SoC Poll - What's most important for you?

[MaxNormal]

Arm linux ultra portable with full keyboard and excellent gaming controls.

That is a real niche that has significant advantages. Intel have never sucessfully competed with ARM on power usage - the heat has always been too high. Why on earth would ED take such a huge risk in going X86? X86 means poor battery life and more heat to dissipate.

What would we gain - some closed source games? What would we lose - devs, the current software, and the current cohesive community (most people running the same setup and able to run the same software)...

I think the OMAP is fine. If another ARM SOC can be used and provides what is needed then that is ok too.

An X86 UMPC is a different product.

 

[Natsu]

How about the fact that he'll get his ass sued by Nintendo, Sega, Sony, etc. if he starts pumping out a mainstream device that advertises emulation as a big feature? Whether he has any legal argument to make or not, they have the money to run him into the ground, and will use it if he starts making noise.

He can market it differently and Android has emulators, but you may be right since this is a smaller company.

 

[commander-beef]

Uh btw, sorry for de-railing the topic.

 

[_wb_]

Arm linux ultra portable with full keyboard and excellent gaming controls.


That is a real niche that has significant advantages. Intel have never sucessfully competed with ARM on power usage - the heat has always been too high. Why on earth would ED take such a huge risk in going X86? X86 means poor battery life and more heat to dissipate.


What would we gain - some closed source games? What would we lose - devs, the current software, and the current cohesive community (most people running the same setup and able to run the same software)...

It is no longer necessarily true that x86 means worse power efficiency. It is clear that Bay Trail is now in ARM territory in terms of power efficiency, whether or not they are actually better is not yet clear because there haven't been any really independent and sufficiently complete benchmarks yet.

Also I don't think we would lose all current devs and software when switching architecture -- for most of us, it's just a quick recompile. As long as we stay with GNU/Linux, the architecture is largely unimportant. Notable exceptions are people dealing with low-level stuff like notaz, and people who are doing assembler optimizations like Exophase. Even among them, some may be fine with switching to x86.

I think the attraction of the Pandora is that it is a portable GNU/Linux computer with excellent battery life and controls. The Pyra, whether it will be ARM or x86, will still be a portable GNU/Linux computer with excellent battery life and controls. I don't think the choice of SoC will have a big impact on the current community (notaz being a notable exception).

That said, I still think it's best to stay with ARM, at least for now.

 

[Rockthesmurf]

2.0 GHz, 6 hours battery life

Like people said, while having the Pandora at 10 hours? I would at least want
something close to that.

Well it is like I already said, the details are what matter. Would people be willing to go from 10 hours to 6 hours if the device was 100 times faster, you'll find that 99% + of people will say yes. If it goes from 10 hours to 6 hours with 1% performance improvement, you'll find that 99% + of people will say no. Everyone obviously loves having decent battery life, but everyone probably have different views of what 'decent' is, for me 6 hours is actually fantastic. A device with a spare battery will last long enough to get me from one charge point to another in all occasions.

 

[Granitehead]

I think the x86/arm question was a bad one. Judging by the discussions going on in that thread, the people who voted didn't have the technical knowledge to actually judge it by technical aspects, mixed it up with the questions 1 and 2, don't know about some non-technical aspects and have rather absurd opinions about some of the other non-technical aspects.

I don't know which architecture is better, but I'm sure that the results of that poll's question 3 don't mean anything.

 

[comradekingu]

All technical things aside, dont underestimate one option has more votes, suggestion people will buy it based on the fact, regardless of what what it technically entails, you cant take peoples opinion away from them, technical backing or not.

 

[Grench]

Arm linux ultra portable with full keyboard and excellent gaming controls.


That is a real niche that has significant advantages. Intel have never sucessfully competed with ARM on power usage - the heat has always been too high. Why on earth would ED take such a huge risk in going X86? X86 means poor battery life and more heat to dissipate.


What would we gain - some closed source games? What would we lose - devs, the current software, and the current cohesive community (most people running the same setup and able to run the same software)...

It is no longer necessarily true that x86 means worse power efficiency. It is clear that Bay Trail is now in ARM territory in terms of power efficiency, whether or not they are actually better is not yet clear because there haven't been any really independent and sufficiently complete benchmarks yet.

Also I don't think we would lose all current devs and software when switching architecture -- for most of us, it's just a quick recompile. As long as we stay with GNU/Linux, the architecture is largely unimportant. Notable exceptions are people dealing with low-level stuff like notaz, and people who are doing assembler optimizations like Exophase. Even among them, some may be fine with switching to x86.

