And finally...

Texas instruments memory selector:
http://processors.wiki.ti.com/index....election_Guide

So, if i'm not wrong, this means that with the DM3730, to reach 1GB of ram we must use 2 x ram chips of 512MBytes selecting the first or the second bank with the CS (chip select) pin.

32GB eMMC Flash + 1GB RAM: http://www.samsung.com/global/busine...=7626&iaId=860
The series offers other combinations too, many of which still not in mass production. Some are allegedly available as POP (dunno if for OMAP).
Anyway a seemingly viable alternative to our flash-only eMMC in N900 that offers same flash storage capacity plus an additional 1GB of fast RAM. http://fuchaoic.en.alibaba.com/produ...00VM_B604.html

CAVEAT: it's still not 100% clear to me that Samsung actually means (M/G)Byte and not bit (as storage manufacturers usually do), in those docs.
Anyway look here for what I consider is a PoP for OMAP et al. Max 512MB RAM. So Neo900 most likely will have 512MB PoP RAM, plus 1GB of really ultra fast swap, which should be even better than 1GB of plain RAM with no swap at all or flash-based swap that's extremely slow. For all practical purposes the user experience will be that of a device with 1.5GB RAM (and we don't need to do any special trcks to maemo5 kernel and core system, it just uses 512MB RAM with 1GB of ultrafast swap - instead of 256MB RAM with 768MB of extremely slow swap as is now on N900).

Evaluation if such chip even can get connected to Neo900 DM3730 SoC and circuitry is ongoing! (usual problem: no datasheets publicly available)

Now, that one is just creepy. I really, really want this modem inside Neo900!

Out of curiosity - aren't you oversimplifying things a little? Is really eMMC bus as fast as normal CPU <-> RAM connections in case of this planned board, to handle RAM-based eMMC area as quickly as real RAM?

Maybe I'm wrong, but it seems to me, that if it would be the case, we would always have RAM's as eMMC's/HD's, without special place for them on POP, or "those" ram sockets in desktops.

/Estel

// Edit

In practice - how fast eMMC bus is, as compared to regular, practical bandwidth of access to real RAM?

I'm simplifying. Sure. Over? Prolly no. You get the bandwidth the RAM on this MCP can do, from the datasheets I linked to. Do the math! And you're probably aware that swap gets paged in and out in chunks, not requiring random access from CPU. You could even figure how CPU would happily work along on another thread, while DMA does swap in/out a needed page concurrently. And you know that swap is not used 10 times per second but only occasionally when you're switching between huge tasks. The (very common) worst case delay introduced by this fast swap is 2 to 5 magnitudes smaller than that introduced by flash based swap which has delays on the nnn milliseconds range for writes.
I clearly said "user experience", _not_ device speed as measured by wetstones or bogomips.

No problem! The friendly aspects of a tiny batch. We build one board without modem for you, you get yourself one GTM609 and mount it, I know you're capable of doing that.
Or maybe we can even build to order, doesn't make much of a difference either. We need to keep track of wich modem goes on which board anyway, if only to write the correct IMEI on the package ;-)

First off I would like to say how much I enjoy the community here at maemo. This is one of the most interesting pojects I have ever had the pleasure of being a part of. I am even more excited to see that joerg_rw is helping to head the Neo900 project. I started a thread a few months back in the brainstorming forum about a possible kickstarter funded device that we here at maemo could call our own. much to my surprise joerg_rw has this very much under controls and I am looking forward to seeing this project take off. Big thanks to joerg_rw i have been fallowing your work since I bought my n900 just as the repositories were being migrated. Big thanks keep it up

I agree Estel...I want one (gtm609) in mine too.
creepy is a good word...downright spooky even.
I notice (in all humility ...with my ever-so-small-understanding) at the least ...it appears to be a freakin' "Planet Walker" ...go anywhere with it in your device..Boston to Frankfurt to Timbuktu to Beijing..and places with less than modern coverage ..more rural..remote areas even..i imagine....want to check that out....hmm... need to read-up more and wrap my feeble understanding around what this modem is all-around capable of....

Just one of many write-ups on the gtm609 I found... good press release statement too.

http://www.engadget.com/2011/01/05/o...allest-with-v/

2 yrs ago is when this article was writ...I'm sure there is even more cutting edge modems out there...but compat. and price are prob. astronomical. Love what I'm reading though and the more I'm reading the more I want this sucker.

Hmmm...I have noticed over the last week or so that my Pavlovian salivation response which used to encompass the basics such as anyone shouting "BOOZE!", "FOOD!" and "WIMMIN!" now seems to include "NEO900!" I'm not sure if I should be disturbed by this new development ...seek help ...and get medication.......... or simply embrace the Drool and get a napkin.

i want to know how many people want to buy.and i am sorry that only one year after can i buy even i want it

Right now I wouldn't buy one of these because I already have a working N900 and no job right now and cant justify spending $ to replace a working phone with a better phone. But when my N900 fails and cant be fixed, I would be very much interested in a project such as this (as long as it had support for the UMTS frequencies used by Optus here in Australia) as it would likely be the only way to get a phone with a decent hardware keyboard, decent specs under the hood and no locks preventing me from installing the software of my choosing.

Regarding the plan to be Fremantle compatible and to re-use blobs like dialer and SMS and stuff, I think it would be worth while for people who know what they are talking about to go through and evaluate all the blobs that remain closed (i.e. those that dont have either source available somewhere or 100% compatible replacements as part of e.g. CSSU) and for each blob figure out whether it should be:
1.Dropped completly because its not necessary going forward

2.Replaced by something existing or new that does the same job but which is not the same functionally (and is not compatible with) the Nokia blob.

3.Used as-is in the new software stack (e.g. the phone dialer app has been mentioned in this thread as one to reuse without replacing it)

4.Replaced by a re-implementation of the Nokia blob by reverse engineering it and creating a clone (as was done for e.g. osso-systemui-tklock in CSSU and others)

or 5.Replaced by something new but which is compatible with the existing Nokia blob i.e. we reverse engineer the interface between the Nokia blob and the rest of the system then we implement a new blob from scratch (or using existing code e.g. ofono) but which supports an identical interface (and is 100% compatible with the blobs that sit on top of it)

Also are people talking about Neo900 on e.g. IRC or other chat? If so which channel?

jonwil - mostly maemo and maemo-ssu on IRC. It is funn to read as I can learn a lot from it.