SoC: Back and forth!


Yeah it does not hurt to shop around for sure. Its good to keep our options open.

If the qualcomm guy is legit, who knows.
 
I'd really like to use a company that supports Linux though.
I know it seems like I'm beating a dead horse, but Intel supports Linux.

http://software.intel.com/en-us/android/blogs/2013/12/11/introducing-4th-generation-intel-atom-processor-baytrail-to-android-developers

figure_5.png


Is there any reason not to use the Linux kernel support and libraries to run something other than Android on a Bay Trail SoC?

I'm ignorant on all of the details - but from a top level view - if they've already got the drivers and libraries done, doesn't that make the difference between Debian and Android more one of UI than OS?
 
And here I thought I was beating a dead horse with Samsung's SOCs, but this guy takes the cake. :p
 
Last edited by a moderator:
"We need a good SOC... with lot power usage... and full linux kernel support for the SOC... and linux driver support for the GPU and Xorg.. and wifi support and cheap and in low numbers"

willy-wonka-good-day-sir.png
 
Qualcomm has it's own share of problems though (compared to OMAP), like:

  • weaker NEON: http://www.openssl.org/~appro/Snapdragon-S4.html  (it's unclear if newer chips are still affected, but rumors say they still are).
  • vendor not caring about mainline kernels at all (only community support), meaning kernel updates will be hard
  • no vendor Linux support in general, TI at least provides something
  • no public chip documentation (NDA only), lacking documentation on Hexagon DSP (only usable from Android SDK?)
  • no 2D acceleration
Sure freedreno is great, but it's all in Rob's hands, and there are limits on things single person can do. For example, GLES3 that some people want so much is only available on vendor's driver (at least for now), which is most likely Android-only.
That's really some downsides of the Snapdragon.

If not too expensive I'd like to get both processors and compare them.

Snapdragon seems pretty powerful to me.

Maybe overall he's better than the OMAP?

I really would like to see some comparison.
 
I still favour TI. Wasn't their 2D acceleration open source? Do we really need a big GPU? Emulators (and most other things) probably profit more by using NEON, good Linux support and public documentation. Snapdragons are designed for smartphones. I'd like to remind you guys, Pyra is not a proprietary smartphone!

Everything about the OMAP5 is nice except no open source 3D GPU drivers. No big deal. The competitors open source drivers won't be useable for a long time, leaving plenty of time to catch up!

Let's find someone capable of reverse engineering PowerVR and feed him money, for growing a big and strong FOSS driver ; )
 
Last edited by a moderator:
I still favour TI. Wasn't their 2D acceleration open source? Do we really need a big GPU? Emulators (and most other things) probably profit more by using NEON, good Linux support and public documentation. Snapdragons are designed for smartphones. I'd like to reminde you guys, Pyra is not a proprietary smartphone!

Everything about the OMAP5 is nice except no open source 3D GPU drivers. No big deal. The competitors open source drivers won't be useable for a long time, leaving plenty of time to catch up!

Let's find someone capable of reverse engineering PowerVR and feed him money, for growing a big and strong FOSS driver ; )
As long as I didn't see how the Snapdragon is working in a Pyra prototype and that it's a lot better than the OMAP I am on your side.

Snapdragon sounds wow, awesome.

OMAP is OMAP. We know what we'll get and we know it's working.

Also I have a really bad feeling about not using OMAP.

Don't ask me why but I strongly feel that Nvidia or Qualcom will lead to a fail.,

Let's see how it develops.

But I wouldn't take the Snapdragon 800.

I think it's not worth the risk.

Only the 805 might be strong enough to compensate it's weakness.
 
I still favour TI. Wasn't their 2D acceleration open source? Do we really need a big GPU? Emulators (and most other things) probably profit more by using NEON, good Linux support and public documentation. Snapdragons are designed for smartphones. I'd like to remind you guys, Pyra is not a proprietary smartphone!


Everything about the OMAP5 is nice except no open source 3D GPU drivers. No big deal. The competitors open source drivers won't be useable for a long time, leaving plenty of time to catch up!


Let's find someone capable of reverse engineering PowerVR and feed him money, for growing a big and strong FOSS driver ; )
I'm confuse. Smartphones seem to be running emulators just fine.

Designed for smartphones. So what? OMAP5 is being used for cars.
 
Last edited by a moderator:
^...and tablets

OMAP is designed for the automotive sector and embeded systems. Not just cars. Basically everything that doesn't need GPS and 3/4G but processing power, availability in low quantity and long lifetime. You don't go buy a new industrial machine or a new car, just because your processor has died. I think availabillity and long lifetime is important for Pyra becuase its a sustainable device.
 
Last edited by a moderator:
Both Nvidia and Qualcomm are pushing their products into Automotive also.  You would have noticed Nvidia make a big noise about it in their last public outing.
 
Last edited by a moderator:
Both Nvidia and Qualcomm are pushing their products into Automotive also.  You would have noticed Nvidia make a big noise about it in their last public outing.
You can use the tech in anything,really, so I don't see his point. 
 
Last edited by a moderator:
The point is one of longevity, and also one of focus. For something smartphonecentric, your main focus is 3G/4G radios, navigation, and stuff like that, and you want quick-to-market (since the mobile market is a cutthroat business), but need not focus much on longtime support (since, well, three years from now noone wants to touch your SoC with a pole anyway). If you're going into automotive/embedded, you're targetting a market with more diverse general computing needs, and where the product is expected to have at least a ten-year lifespan. That rearranges your priorities when it comes to support.
 
^ It actually comes down to the contract , automotives require long term support and availibility and that's something that is assured at the point of sale in contract.  If Qualcomm or NVidia are selling to Automotive companies, they will no doubt be bound to the same required conditions .  

The length of the part availibilty is dictated by the Auto manufacturers country laws.  It can be different in different countries. I worked at Ford for 7 years, from memory in Australia the legal requirement was something like 7-10 years. Of course it gets complicated when a manufacturer exports a particular vehicle.
 
Last edited by a moderator:
Pandora only been out for 6 years and Qualcomm are using same tech (variations included) for other things. I don't think,at least, longevity is a problem.
 
Last edited by a moderator:
Back
Top