I think the attraction of the Pandora is that it is a portable GNU/Linux computer with excellent battery life and controls. The Pyra, whether it will be ARM or x86, will still be a portable GNU/Linux computer with excellent battery life and controls. I don't think the choice of SoC will have a big impact on the current community (notaz being a notable exception).

That said, I still think it's best to stay with ARM, at least for now.

I agree with everything you said except the last sentence.

'Staying with ARM' still requires purchasing the SoC directly from the manufacturer.  Whether that is a TI OMAP or a Snapdragon or other ARM.  We're subject to the manufacturer's whims and changes in direction.  They could decide on a whim to stop selling them to the project.  "The guy" that the project is dealing with might leave and either not be replaced or not want to deal with the project anymore.  The whole ARM marketplace is very trendy and windy - the next greatest thing pops in and all else is instantly considered 'junk'.  Documentation is sparse.  Undocumented binary blobs for proprietary video are the norm.  NDAs just to talk - you get the idea.  The ARM production world controls which devices they allow their product to be in.

I know that the same has been implied about Intel, but appears to be simple fear and rumor mongering.  The fact is that Intel publishes the data sheets without requiring an NDA to view them.  The SoC is available directly through them or through re-sellers.  They don't 'control' which devices the SoCs get put into.  The SoC and video drivers are all already in the current Linux kernel.

It's surprising, and a bad mark for ARM, that Intel's z3770 SoC is more open, more Linux friendly, publicly documented and easier to purchase.  It may even be cheaper at $37 each in thousand SoC trays - a price publicly listed on the Intel product page.  We don't know the price for ARM SoCs as ARM producers require an NDA before purchase so the actual price of an ARM SoC is hidden.  To me, this sounds like the Intel SoC may actually be less expensive, easier acquire, easier to build hardware for - and easier to build software for.

I can't say 100% that the project must use one or the other.  I can say that the project deserves whichever SoC is best - among a wide array of measurements.  That may very well be an ARM SoC.  However, it should be worked through based on facts.

Has anyone from the project given Intel a quick call to find out whether they can get a development board?

http://www.intel.com/content/www/us/en/intelligent-systems/evaluation-platforms/evaluation-platforms.html

Contact link is on the right side of the page.  If the z3770 is the objective, then you're likely wanting a 'tablet' dev board.

I think the x86/arm question was a bad one. Judging by the discussions going on in that thread, the people who voted didn't have the technical knowledge to actually judge it by technical aspects, mixed it up with the questions 1 and 2, don't know about some non-technical aspects and have rather absurd opinions about some of the other non-technical aspects.

I don't know which architecture is better, but I'm sure that the results of that poll's question 3 don't mean anything.

I agree.  However, the way that the poll was worded was highly generic ARM vs X86 without specifying which actual SoCs are options and what the real benefits & faults of each SoC are.  It's clear that a large number of the responses have no clue of the specs and capabilities of the actual x86 SoC being considered

All technical things aside, dont underestimate one option has more votes, suggestion people will buy it based on the fact, regardless of what what it technically entails, you cant take peoples opinion away from them, technical backing or not.

So, by your logic, the Earth should remain flat and we should all buy a new buggy whip.

 

[SeeThruHead]

It's very viable. But it seems the Ed us the only person who really wants to use an soc this way. Everybody else is creating tablets. The only interesting thing I would say is the shield but it doesn't have Linux and isn't x86 so it's mostly uninteresting as well. Arm was really good when it was the only option for low power computing. But it isn't any more.

Maybe thats because, as i said, its a niche platform... The sheild runs linux. Insofar as its more comparable than the failed windows running x86 clamshells of yester year. linux|gnu and android is on every other gaming handheld. Maybe with the omission of the razer thing, that got canceled → X86, windows, mass market, sounds familiar? Im going to say you are partly right, because even Mips (gcw-zero) is more of a goto solution in the real world. If you want x86 devices, go to the last vendors that are alive, they have enough trouble selling laptops at the price it commands. No small company survives in x86, x86 has _nothing_ in small form factor, much less a viable future plan, and if it did, going there is a huge risk, for which there is no tangible reward seeing as dragonbox is a small linux|gnu vendor.

Furthermore, game-centric handheld clamshells isnt generic because it isnt.

Everyone else _wants_ tablets, afaik as evident by this poll, ARM is the arch people want for their niche device, and these are the demographic that makes up the userbase for said niche segment of the market. The diamentrally opposite stance to how that is viable, isnt viable.

