How Much Will Arm Be A Handicap To The Pandora?

[mazza558]

It's nice to know a lot of things are just a recompile away. So it might be something like this:

- A few days after Pandora is released: Angstrom ARM packages/programs working/porting to Pandora
- A week or two after: People start porting some debian apps over
- A month or two later: First native 3D games/apps?

Will a lack of OpenGL ES support mean we won't see any native 3D games for a while?

 

[Chip]

We've been down this road before (discussion starts hereish). Both architectures have strengths and weaknesses. X86's weaknesses just happen to make it totally impractical for a Pandora-like device for at least the next few years.

There really isn't much to say that hasn't been said already.

 

[azmodean]

We will see, maybe Pandora 3 in 2020 will be x86(if x86 still exists, that is B) )

But then you'd lose backwards compatibility with all the great GP32, gp2x, WIZ, and Pandora apps....

 

[Squidge]

One of the specifications of the Pandora was "ability to use pre-existing code from gp32/gp2x". We have a library of code that has been rewritten in arm assembler for performance. If we changed to a non-arm processor, we would lose all these libraries. (I admit, there are a few for the x86, but most people still assume "just use a faster processor" and write in there language of choice).

 

[Svartalf]

- A month or two later: First native 3D games/apps?



This would depend on whether or not the ES support gets here this or next month... If it's here when they say it will be, it won't be a month later (I'm already working on moving some of the stuff I've got access to over to ES 1.1 right now. )

QUOTE

Will a lack of OpenGL ES support mean we won't see any native 3D games for a while?

Unless it's software rendering like Q2 has with it, unfortunately, yes.

 

[krooked]

i have to say not only did this post raise my spirits about the pandora, but definitely provided enough reason why you would make the move to arm processor.

maybe someone who is experienced with porting programs could create a guide to a simple pandora port so people can get the just of how you would rebuild a program for pandora, then they can start working on their own ideas of what they want to see. or just throw us a couple links where to find information that is already available and where the recompliler is.

 

[nikkopt]

agree with prometheus. arm is better for a handheld. why do you think the psp, ds and pocketpc's dont use x86. price/performance/efficiency pretty much covers it.

psp is MIPS based (i know you didn't said it wasn't, lol)
but yeah, we got the point

btw, offtopic: power architecture ftw h34r:

 

[VRAndy]

We have a whole host of software almost guaranteed to work out of the box on the Pandora (have a look at Debian's and Angstrom's ARM repos), but do you think the Pandora's ARM architecture will be the true Achilles heel of the device?

No.

You seem to be imagining a magical chip with all the benefits of an ARM chip, but the instruction set of an x86 chip. If anyone knew how to make a chip like that they would be doing it already.

 

[Exophase]

The same features that make lots of x86 cores fast (superscalar, register renaming, deep-pipelining, etc.) could also be applied to ARM cores if they wanted to set fire to every device they were put into.

But, Cortex-A8 is superscalar and pretty deeply pipelined (13 stages), and A9 has out of order execution and register renaming/speculative execution...

 

[quadomatic]

x86 would've been horrible for battery life

 

[AireTamStorm]

The same features that make lots of x86 cores fast (superscalar, register renaming, deep-pipelining, etc.) could also be applied to ARM cores if they wanted to set fire to every device they were put into.

But, Cortex-A8 is superscalar and pretty deeply pipelined (13 stages), and A9 has out of order execution and register renaming/speculative execution...
This is what makes the Cortex line so much more powerful than its predecessors. x86 = bad idea.

 

[darkblu]

The same features that make lots of x86 cores fast (superscalar, register renaming, deep-pipelining, etc.) could also be applied to ARM cores if they wanted to set fire to every device they were put into.

But, Cortex-A8 is superscalar and pretty deeply pipelined (13 stages), and A9 has out of order execution and register renaming/speculative execution...

..but not the heavy u-op translator frontend present in anything p6-derived i.e. anything up to the core architecture, and the latter continues the tradition for sure. that backward compatibility, in some aspects tracing back to the early '70s, does not come free. not by a long shot.

Moreover, it's the API for at least two consoles and the API for all the next gen handhelds (Of which Wiz and Pandora are the FIRST ones...).

what would those two consoles be? sorry, i may be missing something here.

