory settings to 380Mhz to be on the safe side, if you check cubie settings they are at 480Mhz IIRC which if our statistic is correct will cause troubles to 20% of their customers, I don't know why they do this as I have discussed this on IRC list and Tom is aware for this SOC issue.

Some A20 work reliable up to 532Mhz DDR3 settings but these are let say another 20% out of all processors we tested, so I guess you are not lucky and have got some of these SOCs which fail above 400Mhz DDR3 settings

@Morgaine Dinova we can't influence what our distributors to stock, Farnell have lot of items in stock and they have procedures to decide if to add component in stock which takes time, so if/when they will sell A10-LIME is up to them

olimex.co.uk sells mostly on ebay and they usually stock all new boards 1 week after we release them, if you prefer shipment from UK they are good alternative

tsvetan very interesting information, thankyou. 

You were of course correct, dram was set to 480. the sunxi-tools docs suggest it needs to be multiples of 24, so I've just set it to 384. I'll have to wait and see what effect this has, it's taken up to 5 days to lockup previously.

I'll have to check what the android image in NAND is using as well. It seems much more stable, but disables one CPU core and I'd initially been wondering if that was related as I've seen a few references to A20 SMP support still being WIP on sunxi.org

I just checked the supplied android image and that sets the DDR3 to 432MHz instead of the 480MHz that the debian image they supply uses.  So seems you were right, my particular board isn't happy at 480MHz but hopefully is at 432 or less.

Seems like an interesting way of working. Anyone reporting instability problems gets asked to try booting the pre-installed Android image which of course works due to much more conservative settings.

Just a quick update: the CB2 locked up again with the ddr clock set to 384. So while the memory speed may be a problem, there's some other factor as well. Time to try a new kernel I think.

So after getting an A20-OlinuXino-MICRO, I borrowed it's 3.4.x kernel and the CB2 seems to be stable with that.

Allwinner kernels at least up to the 3.4.67 series from sunxi.org don't use devicetree, instead using a script.fex that's plaintext and very much resembles the old ini format. You compile the file to script.bin and u-boot loads it for the kernel to use. This not only makes it easy to copy one for a different board, but at least in my opinion makes it much easier to change.

Moral of the story yet again is to ditch what comes with the boards and get something that works.. Unless it's an Olimex board - their stuff seems to work just fine!

So after getting an A20-OlinuXino-MICRO, I borrowed it's 3.4.x kernel and the CB2 seems to be stable with that.

Allwinner kernels at least up to the 3.4.67 series from sunxi.org don't use devicetree, instead using a script.fex that's plaintext and very much resembles the old ini format. You compile the file to script.bin and u-boot loads it for the kernel to use. This not only makes it easy to copy one for a different board, but at least in my opinion makes it much easier to change.

Moral of the story yet again is to ditch what comes with the boards and get something that works.. Unless it's an Olimex board - their stuff seems to work just fine!