Able to upgrade not just the speed but manufacturer of CPU?

[EvilDragon]

We can upgrade to any CPU we like in theory, as people said, but getting one is hard. Or better: Getting a GOOD one.

Getting Allwinner SoCs would be possible, and thanks to the community, Linux works somewhat... but Allwinner violates the GPL, would it be a good idea supporting this for an open Linux device?

Rockchip should be possible as well, but how is the Linux support there...? And the documentation to build a board?

My best hope is Qualcomm. They were interested before and they seem a bit more open than the rest.

Maybe someone else will produce something? The Freescale was awesome, but they stopped after the Cortex A9 sadly.

Still, the OMAP5 is pretty good - and seeing how my Pandora often feels faster than many CortexA9 based smartphones, the Pyra should be on par with many tasks compared to current smartphones.

PPSSPP doesn't run much better on the GPD XD than on the Pandora, as I found out.
And most mobile games work fine on an Cortex A15 as well.

It's mostly some REALLY high CPU usage apps (I.e. DraStic with high-resolution rendering) where you'll see a difference.

We tried a LibreOffice Presentation from askarus and sometimes it felt faster than on his desktop PC.

 

[ptitSeb]

What about the Exynos SoC. It's used on some ODroids. The linux support is pretty good and it should be available as not too big quantities (or are the ODroids XU3/XU4 huge sells?).

 

[Caine]

I would not go for Allwinner. The GPL violation cannot be disregarded in a product as the Pyra.
A fair bit of the user-base are GNU/Linux enthusiasts and supporting a GPL violator can be a deal breaker for them.

Thus, more dragons!

 

[Farox]

I think an Intel Atom chip would be feseable...i don't know if you have already talked to them...but from a discussion (in italian forum) i had with an engineer that work in germany (for Intel) it seems they are happy to sell it to us...i don't know their MOQ, but i think ...when you announce an upgrade like this...instantly could sell 5000 units.

 

[Faust]

I think an Intel Atom chip would be feseable...i don't know if you have already talked to them...but from a discussion (in italian forum) i had with an engineer that work in germany (for Intel) it seems they are happy to sell it to us...i don't know their MOQ, but i think ...when you announce an upgrade like this...instantly could sell 5000 units.

From my point of view it worth a try. Maybe we should wait and see how performs the GPD WIN first?

 

[EvilDragon]

What about the Exynos SoC. It's used on some ODroids. The linux support is pretty good and it should be available as not too big quantities (or are the ODroids XU3/XU4 huge sells?).

Samsung were the first I talked with, and while the german branch was enthusiastic and wanted to help me get it, they couldn't convince the korean HQ for that.

AFAIK the ODroids are the official devboards for the Exynos (similar like the EVM we have for the OMAP5), so evaluation boards.

That's why quantity doesn't matter.

From my point of view it worth a try. Maybe we should wait and see how performs the GPD WIN first?

Intels don't have any MOQ or NDA, AFAIK, it's certainly an idea I want to follow.

 

[vcoleiro1]

Hasn't the boat sailed on a Intel SoC replacement board though. I mean, if the community develops a lot of ARM based SW for the Pyra, that would make it hard to choose a Intel SoC replacement board down the line knowing the community development to that date is all ARM. Also, who then would get Linux running on that Intel SoC , etc. I don't know , it seems to me , there was a fork in the road, and the ARM direction was chosen with no coming back.

 

[Djoga'Ro]

I see a new project - an open SoC.

 

[LoopStar]

I see a new project - an open SoC.

This'll open so many possibilities for endless debates & polls! But also the chance to build the only SoC with blast processing!

 

[Galaxis]

Still, the OMAP5 is pretty good - and seeing how my Pandora often feels faster than many CortexA9 based smartphones, the Pyra should be on par with many tasks compared to current smartphones.

Performance-wise, the OMAP5 should be perfectly fine. Negative comments about Cortex-A15 have mostly been about their peak power consumption when under heavy load. Pity it will be hard to make any good use of the M4 cores, too - later ARM SoCs with the big.LITTLE combination of high- and low-power cores that support the same instruction set are more convenient.

On the up side, I assume the Pyra can be run on a powersave governor most of the time, with enough reserves for CPU-intensive tasks, when they are needed. I really hope that some people with Linux kernel development knowhow get a Pyra though, and work on optimizing things for the platform. (I think the Pandora, despite all it's mishaps, was extremely lucky in that regard - probably would have ended up a lot less useful without Notaz' efforts.)

 

[Urben]

I see a new project - an open SoC.

This and an open baseband processor!

Step one: Take multiple million $ in your right hand and...

Oh

 

[Caine]

I see a new project - an open SoC.

You have no idea how costly that is.

 

[Djoga'Ro]

You have no idea how costly that is.

Maybe I have an idea that it'd be an immense effort and was just not that serious.

 

[comradekingu]

NXP bought (all of?) Freescale, and they aren't just developing silently anymore.

LS1046A

Is the most beastly thing they got.

LS2085A

LS2080A

LS1012A

Edit: No GPUs on these, but they have x4 PCIe.

As for totally open SoC, see OpenRisc right now that means FPGA, and makes it a different prospect alltogether.

 

[EvilDragon]

Hasn't the boat sailed on a Intel SoC replacement board though. I mean, if the community develops a lot of ARM based SW for the Pyra, that would make it hard to choose a Intel SoC replacement board down the line knowing the community development to that date is all ARM. Also, who then would get Linux running on that Intel SoC , etc. I don't know , it seems to me , there was a fork in the road, and the ARM direction was chosen with no coming back.

That wouldn't have been any different before, as we had the Pandora with ARM software already, right?