 

[fanoush]

You seem to be imagining a magical chip with all the benefits of an ARM chip, but the instruction set of an x86 chip. If anyone knew how to make a chip like that they would be doing it already.

I guess it is not such big problem. It is just that there wasn't need/market for such chip. x86 is not used much in embedded space so nobody really misses it there. The real need starts only now with netbooks and Intel is already doing something with it. Poulsbo chipset is the beginning and after few iterations it may catch up with ARM. And BTW x86 instruction set itself is not a problem, the whole PC architecture (=many useless chips sucking power) is.

 

[mali]

You seem to be imagining a magical chip with all the benefits of an ARM chip, but the instruction set of an x86 chip. If anyone knew how to make a chip like that they would be doing it already.

I guess it is not such bit problem. It is just that there wasn't need/market for such chip. x86 is not used much in embedded space so nobody really misses it there. The real need starts only now with netbooks and Intel is already doing something with it. Poulsbo chipset is the beginning and after few iterations it may catch up with ARM. And BTW x86 instruction set itself is not a problem, the whole PC architecture (=many useless chips sucking power) is.


Agreed. The link to a discussion on the other board I posted earlier in this thread is worth to read IMO, because it is a cool-headed approach to ARM vs. x86 in handheld space. x86 will catch up fastly, because there is much money behind. I hope ARM will squeeze even more out of its chips in the future, competition is the user's best friend and I've been an ARM fan for years

 

[sold]

I think it's kinda funny that x86 processors keep getting faster and faster, but Windows keeps getting more and more bloated. What is the point of all these advances in clock and bus speeds and 45nm power saving toaster ovens just to keep up with Microsoft?

In the end, Microsoft and the inherent flaws in the architecture itself have become the x86s' Achilles heel. Hence the reason why people look for alternative hardware and operating systems for varying purposes. Using an ARM processor in an open source, portable device like the pandora is nothing but a strong point IMO.

 

[vputz]

I think it's kinda funny that x86 processors keep getting faster and faster, but Windows keeps getting more and more bloated. What is the point of all these advances in clock and bus speeds and 45nm power saving toaster ovens just to keep up with Microsoft?

Why would you buy a faster computer if the one you had was fast enough? It's Microsoft's responsibility to drive the hardware industry by overbloating their software so people have incentive to upgrade :lol: .

Still, I agree that x86 is primarily useful just for legacy code and the fact that more resources are dumped into it. Sure, things like Atom help with power consumption, but unless there's a REALLY COMPELLING reason to run an x86 core, there are better architectures. The advantage of an open-source-based console like the Pan is that we DO have that freedom, because we can just recompile everything. Only thing we don't have is something like WINE, but honestly that's pretty livable (I'm REALLY curious to see how DOSBOX goes for really old-school stuff, though!).

 

[Tor]

I saw some numbers a few months ago.. ARM CPUs are outselling x86 CPUs by a lot. Sounds incredible, but the ACM CPU sales is up to almost 3 billion a year now.
(Reference: http://www.arm.com/news/19720.html)
I couldn't find the numbers for x86 right now, but IIRC it was just above a billion last year.

The only "problem" with ARM is if you want to run some proprietary, Windows XP/Vista-only program. Which is for me really a niche problem. Doesn't bother me a millisecond.

 

[mali]

The only "problem" with ARM is if you want to run some proprietary, Windows XP/Vista-only program. Which is for me really a niche problem. Doesn't bother me a millisecond.

The brilliant side-effect of this is, that it pushes Linux and OS development in general because embedded Windows just isn't up to date. If the ARM netbooks will be successful that might change.

 

[fanoush]

The only "problem" with ARM is if you want to run some proprietary, Windows XP/Vista-only program. Which is for me really a niche problem. Doesn't bother me a millisecond.

Yeah like Fallout, Half-life, Planescape Torment and tons of other great games. Really niche problem on game machine ;-) Doesn't bother me a millisecond ;-)

 

[torpor]

Speaking as someone who works on 'next gen' Intel chips for a living (tolapai, atom) I can say this: ARM RULES!!!