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

[Kev2442]

Great, I like Nvidia Tegra K1.

Some time ago, we had a discussion concerning the X1 to be sold directly as a board the size of a credit card. Is that actually a thing and can we adapt it to the mainboard ? That would be awesome performance-wise, a way to ridicule the Win.

Otherwise, I think we should keep ARM. There's no real need for x86 except for some closed source software that would either run fine on Exagear+WINE or just need something else and just crash.
Plus, I think ARM might be the future of desktop computers as a competitor to the PC standard. RISC is a very efficient concept, and as mobile devices progressively turn into full-fledged computers, there sure will be a crossroads.

EDIT : came across this, might be of interest to someone : http://nexvision.fr/solutions/reference-design-products/cam-master-plus/

EDIT 2 : forgot to say, may I add that we should really make LPDDR a high priority on the next SoC ? current DDR3 is a major battery leaking point.

 

[EvilDragon]

EDIT 2 : forgot to say, may I add that we should really make LPDDR a high priority on the next SoC ? current DDR3 is a major battery leaking point.

That depends on the SoC. The OMAP5430 (the mobile version of our SoC) was supposed to be using LPDDR, only the OMAP5432 (the industrial version) uses DDR.

Sadly, TI decided to never release the 5430 (pretty late as well, we were already designing the PCBs at that time)

 

[Ocean112]

I think it would be a wiser idea to go with an AMD chip instead of an Intel one. That is, if they release something along the same line. They have several products that are close, but we'll have to wait for the 14nm node on AMD before we know whether they'll have a competitive product that would fit the design specs. I'd rather have an AMD gpu if I had the choice. Another decision that will have to be made is BIOS or UEFI. I would assume that BIOS would probably be easier and more compatible, but then I'm not a firmware programmer myself so I don't really know.

 

[EvilDragon]

Is AMD planning a mobile SoC?
I thought they are concentrating on servers?

AFAIK, they even got ARM CPUs, but only for servers...

 

[ekianjo]

AMD is planning 14 nm mobile SOC (x86) combos down the line. Probably within this year.

 

[TrashyMG]

AMD is planning 14 nm mobile SOC (x86) combos down the line. Probably within this year.

Yeah I'm curious how available these would be and how willing AMD would be to deal with such a small quantities and how much native Linux support they would be willing to give. Although right now AMD needs all the helps it gets.

 

[AVahne]

I second the idea of using an AMD chip (if they're indeed making x86 mobile SoCs), especially if ED can get ahold of them before any other gaming UMPC manufacturer.
The red (assuming this is a choice) UMPC with a name of fire using a red team chip just seems to fit.

 

[Ocean112]

As far as SoC's go, they have the GX-217GI (Which is a bit too large I think) and the GX-218GL which is bga-769 (they also have a 6W TDP version of this chip). What I'd like to see is a future variant of the LX series with HBM, We'll see. They don't have a "mobile" SoC per se but they do have something really close with their embedded chips. In any case, I wouldn't waste my time with any of the current offerings from AMD. I'd wait about 6 to 8 months and then check again once zen is released. They should have some info about new future offerings by then.

Edit https://en.wikipedia.org/wiki/Socket_FT3 It looks like the next release will be around Mar 2017 if they release an umpc chip at the same time.

 

[Urben]

Isnt it utopistic to look for upcoming SOCs as we actually want very open source SOCs? I bet the upcoming ones are not likely to be very open at release. Is there a chance for more powerful and released SOCs opposed to the OMAP5 to get more open anytime soon?

 

[Ocean112]

Open source SoC? What? Do you mean the drivers for the SoC? Like the graphics driver? Drivers will be available for linux, that's a given. The xorg driver should already support it with minor tweaks if it sticks with a GCN architecture. There's always fglrx though. You'd be dealing with x86 Debian at that point, and there's a ton of support out there. What in the world is very open source anyway? GPLv2? Free Software? Are you a Stallmanite? Have you converted to the church of emacs; do you worship Saint IGNUcius?
If that's the case, then we should drop Debian and use Trisquel.

Besides, I wasn't even talking about replacing the OMAP5. The cpu board is fine as it is. I'm just looking towards the future, getting my hopes up...

 

[Urben]

The Stallman church is funny, but I cannot deny they have their point...

 

[Ocean112]

They do, and all I will say about it is that I agree with most of their points, if not all. People need the freedom to live in a cage if they wish, too.
I'll leave it there and talk no more about it, it can be a very contentious subject.

