the open pandora 2 must use an amd apu!

[ferrule]

A bit more googling turns up this x86 Android phone running the previous generation of Atom (medfield):


http://www.anandtech.com/show/5770/lava-xolo-x900-review-the-first-intel-medfield-phone/6


It has a small battery, but actually does very well and beats some top end phones.

 

[craigix]

I could go in to optimised ASM CPU cores and such. Pretend I did.

 

[ghostpatch]

What? How is GIMP missing? GIMP was one of the very first things I booted on the first prototype Pandora.

Also XnView comes up on google as being available for arm as long ago as PocketPC?

arrgg damn it. I should have chacked it (played Borderlands2 the whole night and only pure thick coffee that pumps through my vains (and brain hehe) is keeping me from dying right now)


With Steam its one of three he he he

This GIMP?
http://repo.openpand...p-gimp-2.6-4644

you know, I even have it on my unit... -.-

The reason why there is a lack of interest in debating this is because we all already know the advantages of ARM and that this conversation has been done to death.

No one can be bothered to go though it yet again.

except the one guy who started this thread, nobody asks for the Pandora be x86. Its just that with Medfield and Clovertrail x86 gets more and of an alternative for ARM in the mobile devices market and I don't think thier improvements should be ignored.

Android will soon be the most popular OS in the world, almost every device, laptops included, will move to ARM.

who cares? that doesn't say anything


Edit: by the way, can someone explain the video with the two cars and its headline "Welcome to the future."? Guess I don't get it

 

[monstercameron]

the tesla is probably representative of arm, the telsa being all electric ismthe future of cars, so maybe thats his analogy

 

[ferrule]

I could go in to optimised ASM CPU cores and such. Pretend I did.

OK then, with ARM's optimized ASM CPU cores and such, how come they're not holding up against x86 like they used to? I'm ALWAYS interested in nuts and bolts, pipelines and instruction caches, but that's all academic in this case. There's benchmark data which suggests ARM might be in for a competition.

 

[ghostpatch]

the tesla is probably representative of arm, the telsa being all electric ismthe future of cars, so maybe thats his analogy

electric with battery and so? I hobe he didn' mean that. Sorry to say it, but electic cars are worse then normal cars from an ecological perspective. Maybe they can do something about the battery in the future but now?? LOOOK AT THAT smilie -> : | <- it is not looking happy!

 

[sswam]

I guess anyone who wishes to may attempt to make or fit an x86 board in his Pandora case.


I'd probably drop $500 on an "x86 Pandora" (Raptor Jesus eats another kitten).


Meanwhile, the actual OPT developers seem to prefer ARM to x86 for Pandora, and I do too.

 

[craigix]

the tesla is probably representative of arm, the telsa being all electric ismthe future of cars, so maybe thats his analogy

electric with battery and so? I hobe he didn' mean that. Sorry to say it, but electic cars are worse then normal cars from an ecological perspective. Maybe they can do something about the battery in the future but now?? LOOOK AT THAT smilie -> : | <- it is not looking happy!

Madness.

 

[ShleeDragon]

Shouldn't the modular nature of the Pandora 2 allow people to buy x86 SOCs


If they're willing to put it time to get it booting.. HAHAHA

 

[monstercameron]

what format will the system board be for this modular design? qseven, mobile itx?

 

[inkling]

the tesla is probably representative of arm, the telsa being all electric ismthe future of cars, so maybe thats his analogy

electric with battery and so? I hobe he didn' mean that. Sorry to say it, but electic cars are worse then normal cars from an ecological perspective. Maybe they can do something about the battery in the future but now?? LOOOK AT THAT smilie -> : | <- it is not looking happy!

yes and no. depends on where you get your energy. if it is primarily from coal then the math is bad. i think this focus may be off though. the cars are good if the utilize clean energy. we need to produce the clean energy any way since coal is bad.


there are steps in this direction. largely blocked by entrenched interests much like the electric car. hopefully it will move forward in a similiar manner.

 

[diligentcircle]

Android is not going to become the most popular OS in the world. There are so many reasons why this is a ludicrous claim, among them:


- Android sucks.


