I know RAM is stacked with NAND on same chip but does TI or Samsung (or
whoever makes complete package) offer option with more RAM (256, 512)
and possibly zero NAND? I am just wondering if there is remote
possibility that once third parties start to reuse board design it could
be just matter of soldering different chip.
Beagleboard is still perfect for many solutions but I guess that the
desktop use case would benefit from more RAM a lot. Also since the chip
can boot from SD/MMC directly (and NAND is small and jffs2 has its
issues) it does make sense to omit the NAND altogether. Also the chip
has 3 SD/MMC interfaces so having SD or even eMMC expanding board as
permanent boot option while having 1 or 2 free card slots is doable.
So to summarize into few questions:
Are there more RAM/NAND options for OMAP 3530 part (now or planned)?
If yes, is it just matter of sticking different chip to same board?
If not, are there any other sensible ways to expand RAM?
Frantisek
Great. I would take 256MB SDRAM with zero NAND over 128MB RAM/256MB NAND
any day. If booting from MMC works, onboard NAND is just dead weight for
me. If you are considering this for revision C (or D), chalk me up. This
even aligns nicely with the idea of the board - strip everything you can
add externally. You can add flash storage but can't add SDRAM.
> However, managed NAND devices are more expensive than standard
> NAND.
Well, SD/MMC cards of equivalent size (256MB compressed jffs2 ~= 512MB)
are quite cheap, easily replaceable and may be even faster to boot (due
to jffs2 compression). This is true on current internet tablets and the
beagleboard slot has full mmcplus interface so full 8 bit I/O could give
even better speed. Also MMC cards can do 1.8V so maybe there is no
disadvantage even for embedded solution where 256MB of onboard NAND
would be enough.
Frantisek
|
The Ram, Flash and (CPU/GPU/DSP) are
stacked on top of each other in one package. Thats why they can make
the mother board so small. So I don't think it likely that a 256Mb Ram will be possible unless a new version on the omap chip
created. it's also effects cost power. I think the Flash Ram combo his here to stay --- On Tue, 29/7/08, Frantisek Dufka <duf...@gmail.com> wrote: |
|
If you look at the BOM list
http://www.beagleboard.org/uploads/Beagle_BOM_B4.xls there are two parts
55 IC_OMAP3430C_BGA_515 PIN U3 1 OMAP3530
56 IC, POP Memory 2Gb Flash/1Gb SDRAM, Micron U3x 1 MT29C2G24MAKLAJG-6
IT ES
So it looks like is just matter of finding different pop memory chip and
persuading someone to build such board if there is sufficient interest.
Nearest match I got for similar parts is
http://www.micron.com/products/dram/mobiledram/mobileddrpartlist
http://www.micron.com/products/partdetail?part=MT46H64M32L2CG-6
That is the only 2Gb one with status "Production" but maybe we would
need the 168 ball POP one (xxxxJG-6 (IT)) with status "Sampling".
They don't list combo chips directly
http://www.micron.com/products/mcps/
but at least they offer nice animation of this technology :-)
http://www.micron.com/products/mcps/mcp_webinar.aspx
> it's also effects cost power.
I wonder how much. Also I wonder what is price difference between 2Gb
Flash/1Gb SDRAM combo vs 2Gb SDRAM only.
> I think the
> Flash Ram combo his here to stay
>
I hope not :-)
Frantisek
Thanks for designing great board :-)
> The largest DDR we can get in POP is 2Gb, but it is DDR
> only and no NAND. This is where the complication comes in.
What is the complication exactly? Is is impossible to build board with
no onboard NAND? I was under impression that the boot ROM on OMAP3 can
load stuff from MMC directly without needing NAND at all. I don't see
problem with dedicating one existing SD/MMC slot to permanent rootfs. If
more SD slots are needed they can be attached over USB (easy, cheap,
fast) or the expansion interface has option for second MMC slot (third
one is not available?).
I see that for embedded board usage the 256MB NAND may be useful and
256MB RAM is overkill but still, as I tried to explain in previous
posts, cheap MMCmobile card in SD slot may be good (or even better)
solution even for such embedded use case. So why to bundle slow (mainly
because of jffs2) and relatively small NAND memory with the board?
This (maybe a bit radical) opinion comes from my experience with Nokia
internet tablets. The first thing I did on each tablet so far
(770,N800,N810) is copying rootfs from internal flash to mmc and booting
from it as main root filesystem. System is faster this way and there are
other advantages too (easier recovery, much more space available).
Frantisek
I see, forgot about SDIO, it certainly makes sense to leave this option open.
> We
> could add managed NAND onto the board through the second MMC port, but
> managed NAND is more expensive than the regular NAND. Our goal is to keep
> the cost as low as possible. If low cost is not a requirement, then a lot of
> things become possible.
Agreed. I admit I don't know what is the cost of various parts and
understand there are also other design restrictions. One example - Is
adding second SD/MMC slot socket cheaper then eMMC chip or NAND part
in POP package? Because that would provide more expansion
possibilities. Price of SD/MMC cards goes down, any hardwired NAND or
eMMC becomes small and outdated quickly.
>
> I will take your feedback into consideration for the next spin of the board.
Thanks for listening. Also sorry for being thick but does it mean that
there are no HW issues with replacing that ram/flash combo with pure
2Gb DRAM part so anyone else reusing current Beagleboard design 'as
is' can just use different part with no other HW changes? Thanks.
Frantisek