Wish List for Am3894 Beagleboard

GoogleGroups
1

Have you considered using the AM3894 on a future Beagleboard?
http://focus.ti.com/docs/prod/folders/print/am3894.html

The AM3894 is available through distributors because it’s already
listed on digikey and farnell with prices but with zero stock at
present. TI classifies the AM3894 as suitable for single board
computing amongst others.

Here is my wish list.

Standard Memory Module Socket(s)
Having looked at building our own beagleboard variant, pop memory
complicates manufacturing and therefore increases production cost to
some extent. And it’s not possible to increase memory capacity post
production.
Rather than using pop memory it would be great to have a standard DDR2/
DDR3 memory socket so that we could decide how much memory we wanted
to fit ourselves. Fitting a socket must be cheaper than fitting pop
memory. Would DDR2/DDR3 memory modules be cheaper than pop memory?
Does anyone have an idea? The AM3894 has DDR2/DDR3 support.

SATA Interface
Current beagleboards can only use flash based file systems via SD/MMC
or USB Flash devices or use a USB to SATA bridge to connect a hard
disk or SSD. Flash based file systems tend to be slow and I don’t
think anyone has reported that they have got a USB to SATA bridge
working successfully. It’s not referred to in the reference manual.
Correct me if I’m wrong.
It would be fantastic to see a beagleboard with a true SATA
controller. We could then easily connect either a hard disk or an SSD.
The AM3894 has an SATA interface.

PCIe 2.0 Interface
Current beagleboards have the expansion connector(s) which is great
and it should be retained. If the beagleboard had a PCIe 2.0 interface
you could buy many PCIe 2.0 peripherals. The main one I’m thinking of
is a wireless interface but it could be additional SATA ports or a
USB3.0 port when they finally arrive. The AM3894 has a PCIe2.0
interface.

Ethernet Interface
The beagleboard xm has a USB ethernet bridge chip. However it would
not be needed if the processor had an embedded ethernet controller. If
more USB ports are needed you could buy a USB hub and plug it in as on
the original beagleboard. The AM3894 has embedded ethernet interface.

Disadvantages of using the AM3894
According to digikey, the AM3894 is likely to be more expensive than
the AM37x but the extra functionally would be worth the extra cost.
And it’s a shame the AM3894 is not a dual core device. Maybe there is
one on the drawing board??

2

What is the connection between the AM3894 and DM3730, specifically Beagleboard? Did you see the complexity of AM3894 and Beagleboard?

What’s your idea about memory slot? Probably you saw the memory module installed at C6A16x/AM389x EVM? It is not a PC memory module and you won’t buy it at a local PC-store. It is custom design for this specific board. By the way in revision C of this EVM memory is installed on the main board.

If you need all the features available for the Spectrum Digital C6A16x/AM389x EVM then spend $2k and buy it. Do you expect guys from Beagleboard.org redesign AM389x EVM leaving all features unchanged and give you a price of $149? U’re a dreamer

2011/2/21 Andrew <andrew.czopiw@czworks.co.uk>

3

We have no plans for an AM3894 Beagle. Any futre Beagle will follow the -xM design. We will of course take yur input and consider it. Cost is the main driver on nay new board and there are other factors that will determine any direction that we go.

Based on your list, this will not be a low cost board. It will end up more hte the $250 to $300 range which is well outside or target audience.

Gerald

4

It's worth noting discussion on the TI forum regarding this:
http://e2e.ti.com/support/dsp/integra_dsparm/f/625/p/71675/324653.aspx
http://e2e.ti.com/support/dsp/integra_dsparm/f/625/p/92816/324936.aspx

5

Thanks for pointing out the discussion on the TI forum. It was very
informative i.e. a low cost EVM is on the way and that standard memory
modules cannot be used. Anyone with similar questions should check out
the above forums.

It’s a shame that the AM3894 is not a contender for a future Beagle
and I understand that cost is a major factor. Is there a discussion
or a web page that defines the future direction of the Beagle? Or
could you briefly tell us so we know what’s coming down the pipeline?

Will there be a Beagle with 1GB of RAM for example?

It has recently been reported in the trade press that Microsoft has
allegedly 1000 engineers working on porting a future Windows 8 on to
ARM. I’m sure if this is happening then major computer manufacturers
will have ARM motherboards on the drawing board with Memory Sockets,
SATA, USB, Ethernet, Wireless, PCIe and hopefully the more usual SPI,
I2C and UART interfaces too.

