SoC: Back and forth!


Some people from intel vectorizing scene are posting on source forge vectorizing mailinglist. They are talking about how to auto detect parallel loops and build data structures which support vectorization on the fly. This would be very nice to have for the Pandora, if people would use intel compilers ( I think gcc is starting to support this too). It can give you a very big speed boost, since it  uses SSE4 and multiple cores without much manual work. This isn't as fast as hand coded neon code, but for the average Joe coder, it can give you lots of speed without much knowledge. This could maybe mean less power consumption but I am not sure, how 1 SSE instruction performs against several normal FPU instructions.
 
Last edited by a moderator:
I can't imagine an x86 chip can reach 10h battery time like an ARM can, at least not under load and especialy not with a Quadcore Chip. Because Battery is a important factor for mobile Devices, I guess it's clear then that it can only be a ARM design.

Another factor that should not be forgotten is the price. Which one will be cheaper to get, OMAP or Snapdragon?
 
I can't imagine an x86 chip can reach 10h battery time like an ARM can, at least not under load and especialy not with a Quadcore Chip. Because Battery is a important factor for mobile Devices, I guess it's clear then that it can only be a ARM design.


Another factor that should not be forgotten is the price. Which one will be cheaper to get, OMAP or Snapdragon?
Under load the current ARMs also won't be able to run 10 hours.

The current Intel and ARM should be similar in battery life when running under full load, that's not the problem.

However, from my experience, ARM stuff is more optimized as ARM had been a lot slower for the last years than any current Intel x86 CPU.

So people needed to optimize, whereas on Intel desktop, it didn't matter, since the CPU power was there and it wasn't a mobile device.

That's the concern.
 
I can't imagine an x86 chip can reach 10h battery time like an ARM can, at least not under load and especialy not with a Quadcore Chip. 
Why not if it's overengineered ? (with the finest state of art lithography, transistor layout and other tricks) 


Well it's true that it may cost an arm... ;)
 
One of the principles the founding fathers of the OpenPandora was that the machine must run Linux and only Linux.  Therefore, Pyra, the child of OpenPandora must adhere to the principles set forth by the founding fathers.  That is, it must run Linux and Linux only.  Therefore, any SoCs that can even remotely enable the possibility to run windows (natively) are not allowed.  Hence, ARM SoCs and not Intel :p

-I Duck, I Duck, I Duck-

bush-shoe.gif
 
Last edited by a moderator:
I'm definitly agreeing with the reasoning for using OMAP 5 instead of trying to go x86. x86 would definitly make it more unique then say smart phones, but there's a reason that x86 isn't used in mobile devices, the battery life is horrible. Though as a guy with the raspberrypi and such I totally love the idea of a powerfull portable ARM opensource friendly mobile device.
 
^

http://www.dell.com/us/p/dell-venue-8-pro/pd

10 hour battery life, x86

EDIT:

disclaimer, I'm for ARM processor in the pyra
Careful - tablets are huge, so they usually have batteries included that is WAY bigger than the ones in a Pandora/Pyra or Smartphone.

Another issue we need to be careful about are the display drivers.

We need to connect a MIPI display to the SoC (which is not easy, Nikolaus is working on that for a few days with the OMAP5 already), and while there is work happening, we don't know how complete it is right now.
 
I sincerely doubt that x86 CISC overhead is free. Part of the bloat is hidden by the overengineering of these chips. That's why independant reviews of the intel chips are needed :( .
Internally even those CISC CPUs are pretty RISC-like these days anyway. I fully agree that you should not trust the manufacturers and their numbers, but that is true for all of them. I still think that the Anandtech comparison is pretty objective. Also, when considering to use any of these SoC, one would probably want to set up some development environment and try out things without solely relying on others.

To Fusion_Power and all the other repeating those old stories of x86 being the power-hungry beasts: I has been mentioned here and elsewhere over and over again. Just check the facts before repeating that nonesense. It is always good verify manufacturer's claims and not believe everything, but it does not help at all to repeat false myths over and over again.
 
Careful - tablets are huge, so they usually have batteries included that is WAY bigger than the ones in a Pandora/Pyra or Smartphone.
Actually when you check the website, Dell says that a 18 Wh battery has been used there to achieve the 9.9 hours runtime. The Pandora battery has about 15 Wh, so it is not that far away. With the smaller Pyra display one could probably have the same runtime.
 
Last edited by a moderator:
^

http://www.dell.com/us/p/dell-venue-8-pro/pd

10 hour battery life, x86

EDIT:

disclaimer, I'm for ARM processor in the pyra
Careful - tablets are huge, so they usually have batteries included that is WAY bigger than the ones in a Pandora/Pyra or Smartphone.

Another issue we need to be careful about are the display drivers.

We need to connect a MIPI display to the SoC (which is not easy, Nikolaus is working on that for a few days with the OMAP5 already), and while there is work happening, we don't know how complete it is right now.
this says it's a 4860mah battery  (18 watt hour @ 3.7v)

http://www.gsmarc.com/dell/venue-8-pro/

I know not the physical size of it though

EDIT: 

https://www.youtube.com/embed/7pi5baams1U?feature=oembed

physical size looks comparable to the pandora's actually
 
Last edited by a moderator:
However, from my experience, ARM stuff is more optimized as ARM had been a lot slower for the last years than any current Intel x86 CPU.

So people needed to optimize, whereas on Intel desktop, it didn't matter, since the CPU power was there and it wasn't a mobile device.
This is what really impressed me about the GP2X community when I first joined. All of the optimizations that people made to make things run in limited cpu or RAM. Seeing OpenBOR made working on the GP32 with only 8MB of RAM available, or seeing emulators working to fit within 32MB of RAM before the MMU hack. Not to mention all of the ASM cores that were written and implemented in emulators from the GP2X to the Pandora.

If we go x86, not only will a vast amount of the software be unoptimized, but I fear that we may lose that drive that made this community feel so impressive.

-God Ginrai
 
^ Do we even need a community if we go x86? I mean, everyone knows how to use an x86 system, and all the software will already be available. Maybe not in a shape that runs well, but who cares?
 
I want PCSX-ReArmed, DraStic, PPSSPP, Reicast, and a whole bunch of other ARM-optimized emulators and games to work day 1 when I open my PYRA.  The last thing I want to do is wait for these to be re-optimized for x86 in order to use on PYRA (if there's even x86 people willing to do the work - I DOUBT THIS VERY MUCH).  That would be suck.

With ARM, I know we have notaz, Exophase, ptitSeb, linuxbochs, and other members of this community who are always willing to help.
 
Last edited by a moderator:
Back
Top