- It will have to compete with Windows for users that don't care about freedom and spyware.


- Users that do care about software freedom would never in their right mind use Android if they can avoid it.

 

[Maquis196]

Rubbish, im waiting for my Digital Alpha 21164 pandora 2. So who's going to suggest SPARC? anyone?


For me theres no point in having native ability to run windows to play games from turn of millenium if the SoC is gles only. If youre able to port a game that way, might as well recompile it for ARM anyway...

 

[Exophase]

OK then, with ARM's optimized ASM CPU cores and such, how come they're not holding up against x86 like they used to? I'm ALWAYS interested in nuts and bolts, pipelines and instruction caches, but that's all academic in this case. There's benchmark data which suggests ARM might be in for a competition.

Craig is talking about the quality of emulator code. Where have you seen a comparison between an Atom and an ARM core on emulation performance? Bear in mind, I don't agree with his post at all, I'm just looking at the current situation.


The benchmark data you're referring to is SPEC2k numbers provided by Intel. When Intel submits SPEC numbers on behalf of their competition it's a pretty safe bet that it's not a totally fair test. They've done this in the best with AMD where they totally crippled them. There are a lot of ways you can sabotage a SPEC score. Not just with compiler and compiler settings but also runtime libraries, OS, etc..


Intel probably compiled SPEC for Atom with ICC, and for ARM with GCC (possibly an old version!). ICC often does produce better code than x86 GCC, but on ARM other compilers (like ARM's compiler) have often produced better code than ARM GCC. armcc isn't used that much, but suffice it to say, neither is ICC - not that widely used on Windows and you will especially not see people using it to compile NDK fat binaries on Android.


On the flip side, Atom doesn't win on Phoronix benchmarks vs Cortex-A9 boards, where the same compiler is used and the compiler options are known. And on Android benchmarks, shitty as they are, Atom only wins on Sunspider (and has advantages on Javascript in general) which is an awful benchmark run on engines that probably output better quality code for x86, along with a lot of other relatively artificial reasons for it to be ahead.


Bottom line, never ever put faith in benchmarking numbers used for marketing purposes to make a processor look good. Especially if they don't provide any details on how they did the test much less allow for any kind of third party reproduction. Of course since SPEC2k is an ancient benchmark they have no avenue to even submit the results officially.

 

[craigix]

the tesla is probably representative of arm, the telsa being all electric ismthe future of cars, so maybe thats his analogy

electric with battery and so? I hobe he didn' mean that. Sorry to say it, but electic cars are worse then normal cars from an ecological perspective. Maybe they can do something about the battery in the future but now?? LOOOK AT THAT smilie -> : | <- it is not looking happy!

yes and no. depends on where you get your energy. if it is primarily from coal then the math is bad. i think this focus may be off though. the cars are good if the utilize clean energy. we need to produce the clean energy any way since coal is bad.


there are steps in this direction. largely blocked by entrenched interests much like the electric car. hopefully it will move forward in a similiar manner.

Even if you get your electricity from coal it's still far cleaner and efficient than an oil burning engine.


Teslas supercharger network is free to charge from and runs on sunlight.


The oil industry is currently doing everything they can to confuse, twist and discredit the electric car industry.


Fortunately Elon seems to be an exceptionally determined and driven individual. People have been saying he will fail for years.


Their first 'real' electric car just out dragged the 35 years of engineering in an M5 from 0-100mph (where it counts) - imagine what the next generation of electric cars will do.


Oh and it's cheaper to run than a Prius.

 

[McLovin]

Here in germany, 0-200 is where it counts


Anyway, I want to throw in one real reason to keep arm: compatibility. We have some great software for arm and with next generation socs, it will truly shine.


Also, I don't think x86 will do the magic things like automatic kernel compatibility, it would still be a pretty custom system.

 

[craigix]

Here in germany, 0-200 is where it counts

There is no particularly difficult challenge of beating an ICE car to 200MPH either, it would just be a case of different gearing.


The elegance of electric cars means ICE just can't win in the long term.


I am looking forward to BMWs electric car range, I like BMW.