Nufront already has a working prototype for such a device which will
sell for $250 according to Nufront. No need to buy expensive EVM’s at
$2K. Check out the NuSmart 2816 ARM device.
http://www.nufrontsoft.com/en/cpzx/eed31e97-d916-4441-8aa1-6f6413a692f9155.html

If anyone has successfully connected a Beagle to either a SATA hard
drive or SSD I would like to hear how they achieved it.

6

Ok, so there will be ARM based computers with Windows 8,9 or whatever, equipped with PC compatible PC sockets, SATA PCIe and a lot more… So, what’s then? Do you expect the same performance or functionality as contemporary PCs provide?

What do you want from ARM architecture? x86 is far-far more advanced, powerful, reliable and of course hot in a manner of heating. You expect to have $200 single board solution based on ARM and fully similar as x86?? May be you just buy ATOM motherboard for $70 and $20 for memory and would not blow minds of good people here? :slight_smile:

Every task has a solution. You solution can’t be based on the ARM architecture. ARM is for mobile platforms. PCIe and SATA is not for mobile applications.
If you need x86 SOC have a look at DMP Vortex86dx chips.

2011/2/23 Andrew <andrew.czopiw@czworks.co.uk>

7

I started this thread as a wish list for a future Beagle and with a
suggestion to use the AM3894 because of its extended peripheral set
which I’m interested in using.

Other members of the forum have contacted me directly regarding this
thread and have they been polite and very helpful with their advice.

Many other members in this forum have asked about mounting hard disks,
SSD’s and wireless adapters to the current Beagle but I’ve not read of
any successful attempts. Hence the questions.

If you look at TI’s published roadmaps or press announcements they all
talk of Cortex-A8, Cortex-A9 and Cortex-A15 ARMs, some with dual
cores, some with SATA ports and some with PCIe ports. So a discussion
of a future Beagle with these peripherals is legitimate on this forum.

ARM is no longer for mobile applications. See ARMs website for their
view.
http://www.arm.com/products/processors/cortex-a/index.php

If I wanted an x86 SOC I would have bought one.

8

Attach a USB to SATA thingy and the SSD just works. Same goes for wireless dongles.

9

Andrew, ARM corporation can image anything about their processors applications but this does not change the truth of life. ARM processors are just intended to work in low power environment and they can provide only low performance for all of fast or slow devices connected to them.

Indeed you can connect any USB SATA or SSD drives and they work out-of-box because Linux recognize them without user efforts. But, what performance can they give? They are just slightly faster (or even not) comparing with conventional SD cards.

2011/2/23 Koen Kooi <koen@beagleboard.org>

10

Ok, so there will be ARM based computers with Windows 8,9 or whatever,
equipped with PC compatible PC sockets, SATA PCIe and a lot more…
So, what’s then? Do you expect the same performance or functionality as contemporary PCs provide?

Am I missing something here? from your post it seems that not even the ATOM platform should exist

I suppose what ARM aims with the new multi-core processors is to be a serious option to Atom or Tegra, I dont see why this is necessarily a bad thing (think of netbooks, UMIDS, infotainment, you will not run VS 2010 IDE at the same time with your video editor along with your fermi spec’ed game at the same time on that platform, but if it does what it currently does but faster, that is a good thing in my book )

Lioric

11

gerald,

i’m re-visiting this thread because of all of TI’s processors, the AM3892/4 is the closest to EOMA-68 and being FSF Hardware-Endorseable that there is. i am evaluating getting a FSF Hardware-Endorseable tablet/laptop PC made up - EOMA-68 compliant - crowd-funded through the FSF. can i therefore please ask you some questions?

  • would TI be interested to aid and assist such a project - the key part being the creation of the CPU Card - by e.g. providing reference design schematics (ORCAD and PADS) for an AM3892/4 CPU Board, or pointing me towards a resource where such files already exist?
  • is the AM3892/4 part of the “core embedded range by TI” i.e. is it exempt from the deal we heard about Amazon being in advanced negotiations to buy out parts of TI’s ARM SoC division? perhaps a better question would be “will TI honour its product lifetimes and lifecycles on existing AM Sitara ARM SoCs”?

as this product would be sold primarily to software developers and FSF supporters, there would be pretty much zero requirement for T.I to spend time and resources on software support. my key concern however is the hardware. beagleboards etc. are dead-easy to convert to other form-factors: download the files and go. the AM3892/4?

really not kidding about the suitability of this SoC for EOMA-68, btw. SATA, Gigabit Ethernet, HDMI, USB2x2 etc. - perfect. video in is a bit of overkill, as is PCIe, but there you go.

l.