TL;DR: wintel is subsidizing the x86 mobile dream for now, but its not happening. If its anything that is more mainstream now as opposed to when pandora came to be, its the platform its built on. Steady growth with less risk makes sense.

You're basing your argument against x86 on mobile based on years of failures doesn't actually have any meaning. Because there hasn't been an x86 chip with low power usage available in a consumer device until a few months ago. So there is no history of failure for this type of soc. 2014 will have tons of windows and android tablets based on intels x86 chips. Amd is going to bring mullins which should have even more gpu performance and hopefully similar battery life. I see zero reason to buy any arm device in 2014 as it's merely limiting your options. I see ARM as a solution to the power problem of yesteryear, a solution that traded software compatibility. Now a device designer no longer needs to make that sacrifice. 

 

[Exophase]

Z3770 isn't easier to purchase than OMAP or even Allwinner SoCs. Both are sold through distributors and not just directly from the manufacturer. And I don't see a BayTrail eval board listed on Intel's site so there's nothing to try to get..

 

[second exodous]

I voted ARM did some reading on Bay Tail, got excited, then watched some video and think ARM again. I think it is wise to stick with ARM for now but the low power x86 stuff is pretty impressive and maybe the Pandora3 will be developed in a time when it will make perfect sense to run x86.


I really don't think keeping backward compatibility with the Pandora should be a point to stick with ARM, or really moving to one handheld to the next, it could hold the next gen back. When using open source you never start from square one anyway, you stand on the shoulders of those who coded before you.

 

[hollowshell18]

I don't normally post on forums, but given the state of both this thread and the "SOC: Back and Forth" thread I feel I need to.

I bought my Pandora for many reasons.  I didn't want some disposable piece of garbage from sony or nintendo, nor did I want a tablet.  I wanted a linux laptop with excelent multi-media and gaiming capabilitys with a keyboard.  Its obvious the Pandora was designed with these functions in mind (it has a game controller built in after all), but given the tone of these threads I'm beginning to wonder if I'm in the minority here.

So I have to ask, what is the purpose of the Pyra?

Allow me to play Devil's Advocate for a moment.

If we were to switch to an x86 chip, we know what we'll gain software wise.  Going from the Pandora's ability to run Daggerfall at slideshow framerates to potentually running Oblivion via Wine at playable speeds(along with nearly every other x86 wine game from the mid 2000's and earlier) is a compelling argument.

But I'm yet to see anyone say exactly what we'll gain software wise from a set of A15s.  I know we'll be gaining betterN64, PSP, and DS emulation, and likely gaining Dreamcast emulation as well. 
 

The general consensus seems to be that Sixth Generation emulation is almost certinally out no matter the CPU architecture, both due to hardware being not-quite there yet, and the emulators themselves needing massive amounts of work. 

Then of course we have the Android games, but has anyone taken a look at the Play store lately?  We've got the PS2 Grand Theft Auto trilogy and a handful of great indie games that have been ported.  Not exactly a lot of titles if your tastes don't include casual games like Fruit Ninja.

Am I missing something here?  Surely this isn't all we'll be gaining with an ARM chip, is it?

Again, I don't yet have a preference.  I'm merely playing Devil's Advocate before I vote one way or the other.  After all, no matter what decision is made, at least one way or another I'll be playing GTA: San Andreas on my Pyra.

 

[Grench]

Z3770 isn't easier to purchase than OMAP or even Allwinner SoCs. Both are sold through distributors and not just directly from the manufacturer. And I don't see a BayTrail eval board listed on Intel's site so there's nothing to try to get..

Educate me a bit.  What differentiates this:

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

from an 'evaluation platform' like this:

http://www.intel.com/content/www/us/en/intelligent-systems/evaluation-platforms/norco-bis-6630-atom-n2800-development-kit.html

???

 

[slaeshjag]

1) It needs all the GPIO that the Pyra needs, to be accessable.

2) The board needs to be using the right SoC, or at least one that can be drop-in replaced without modifications to the board.

 

[God Ginrai]

'Staying with ARM' still requires purchasing the SoC directly from the manufacturer.  Whether that is a TI OMAP or a Snapdragon or other ARM.  We're subject to the manufacturer's whims and changes in direction.  They could decide on a whim to stop selling them to the project.  "The guy" that the project is dealing with might leave and either not be replaced or not want to deal with the project anymore.  The whole ARM marketplace is very trendy and windy - the next greatest thing pops in and all else is instantly considered 'junk'.  Documentation is sparse.  Undocumented binary blobs for proprietary video are the norm.  NDAs just to talk - you get the idea.  The ARM production world controls which devices they allow their product to be in.

