Am Wed, 5 Feb 2014 14:54:00 +0100 hat "Dr. H. Nikolaus Schaller" hns@goldelico.com geschrieben:
Hi,
Well, the question is, if we can get dev boards,
I think we already considered to use the DevBoard which is the "DragonBoard" (what a nice coincidence to DragonBox, DragonFly :): http://mydragonboard.org/
Yep, those are readily available.
We haven't started to integrate the OMAP5 onto our PCB yet, and the Snapdragon 800 seems impressive... and has OpenSource GPU drivers available.
It is funny with the Freedreno that the main driving person is an TI employee: http://www.phoronix.com/scan.php?page=article&item=qualcomm_freedreno&am...
Heh - neat ;)
Maybe this is something for a statement for the discussion about an "IT" petition :)
Not sure if they will care...
But if I look here: http://freedreno.github.io/ http://bloggingthemonkey.blogspot.de/ indicates that it is still work in progress.
Yes, but we could still use the closed driver as well. People would still be happy that they could get an open driver if they wanted to.
I have no details but like with the modems those SoCs are to some extent exchangeable. Because everyone wants the same features from them:
- CPU
- Display
- RAM
- USB3.0 + other ports
- some I2C, SPI, GPIOs etc.
Yep, that shouldn't be a deal with our current design.
One thing has to be considered: the Snapdragon is designed as a complete SoC including UMTS/LTE and WLAN/BT. So we have to take care to get a (cheaper?) version without these functions.
The Snapdragon 800 can be bought with and without modem.
So I would still stay with external 3G module + external WLAN/BT module, even if Qualcomm will try to convince us to user their fully integrated chips.
Yes, definitely.
notaz, does it have all the stuff we need for software? (Linux drivers should be available, but does it have Neon and 2D acceleration we can use?)
For the MIPI display we could most likely use the Sony Xperia Z driver I have found yesterday :)
Oh, okay ;)
Just to make sure: We don't know yet if we can get it - but the question is: IF we can get it, would it make sense to switch?
Well, it is not answerable without having it seen and evaluated. The easier to answer question with the current level of information is:
I can answer some of them.
- CPU+GPU performance is at least as good (we really need to
benchmark it carefully)
CPU: Yes, definitely. The Snapdragon is very similar to an A15, but we've got 4 cores with over 2GHz here.
Also, it's using lp-ddr3-memory, so a bit faster as well.
GPU - I don't know, but it should be fast enough for most things anyways.
There's OpenCL and OpenGL ES3.0 available (only OpenGL ES2.0 for the OMAP).
- it is not more expensive (assuming we can get it in the low
quantities we want)
Yeah, well, TI is one of the most expensive ones out there, so that should work.
- we can handle the chip in PCB dimensions and production (BGA pitch)
PCB dimension shouldn't be an issue, as it is smaller than the OMAP5432. Die size (edge seal) 11.14 mm x 10.62 mm
BGA Pitch is something I need to find out.
- we have enough information to "copy" the Dragonboard circuits
Yes, that will be checked as well.
If any of these is worse than with the OMAP5, I would say the "open GPU driver" is not outweighting the downsides.
Qualcomm processors are pretty good, that's why they're being used in most High-End-Smartphones.
But we could also ask TI for help to open the SGX drivers... E.g. by permitting Rob Clark to work on it. If they ever want to see an OMAP5 in a consumer device :)
They didn't really care much about the Pandora, sadly.