12

This is not a device on the currently on the BeagleBoard.org road map. If we made a board for all the different processors that TI has, we would have to turn out a new board every month.

So, as far as having something ready to go on this similar to a BeagleBoard, I know of no such effort. There should be EVMs available for the processor. Generally speaking TI does not undertake such an effort as this. There should be other companies out there that, if the market was large enough, may be interested in such an effort.

It is a different product line as to the one that somebody is mentioning in the rumors. But, I don’t know what the long-term support plans may be. I am not involved in this product line. Your best bet would be to contact you local TI sales person and let them work back into TI and see what support/interest there may be for such an effort.

Gerald

13

This is not a device on the currently on the BeagleBoard.org road map. If we made a board for all the different processors that TI has, we would have to turn out a new board every month.

sorry, gerald, i should have been clearer: i’m not interested in beagleboards, nor in the beagleboard roadmap. however the EOMA-68 board that i am considering getting made, using an AM3892, would have all other pins not going out of the EOMA-68 interface (at one end) or HDMI, USB, SD/MMC (at the other end) would be placed onto jumper headers in a similar fashion to a beagleboard.

basically i read what was written: i hear what you have said: TI is not interested to, itself, create a beagle-board-like board using the AM389x. what i want to know is: will TI support an initiative done by someone else to make a beagle-board-like board using the AM389x (one that is conformant to the EOMA-68 specification).

So, as far as having something ready to go on this similar to a BeagleBoard, I know of no such effort. There should be EVMs available for the processor.

great. question. can the schematics for those EVMs [for the AM389x] be made available so that i can pass them to engineering teams and get them to price up the creation of a beagle-board-like board that is conformant to the EOMA-68 specification?

if we can get those EVM schematics made available under a Free Software License then like the Beagleboard schematics we will ensure that the resultant schematics are also made available under the same license i.e. made publicly available.

Generally speaking TI does not undertake such an effort as this.

if TI is prepared to get us the EVM schematics, we’ll undertake the required effort.

There should be other companies out there that, if the market was large enough, may be interested in such an effort.

yes: the one that i am working with would be interested in such an effort, and in doing the market research, and in leveraging the free software community resources, and creating the P.R., and working with the Free Software Foundation etc. and so on.

It is a different product line as to the one that somebody is mentioning in the rumors. But, I don’t know what the long-term support plans may be. I am not involved in this product line. Your best bet would be to contact you local TI sales person and let them work back into TI and see what support/interest there may be for such an effort.

my local TI sales person isn’t responding, i don’t know why. could you tell me: who makes the decisions about releasing of EVM schematics, and could you put me in touch with them? it will save us at least $10k in development costs to have some working schematics. adaptation of existing schematics is about $6k. starting from scratch would take months and would be around $20k.

l.

14

I have no experience in TI not supporting customers in anything they want to do. If you were to create this product, I am sure TI would be willing to assist in promoting it to other potential customers. As to the EVM schematics, I believe they are indeed available. If my memory serves me correctly, Mistral designed the EVM so you should be able to get he schematics from there.

Again, the shortest easiest path on this is to work with your local TI sales contacts to get you to the right people. If you have trouble getting some traction, you may contact me direct.

Gerald

15

I have no experience in TI not supporting customers in anything they want to do. If you were to create this product, I am sure TI would be willing to assist in promoting it to other potential customers. As to the EVM schematics, I believe they are indeed available. If my memory serves me correctly, Mistral designed the EVM so you should be able to get he schematics from there.

ah brilliant! and, actually, as they have experience with the chip, they may actually be the best people to quote for the conversion to EOMA-68. thanks gerald: good lead.

Again, the shortest easiest path on this is to work with your local TI sales contacts to get you to the right people. If you have trouble getting some traction, you may contact me direct.

all right sir. will keep you informed. thank you.

16

I have no experience in TI not supporting customers in anything they want to do. If you were to create this product, I am sure TI would be willing to assist in promoting it to other potential customers. As to the EVM schematics, I believe they are indeed available. If my memory serves me correctly, Mistral designed the EVM so you should be able to get he schematics from there.

ah brilliant! and, actually, as they have experience with the chip, they may actually be the best people to quote for the conversion to EOMA-68. thanks gerald: good lead.

hiya, gerald - hum, apologies, i haven’t heard back from mistral, it’s been almost a week. i’ll try again, let you know how it goes.

