Pandora 2, A Year Or So From Now?

[Kangal]

I'm still interested with a Pandora II with a multi-core Cortex-A9 (two minimum)

It would be dual-core. A quad-core will eat out of the battery life without giving any real performance boost. Why? Most programs are designed to be handled (not handheld) by one or two cores, except for encryptions which wont be something used on a "low power" computer, we'll use our desktop PC for those. And you can see evidence of this; sales of Quad-Cores slowing while Dual-Core sales still increase.

 

[WizardStan]

It would be dual-core. A quad-core will eat out of the battery life without giving any real performance boost. Why? Most programs are designed to be handled (not handheld) by one or two cores, except for encryptions which wont be something used on a "low power" computer, we'll use our desktop PC for those. And you can see evidence of this; sales of Quad-Cores slowing while Dual-Core sales still increase.

A multi-core system would allow you to just turn off cores you aren't using, thus saving power. I've yet to encounter a desktop that couldn't do this, but they all require a reboot unfortunately. An ARM based system can probably arbitrarily turn a core on or off as it's needed. If the OS notices it's doing a few single threaded things, shut off three cores to conserve power. If it suddenly sees a lot of multi-threaded activity, crank up all 4. It can already automatically adjust the clock speed to conserve power (in theory, I'm not sure if the image the Pandora comes with has this built into it yet) so this would just be one more step in that direction. So your power vs battery life concern is irrelevant. Dropping a quadcore in a handheld should not cause any loss in battery unless they're being used, and if they're being used, then it's an automatically a performance boost.

 

[ldesnogu]

That diagram is completely wrong.

 

[juniorm33]

That diagram is completely wrong.

Wow talk about a blanket statement.
The title isn't that bad
The axis are ok.
Names of chips doens't seem wrong.
:lol:

 

[calc84maniac]

That diagram is completely wrong.

Wow talk about a blanket statement.
The title isn't that bad
The axis are ok.
Names of chips doens't seem wrong.
:lol:

Maybe he's talking about the claim "higher is better"

 

[laurencevde]

If we disable all power-management on all chips, put them to full use (including DSP, GPU etc.), and make the scales logarithmic, then that graph might make some sense...

 

[mali]

That diagram must be taken with a whole block of salt,

ROFL
Signature alert!

 

[fischju2000]

Some semi-relevant articles:
http://www.engadget.com/2010/02/15/texas-instruments-introduces-arm-based-omap-4-soc-blaze-develop/
http://www.engadget.com/2010/02/15/st-ericssons-u8500-brings-dual-core-1-2ghz-arm-cortex-a9-to-the/

 

[Jeffery Mewtamer]

It would be dual-core. A quad-core will eat out of the battery life without giving any real performance boost. Why? Most programs are designed to be handled (not handheld) by one or two cores.

While each individual program would require an overhaul to benefit from multiple cores, could an overall performance boost during heavy multitasking be achieved by the OS delegating CPU intensive software to a separate core?

 

[Kangal]

It would be dual-core. A quad-core will eat out of the battery life without giving any real performance boost. Why? Most programs are designed to be handled (not handheld) by one or two cores.

While each individual program would require an overhaul to benefit from multiple cores, could an overall performance boost during heavy multitasking be achieved by the OS delegating CPU intensive software to a separate core?

Worth reading:
http://apcmag.com/is_dualcore_better_than_quadcore.htm
http://www.codinghorror.com/blog/2007/09/choosing-dual-or-quad-core.html

Perhaps a Tri-Core might give the performance boost you were looking for without sacrificing battery life like a quad core. Although a tri-core would mean; a 3x solo-cores setup unlike a 2x dual-cores (quad), so making a good OS will be much more difficult.
Then again that's a HUGE "if we tried, completed the project and benchmarked it" scenario.
As far as anyone knows (for personal use) dual core is the sweet spot (without large networking/video encoding)

 

[hlidegp2x]

Worth reading:
http://apcmag.com/is...an_quadcore.htm
http://www.codinghor...-quad-core.html

Perhaps a Tri-Core might give the performance boost you were looking for without sacrificing battery life like a quad core. Although a tri-core would mean; a 3x solo-cores setup unlike a 2x dual-cores (quad), so making a good OS will be much more difficult.
Then again that's a HUGE "if we tried, completed the project and benchmarked it" scenario.
As far as anyone knows (for personal use) dual core is the sweet spot (without large networking/video encoding)

1) Those figures are for INTEL/AMD cores (energy hungry). Besides, dual core and four core don't have the same frequency so their benchmarks are not valid for what you try to state.
2) Most games are not coded to handle all cores. Most are probably considering only two permanent threads to run in concurrency. In fact, most games are not optimized for a variable number of cores. Again not valid argument.

For a PC gamer, there is no point to buy a quad-core with a lower frequency to replace his/her duo-core as long as games are still developped to be optimal in duo-core. Note that i5 and i7 has turbo mode to address this issue (less core used, higher frequency). And AMD plans to have an equivalent turbo mode too.

From developper point, I'm most interested to have a quad-core than a duo-core.

Anyway, I just want a multi-core for Pandora II (which means at least two cores, not the solo core) so you can have several threads running in concurrency. If not for game, it will be useful for some emulators.

 

[Awakening]

IIRC at least Windows (have no experience of other OSs) have for a very long time split up multiple single threaded programs to run on multiple cores/CPUs, improving the speed when running a bunch of stuff in the background.

 

[lulzfish]

IIRC at least Windows (have no experience of other OSs) have for a very long time split up multiple single threaded programs to run on multiple cores/CPUs, improving the speed when running a bunch of stuff in the background.

Yes, so if you have a video encoder dominating one CPU core, your web browser, music player, etc. can share the other core. I think all OSes do this if they support multiple cores.

 

[urjaman]

video encoder dominating one CPU core

I lol'd. I must be weird...

 

[sullrosh]

it needs a chip like this

WiLink 7

 

[daehremmah]

it needs a chip like this

WiLink 7

+1
I figured I wasn't the only one to see that.
Plus, I can only imagine some of the fun you could have with a small FM transmitter...

 

[fischju2000]

At the MWC Samsung showed it's new Super AMOLED tech outdoors, the biggest problem with OLEDs, next to the whole slow death of their organic parts, is sunlight viewing which seems to be overcome: http://www.engadget.com/2010/02/16/samsung-wave-shows-whats-super-about-its-amoled-display-by-goin/ - If it were available, this tech (which is 3.3" and 800x480 so no resolution problems on a Pandora sized screen) would be an amazing choice for a successor device. Many comparison pictures here: http://www.mobile-review.com/review/samsung-wave-s8500-en.shtml#3 Though I doubt the tech would be available or feasible for OP in a 'year or so' AMOLED is the way to go.

 

[Dunny]

Most emulators don't scale that well beyond a few cores, because some major portion of the runtime has to be ran serially. Actually, most emulators don't even give that much benefit past one core, although this has been changing a bit.

In my own considered opinion, you cannot, under windows emulate a Sinclair Spectrum and give a smooth, fluid experience with less than three cores.

D.

 

[calc84maniac]

Hmm, so would DS emulation (or dare I say virtualization?) be easier if you have separate cores to emulate ARM7 and ARM9?