 

[Letalis Sonus]

There's always fglrx though.

No, there is not. Did you really miss the big happening of the past few months? It's amdgpu all the way now.

 

[Ocean112]

Yeah? I'll have to look into it. Thanks for the heads up.

 

[second exodous]

The OMAP doesn't get much love here atm.
Maybe you should all just wait and see how it's doing in actual usage.

I'm just asked because maybe future Ti chips won't be that good and maybe we will want to upgrade to another company down the road. I wasn't suggesting this one wasn't good, just wanting to know how future proof it is.

 

[God Ginrai]

Nvidia outrightly refused to work with Evildragon... So I doubt they would have a change of heart any time soon.

IIRC, they merely told him that he must work through one of their approved partners to design the circuit board. For the Pyra, that may not have worked, but if all ED is trying to do after that is design a CPU upgrade board, then why wouldn't he consider this avenue?

-God Ginrai

 

[rygD]

IIRC, they merely told him that he must work through one of their approved partners to design the circuit board. For the Pyra, that may not have worked, but if all ED is trying to do after that is design a CPU upgrade board, then why wouldn't he consider this avenue?

-God Ginrai

Maybe he will. Still, this is his baby, so what gives a component manufacturer the right to dictate who will design it? I know that would leave a bad taste in my mouth, and it would be even worse if I was trying to design something to support freedom and openness.

 

[Grench]

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?

It's cool that you are giving this consideration. The Pyra hardware platform and form factor would be cool with any SoC.

Yes, that work would have to be done - however much of it may be already there. I'm not sure how much of what these guys did got submitted upstream, but they would have had to have done similar work to produce this:
http://liliputing.com/2016/04/mele-pcg02u-is-a-70-ubuntu-pc-stick.html
I have one of those on order, shipping to take 7-28 days, so not sure when I'll get it. I would loan it to you, but shipping it to you and back to me would be 80% of it's $70 cost.

It is Bay Trail, but Cherry Trail being a die shrink of Bay Trail should have very similar drivers.

Ubuntu, based on Debian, those drivers may not work out of the box, but they should bear a strong resemblance to what would need to be created right? Leverage the open part of the open platforms - no need to invent the wheel?

So, I suspect that the hill to climb may not be as substantial as it is/was starting with a relatively new/obscure ARM chip with limited Linux support. Intel already pushed the SoC with it's graphics into the kernel after all - which was probably a big chunk of it.

Counter-intuitive as it may sound, Intel is more open than the vast majority of the ARM producers.

If you produce an Intel SoC for the Pyra in the next year or two, I'd buy one - and maybe a whole new Pyra to wrap around it. 'Both' is better than 'either'.

 

[Faust]

It's cool that you are giving this consideration. The Pyra hardware platform and form factor would be cool with any SoC.

Yes, that work would have to be done - however much of it may be already there. I'm not sure how much of what these guys did got submitted upstream, but they would have had to have done similar work to produce this:
http://liliputing.com/2016/04/mele-pcg02u-is-a-70-ubuntu-pc-stick.html
I have one of those on order, shipping to take 7-28 days, so not sure when I'll get it. I would loan it to you, but shipping it to you and back to me would be 80% of it's $70 cost.

It is Bay Trail, but Cherry Trail being a die shrink of Bay Trail should have very similar drivers.

Ubuntu, based on Debian, those drivers may not work out of the box, but they should bear a strong resemblance to what would need to be created right? Leverage the open part of the open platforms - no need to invent the wheel?

So, I suspect that the hill to climb may not be as substantial as it is/was starting with a relatively new/obscure ARM chip with limited Linux support. Intel already pushed the SoC with it's graphics into the kernel after all - which was probably a big chunk of it.

Counter-intuitive as it may sound, Intel is more open than the vast majority of the ARM producers.

If you produce an Intel SoC for the Pyra in the next year or two, I'd buy one - and maybe a whole new Pyra to wrap around it. 'Both' is better than 'either'.

I'm quite agree, I will do the same, and we are not the only one.

 

[txus]

I'm quite agree, I will do the same, and we are not the only one.

No, you are not. Count me in on day 1 an x86 board starts selling/preordeing/whatever for the Pyra. In fact, the OMAP5 is my only gripe with the current Pyra, with its lack of support for all the wonderful dedicated hardware it has, since even being the most open ARM manufacturer, it's still far from the hardware support you get from Intel or AMD. Having said that, I'm still eagerly awaiting a 4GB EU model, OMAP and all.