Again, the shortest easiest path on this is to work with your local TI sales contacts to get you to the right people. If you have trouble getting some traction, you may contact me direct.

oh btw - i’m waiting to hear back from mistral before proceeding here. apart from anything, “local” TI sales contact for me might not be appropriate: the design work and sales could take place in the USA, or germany, or the UK - anywhere: it just depends on whom we find to do the board.

l.

17

http://support.spectrumdigital.com/boards/evm816x/revg/

ahh ha haaaaa :slight_smile:

18

andrew, hi,

we’re evaluating the feasibility and the desirability of an AM389x beagle-like-board, and wanted to know if you (or anyone else) would be interested to buy it.

Here is my wish list.

andrew, let me go through this, and you let me know if you’d be interested to buy one, ok?

Standard Memory Module Socket(s)

Rather than using pop memory it would be great to have a standard DDR2/
DDR3 memory socket so that we could decide how much memory we wanted
to fit ourselves. Fitting a socket must be cheaper than fitting pop
memory. Would DDR2/DDR3 memory modules be cheaper than pop memory?

it’s a nice idea, but:

a) there would need to be a published list of “compatibility” maintained
b) we’re planning to make the PCB approximately 75mm x 45mm: there’s simply no room. it’s just possible to get 4of 2Gbitx8 82-pin TFBGA RAM ICs onto a 75x45mm board, if you’re prepared to put the NAND flash on the other side.

so we’ll be putting on 1gbyte of DDR3 RAM as a 32-bit interface, probably 1066mhz but if the prices of 1600mhz DDR3 2GBitx8 RAM ICs aren’t completely insane i’m all for putting those on.

SATA Interface
It would be fantastic to see a beagleboard with a true SATA
controller. We could then easily connect either a hard disk or an SSD.
The AM3894 has an SATA interface.

and it’s exactly why i picked it as an EOMA-68 candidate. i plan to put the SATA of the AM3892 out of the EOMA-68 interface: you would be able to put a micro-engineering board header onto the EOMA-68 interface, with an eSATA connector being on the micro-engineering board.

PCIe 2.0 Interface
USB3.0 port when they finally arrive. The AM3894 has a PCIe2.0
interface.

yes. i will try to see if it’s possible to cram in an expansion header: space is extraordinarily tight for the form-factor we’ve picked. it may have to be an absolutely tiny 1mm DIL-50 connector. bear in mind that the RGB/TTL will be out the EOMA-68 interface, and the HDMI interface will be on the other end of the PCB, as well as a Micro-USB and a Micro-SD slot, so there won’t actually be that many functions needed to go onto the expansion header.

Ethernet Interface
the original beagleboard. The AM3894 has embedded ethernet interface.

yes. again, this is why i picked the AM389x. this again will be routed to the EOMA-68 interface, where a micro-engineering board can be plugged in with an RJ-45 socket.

Disadvantages of using the AM3894
According to digikey, the AM3894 is likely to be more expensive than
the AM37x but the extra functionally would be worth the extra cost.

And it’s a shame the AM3894 is not a dual core device. Maybe there is
one on the drawing board??

TI don’t need to. the AM3894 is a very specialist device, for security cameras, cinemas, real-time (RAW!!) high-definition video recording and so on. it’s not actually the ARM CPU that needs to be fast for those purposes because everything that’s “special” about this SoC is done via DMA.

it just so happens that those features make it a stonkingly-good SoC for general-purpose computing. i’m just a bit alarmed about the 1,000+ pincount: typically the yields on such larger ICs tends to be lower, pushing up the price. ahh, can’t be helped :slight_smile:

l.

19

ok, i’ve set up a description page where i’m inviting people to comment on it, and also perhaps register preorder interest if they so wish:
http://rhombus-tech.net/ti/am389x/

there’s also a preorders page, if people are interested.

20

http://rhombus-tech.net/ti/am389x/

right - after finding the schematics from spectrumdigital, here:
http://support.spectrumdigital.com/boards/evm816x/revg/
i’ve rapidly modified them into a roughly EOMA-68 style by cut/paste
from the ones i did for the A10 CPU Card, and have uploaded them here:
http://hands.com/~lkcl/eoma/ti/am3892/AM3892_ORCAD_EOMA68.DSN

if anyone has ORCAD and can do “print to PDF” (i can’t) that would be real handy. also some input about which interfaces would be desirable on say a 50-pin DIL2 1mm pitch connector would be handy.

i’ll be handing this over to someone else to verify that the schematics are ok, and also getting them to do the PCB layout, that’s waaay outside my expertise :slight_smile:

l.