oxym | # 461 | 2013-09-16, 14:08 | Report

I know that memory is the bottleneck here.
But regarding Power consumption issue, this page :
http://forums.arm.com/index.php?/top...r-consumption/
States that OMAP4 takes less Power than OMAP3.
So I don't see any disadvantages right here.

---

Last edited by oxym; 2013-09-16 at 14:26. Reason: hate virtual keyboard on non N900 decices

misiak | # 462 | 2013-09-16, 14:32 | Report

The page you linked just link to 2 further pages, one stating that pure Cortex-A8 core under heavy load takes <300mW while pure A9 core takes <250mW per cpu, in power-optimized mode, while at the same time having two cores (so like.. two cpus? so 250mW would change to 500mW for whole chip, thus more than A8? so even that statement would be wrong? or am I misinterpreting something?). Other poster, however wrote that in fact one should compare whole SoCs and clearly noone over there did that publicly, even the post you're probably relating to. And there is no OMAP4-OMAP3 comparison over there either, I still don't know where you got that info from...

biketool | # 463 | 2013-09-16, 15:08 | Report

What software is the board design being done on? Can we have access to the design files so we can propose hardware hacks, file in hand, for actual production?
While the cement is quickly drying on the Neo900 mk-I mostly based on the proven GTA04 we can already be working on the Neo901.

joerg_rw | # 464 | 2013-09-16, 17:45 | Report

OMAP4 -> not binary compatible to OMAP3 anymore, ergo main design goal of using same app repos on N900 and Neo900 failed.

Project files -> see post#1

---

Last edited by joerg_rw; 2013-09-16 at 21:22.

zoner | # 465 | 2013-09-17, 02:23 | Report

Really nice work by all involved!

I'm throwing money at my screen now. When does this roll? Do we still need 200 people with 700 eur?

Has a timeline (after commitments are in place) been discussed?

joerg_rw | # 466 | 2013-09-17, 03:32 | Report

Thanks! :-)
Yes, in the end we will need 200 people placing a real preorder so we can start building a batch of 200 devices.
The timeline until this starts is:

• in 2 or 3 weeks (we hope) we will have evaluated the complete concept against a real N900 and we will know what we plan to build, I.E. what we need to change in GTA04 and which components we need to and actually can source.
• Then we know which product properties aka specs you want and we can actually offer, and we will have a more clear-cut BOM and expense evaluation so we can tell something reliable about the final pricetag.
  Only then we will start some kickstarter or other means to raise funds for developing prototypes. It's not yet decided what options we will offer, like special editions and/or premium memberships with direct channel to the lab, and/or an actual preorder already - with an ahead down payment of maybe 20%. Anyway the goal for such fundraiser/preorder-early will be ~10k .. 15k EUR that's needed for the prototypes. There's a certain very minimal (0/00) but non-zero risk that by circumstances we have no control about (components might show fatal defects, known as Silicon Errors, etc) the project still may fail during that phase, so this funding must be considered some kind of venture capital.
• Then - given we can raise the funds - we design and build prototype board GTA04-NeoN_A1, cross fingers for components and all will behave (see SiErr), and test it. Fix any bugs (sure there will be some, there always are), test again,
• build better protoboard GTA04-NeoN_A2. After a few cycles in this process which will take a few months
• we will reach quality that's ready for MP.
• Then the latest we will show off first prototypes and open a regular preordering form (or ask for payment of the remaining 80%) and hope for those minimum 200 needed preorders to rush in during a few days.
• let's hope 200 preorders come in, we start "Mass" Production, and 4..6 weeks later the devices should be ready to ship (disclaimer: sourcing of N900 case parts from some obscure asian channels may take longer than 6 weeks, depending on shipping and whatnot, so while the GTA04-NeoN boards may already be ready to ship, this might not apply for the complete devices)

cheers
jOERG
(disclaimer: all time spans and values are my own rough estimations and not backed up by those who are the experts of managing this process. [edit] meanwhile "confirmed" by Nikolaus)

---

Last edited by joerg_rw; 2013-10-03 at 00:51.

c12mech | # 467 | 2013-09-17, 10:22 | Report

I put my vote in for 700€. If this version outperforms the current n900 and has more features it will be exactly what I am looking for. I know a lot of work goes into these things so I'm in no rush. All I care is that it works and I can use it as a portable pen testing ool like my current one. Keep at it and thanks for all the effort.

ffha | # 468 | 2013-09-17, 12:33 | Report

How so?
They both support the same architectures.
I can run the same precompiled Debian binaries on both a Cortex-A8 and a Cortex-A9 without any side-effects.
I can run the same x86 binaries on both an Intel Sandy Bridge and an AMD K8.

Did OMAP4 drop the DaVinci technology or something?

joerg_rw | # 469 | 2013-09-17, 13:43 | Report

It been what I've been told. I think it might be related to GFX core and whatnot else. Sure the *basic* CPU opcode should be portable, but the SoC differs too much - aiui.
If we could guarantee that fremantle incl all blobs can work on a OMAP4, all the better! Nikolaus mentioned that we might even have the needed additional 0.8mm headroom in N900 case that were missing in GTA02 case, so we *could* go OMAP4. Then OTOH I know about SiErr in all OMAP4 that are declared "WONTFIX in OMAP4" by TI. So maybe we do not even want to go for OMAP4.

Ariadeno | # 470 | 2013-09-17, 14:08 | Report

TI never keep their promises. It's too bad as they make really nice SoC's. I have been waiting since the release of BeagleBone Black for the SGX530 driver for kernel 3.8.

If you ask them about progress, "It's done when it's done" is the answer you get. Relying on TI is like asking a deaf man for directions.

So it might be a good choice not to have OMAP4. But we'll see.