We're with Linux here, any software can be compiled for any architecture, so whether it's x86 or ARM is not a problem. Except for that a software uses specific features (e.g. NEON) or assembler language, that would be hard to port.

Otherwise, everything could be compiled using x86 as well.

Who would get Linux running? Well, software developers.
Same as on the ARM system, or do you think Linux and the drivers simply popped out of thin air?

 

[Djoga'Ro]

Who would get Linux running? Well, software developers.
Same as on the ARM system, or do you think Linux and the drivers simply popped out of thin air?

Thought Seti received it. And then here HW was designed to match that stuff. ?

 

[vcoleiro1]

That wouldn't have been any different before, as we had the Pandora with ARM software already, right?

We're with Linux here, any software can be compiled for any architecture, so whether it's x86 or ARM is not a problem. Except for that a software uses specific features (e.g. NEON) or assembler language, that would be hard to port.

Otherwise, everything could be compiled using x86 as well.

Who would get Linux running? Well, software developers.
Same as on the ARM system, or do you think Linux and the drivers simply popped out of thin air?

The question as you yourself once said in defense of not going Intel, is who would do it from this community. Your argument back then was that no one you knew from the current community would do it - are we meant to materialise a x86 dev to do it from thin air - your words. Even if they did, the community would then be split. How,much support would the x86 side get (probably end up like Android support). Frankly I don't see it working. For the exact reasons mentioned which an ED of past days mentioned him self.

Yes it would have been different if you went Intel now, as time goes by, more ARM based SW will be developed - so a future jump is worse than if you had done it now. Not that it wouldn't be a big jump now, but there is that.

Unless someone was found(paid) to get Linux running(drivers and all) & support on that optional x86 daughter board. I doubt people would get it . You'd want that assurity before laying out the cash to buy one.
And there's the problem, not many people are going to buy a $150 - $250 x86 daughter board without the assurity of x86 Linux having been already developed and supported for it. And it would be hard to convince a dev to do that if they didn't know how many if any people would buy that daughter board . A catch 22 situation. Unless of course you payed a dev to it and gave people assurity that way. That's the only way I see it working.

 

[Faust]

Actually if we get a Intel Cpu I'm almost sure that new devs would come to port and develop stuff for the new one, some more Linux has already lots of SW and devs.

 

[EvilDragon]

The question as you yourself once said in defense of not going Intel, is who would do it from this community. Your argument back then was that no one you knew from the current community would do it - are we meant to materialise a x86 dev to do it from thin air - your words. Even if they did, the community would then be split. How,much support would the x86 side get (probably end up like Android support). Frankly I don't see it working. For the exact reasons mentioned which an ED of past days mentioned him self.

I think you are confusing a few things here, it sounds like I was misunderstood by you a bit.
Let me summarize my view on it:

1. We'd need a BIOS developer, that's the complicated thing. But that can be solved, we also needed to create or our board files for the ARM SoC.
2. Except for the drivers, there's not really anything that NEEDs to be ported. You could simply run all Linux stuff. However, and that's probably where you misunderstood me back then: I'm not a fan of that. Because it can be very frustrating playing games that are not OPTIMIZED for that system (controls, resolution, speed, etc.). That's the reason I prefered (and still prefer) an ARM-Based system: Stuff needs to be ported, but you can be sure it will run as it should when it's released. On x86, most stuff isn't optimized as it's not needed for Desktop PCs. People would simply use the unoptimized software though, as I don't think many devs would see a reason to work on those games, as they basically work, just aren't optimized.
3. I don't think that would split the community. My guess is that the x86 users would mostly use it similar to a normal PC. There's not much to talk about, as there are no custom releases for it. Similar to the GPD Win or GPD XD. What's there to talk about? It's Windows / Android, there are no custom releases or developments, so not that much to talk about. It's different with more uncommon devices like the Pandora. There's plenty to discuss about

Yes it would have been different if you went Intel now, as time goes by, more ARM based SW will be developed - so a future jump is worse than if you had done it now. Not that it wouldn't be a big jump now, but there is that.

Honestly? I don't see any problem here.
The x86 users would simply get the system and mostly use it like a portable PC. Not much discussions and developments would happen, as it isn't anything more than a normal portable PC. If they even register on the boards, they would mostly be silent, I guess.
Well, at the beginning there would be happening a bit more, while powersaving, etc. is being improved... but once that has happened, it's a normal PC.

The ARM users would still continue to release games and ports and discuss about every possible improvement.

There is not really ARM-Based software. How many files in the Pandoa repo use ARM-specific stuff? Maybe 5 - 10 out of 1500. If even that many.
The rest could be compiled both for ARM and x86 without any additional changes.

Unless someone was found(paid) to get Linux running(drivers and all) & support on that optional x86 daughter board. I doubt people would get it

And what's different from ARM here?
Nikolaus is being paid by me, and he did most of the low-level hardware and driver stuff until now as well.

If I decided to produce an x86 board, that would be part of the job as well.

Did you think we only out the OMAP5 onto a CPU board and didn't care about anything else?

And there's the problem, not many people are going to buy a $150 - $250 x86 daughter board without the assurity of x86 Linux having been already developed and supported for it. And it would be hard to convince a dev to do that if they didn't know how many if any people would buy that daughter board . A catch 22 situation. Unless of course you payed a dev to it and gave people assurity that way. That's the only way I see it working.

That's not any different from the ARM. No one would buy the Pyra if we wouldn't at least deliver a basic working system at the beginning, would there?

 

[vcoleiro1]

If I decided to produce an x86 board, that would be part of the job as well.

I'm glad you would do that ie pay some one to port x86 linux - particularly get drivers working. If that's the case, it could work.
The impression I had, was that you would leave it to the community to sort out, I'm glad you corrected that , and cleared it up.