 

[steved]

Here in germany, 0-200 is where it counts

There is no particularly difficult challenge of beating an ICE car to 200MPH either, it would just be a case of different gearing.


The elegance of electric cars means ICE just can't win in the long term.


I am looking forward to BMWs electric car range, I like BMW.

Can you afford one?

 

[ferrule]

OK then, with ARM's optimized ASM CPU cores and such, how come they're not holding up against x86 like they used to? I'm ALWAYS interested in nuts and bolts, pipelines and instruction caches, but that's all academic in this case. There's benchmark data which suggests ARM might be in for a competition.

Craig is talking about the quality of emulator code. Where have you seen a comparison between an Atom and an ARM core on emulation performance? Bear in mind, I don't agree with his post at all, I'm just looking at the current situation.


The benchmark data you're referring to is SPEC2k numbers provided by Intel. When Intel submits SPEC numbers on behalf of their competition it's a pretty safe bet that it's not a totally fair test. They've done this in the best with AMD where they totally crippled them. There are a lot of ways you can sabotage a SPEC score. Not just with compiler and compiler settings but also runtime libraries, OS, etc..


Intel probably compiled SPEC for Atom with ICC, and for ARM with GCC (possibly an old version!). ICC often does produce better code than x86 GCC, but on ARM other compilers (like ARM's compiler) have often produced better code than ARM GCC. armcc isn't used that much, but suffice it to say, neither is ICC - not that widely used on Windows and you will especially not see people using it to compile NDK fat binaries on Android.


On the flip side, Atom doesn't win on Phoronix benchmarks vs Cortex-A9 boards, where the same compiler is used and the compiler options are known. And on Android benchmarks, shitty as they are, Atom only wins on Sunspider (and has advantages on Javascript in general) which is an awful benchmark run on engines that probably output better quality code for x86, along with a lot of other relatively artificial reasons for it to be ahead.


Bottom line, never ever put faith in benchmarking numbers used for marketing purposes to make a processor look good. Especially if they don't provide any details on how they did the test much less allow for any kind of third party reproduction. Of course since SPEC2k is an ancient benchmark they have no avenue to even submit the results officially.

You're totally correct, benchmarking is almost always skewed for many reasons. Remember the good old days when Apple would contend their machines were orders of magnitude faster than x86? Good times...


The numbers I found interesting though were the power consumption comparison, only because it shows real progress for x86 in comparable real world scenarios.


BTW: Phoronix benched the old N450 and characterizes it as dual core while it's just a hyperthreaded single. Larabel definitely has his biases, though he is very open in methodology.. (not saying I'm a jedi either!)


For onlookers, here's the benchmark:


http://www.phoronix.com/scan.php?page=article&item=gentoo_arm_x32&num=1


And here's the Intel compare between the N450 and the interesting Z2760.


http://ark.intel.com/compare/70105,42503


This isn't going to be solved until someone can hold a Z2760 phone in their hand and see for themself what it's like. I've had many flavours of Atom machines cross my desk at work N330(Acer Revo), Z530(FitPC-2), and a few others which were more generic. They were totally usable machines, not fast, but certainly usable for the work we wanted to do which was at the time digital signage. I have the Revo here, maybe I should see what it's like running an emu or two? Any suggestions?

 

[Exophase]

I've said this several times now, but the power consumption tests you see for phones rarely more than barely stress the CPU. So Intel has done a good job showing that their SoC is all around low power enough to handle phone tasks, but it says little about what the CPU consumes when you're using it heavily. For a phone usage pattern this might be fine, but for a Pandora usage pattern it definitely is not. You need someone to play some demanding emulator (like N64, or better, Dreamcast) for several hours on ARM phones vs Atom phones then get back to you on what the results are like. Of course, the emulator has to be compiled for x86 to begin with, or the performance is going to be awful. I'd actually be surprised if any emulator that uses an ARM dynarec even works on their binary translation.


FWIW, the whole core thing re: Phoronix is just a matter of how many threads are reported in Linux, it's not a problem with the benchmarks. Of course, the benchmarks have their own problems, but at least they use known compilers/versions/flags and the whole thing is easy to reproduce.