I know that the same has been implied about Intel, but appears to be simple fear and rumor mongering.  The fact is that Intel publishes the data sheets without requiring an NDA to view them.  The SoC is available directly through them or through re-sellers.  They don't 'control' which devices the SoCs get put into.  The SoC and video drivers are all already in the current Linux kernel.

It's surprising, and a bad mark for ARM, that Intel's z3770 SoC is more open, more Linux friendly, publicly documented and easier to purchase.  It may even be cheaper at $37 each in thousand SoC trays - a price publicly listed on the Intel product page.  We don't know the price for ARM SoCs as ARM producers require an NDA before purchase so the actual price of an ARM SoC is hidden.  To me, this sounds like the Intel SoC may actually be less expensive, easier acquire, easier to build hardware for - and easier to build software for.

I'm curious, have you confirmed that all 3 of the proposed ARM SoCs have these NDAs you claim? I know the Exynos had NDAs, and seem to remember hearing that the Snapdragon did, however I haven't heard any such thing about the OMAP or the A80 chips.

-God Ginrai

 

[EvilDragon]

Yes, the OMAP5 has an NDA, not for the datasheet, but for additional information like the schematics of the devboard (which makes moving the SoC with all needed components over to the main PCB a lot easier).

It could very well be Intel has NDAs for stuff like that as well.

However, once you signed the NDA and got the okay from a company to buy he SoC, there's no "That Guy" who can decide that you can't buy it anymore.

A contract is a contract, and it would be horrible for a company to spend hundred-thousands of dollars into designing a product just to not be able to buy the SoC later on...

Nope, doesn't work like that. We got a contract here, so we can buy the OMAP5 for the next about 10 years.

 

[vcoleiro1]

Out of interest ED, what does a OMAP5 cost in low volumes?

 

[WizardStan]

Am I missing something here?  Surely this isn't all we'll be gaining with an ARM chip, is it?

It isn't about what we're gaining one way or the other, it's mostly risk management.ARM is more or less known, it's performance, it's power consumption and heat dissipation, the features that can be used on it, it's all familiar to the community. x86 is new and new is scary. We have reports from intel and spec sheets that we can look up and "everyone" has an x86 desktop of course, but in SoC form it isn't quite the same: until EvilDragon has an x86 dev board and gta04 can do some proper performance tests we don't really know anything about it.

Plus there's a fear that x86 will be too popular; yes, I said too popular. EvilDragon runs a small company with a lot of debts right now, he wants to produce a successor that'll hopefully get him out of those debts so that he can start growing but right now he's legitimately worried that support costs of x86 will be too high. I respect him this opinion and can appreciate the concern. ARM is safe, ARM is familiar, ARM won't get a bunch of Windows kids jumping on it without thinking and then getting upset when it doesn't run Crysis.

I'm sure if he didn't have the Pandora debts he'd be more happy to consider an x86 but realistically I won't hold it against him for not wanting to take that risk.

edit: and production channels, very important. ED's already got a contract for the OMAP5 and we know how to put OMAPs on boards. It may seem like such a small thing to put x86 on a board instead but every small thing adds up to bigger things.

 

[Exophase]

Educate me a bit.  What differentiates this:

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

from an 'evaluation platform' like this:

http://www.intel.com/content/www/us/en/intelligent-systems/evaluation-platforms/norco-bis-6630-atom-n2800-development-kit.html

???

That evaluation platform is Atom N2800 which is CedarTrail, not BayTrail-T. It's a totally different, previous generation Saltwell-based SoC. I know it's confusing because the other product you linked is called Celeron N2820 which is current generation, but it's not BayTrial-T but BayTrail-M, which is a different SoC from Z3770. Neither of these linked products would be great for ED's purposes anyway, you want a real eval board that you can easily connect things to, not a little PC.

 

[SNESFAN]

While these non tangible arguments are infinitely readable (infinite as looping back on its self) they seem to never end and the same people seem to be saying the same things over and over and over. All of which seems to stem from personal preference. Get some Dev boards, post real actual measured numbers then let people vote. This spiritual debate for the uninformed can't be the way this device is going to be made. For things like the OS distro or software its fine because it can always be scrapped and switched later in favor of something better. But this is hardware, set in stone for years probably and hardware numbers don't lie, preferences might however.