SoC: Back and forth!


If you would have followed the thread a little further, you would have noticed, that someone there actually has sucessfully used the Intel driver with the Z3770 GPU on X.
 
Modesetting mostly, yes.

What makes you think that? In recent years, Intel has actually been pretty quick with releasing driver updates and supporting new hardware on Linux. Also, the GPU in that Z3770 is from the Ivy Bridge generation, so supporting that chip really should not be that difficult for them. Also, as I said above, it looks like it actually works okay.
The fact that Intel driver development is happening in linux kernel and mesa and these projects have their respective release cycles. Also there is no linux or android devices that use z3xxx so it's not very high priority work.

Also everything that seems easy to a layman isn't always easy.

About snapdragon...

There was a problem with ppsspp performance on these compared to 'standard' A15 implementations (see http://forums.ppsspp.org/showthread.php?tid=7069). Don't know if it was fixed or if it can be fixed with reasonable manpower.

Also sorry for a guest post, I didn't see that I wasn't logged in
 
The fact that Intel driver development is happening in linux kernel and mesa and these projects have their respective release cycles.
Both the Linux kernel as well as Mesa releases happen more often than those hardware releases.

Also there is no linux or android devices that use z3xxx so it's not very high priority work.
This actually is pretty absurd. Especially the Android part. Intel wants companies to actually use their chips (yes, also on Android not only Windows), so they have to supply working drivers.

Btw, the mode setting issue people reported in that forum thread does not even have to be a driver issue, but could as well be caused by the hardware configuration (e.g. the display not correctly reporting its resolution - it is not like this has not happened before).

Also everything that seems easy to a layman isn't always easy.
I never said the opposite was true.
 
Last edited by a moderator:
The Snapdragon 805 is announced in Q2/2014, so that would throw us back about 6 - 8 months... not feasable, I think.
 Design for Snapdragon 800 and let's do the upgrade thing again in a few jears.
If it does come to that point, we would probably be looking at the Snapdragon 1000. :lol:
 That's what I was hoping for :) .
Why are you so fixed on the Snapdragon?

Unless Qualcomm can provide Linux drivers and a 2D accelerator (which seems unlikely, since they're concentrating on Android), we'll most probably not use it.
 
I wonder if Intel would subsidize their SOC and other components for the Pyra. Maybe it's a long shot, but doesn't hurt to ask , might be a factor to weigh up in the SOC comparison should it be offered.


Actually from what I've read , it seems to be that the SOCs are being sold for less than they should be. I thought it was some kind of subsidizing offer at first.


See here: http://liliputing.com/2014/01/bay-trail-tablets-cheap-intels-footing-bill.html


..
 
Last edited by a moderator:
Three two things I like about Snapdragon compared to OMAP 5432:

  • supports LPDDR3 (14.9 GB/s), versus only LPDDR2 (8.5 GB/s) on OMAP (or DDR3, but not low power) -- so it can have lower power and/or faster RAM
  • Adreno GPU has a better chance of getting a FOSS driver, and it is probably faster too (~129 GFLOPS vs ~38 GFLOPS, if that is a relevant metric)
  • it can have built-in wifi/bluetooth/3G/GPS, which probably saves PCB space and guarantees that everything works well together, and simplifies design. It also means that wifi etc are using 28nm nodes, which means they're probably lower power than external chips which tend to use older processes.  [Edit: this is not really an option, for several reasons, see posts below]
Depending on how much Linux and PCB design support Qualcomm gives, it is superior in my opinion. The 2D acceleration and faster NEON of the OMAP are minor advantages, but mostly irrelevant given that the single-core CPU performance and the GPU are faster on the Snapdragon. Long-term availability is perhaps an issue, but successors are already planned and will hopefully be mostly compatible. It's not that big of a problem if we are forced to do a SoC upgrade in a few years, as long as it does not require a PCB redesign.

So if Qualcomm wants to sell it in the low quantities we need, with good Linux support and PCB design guides etc, and if notaz is OK with it (it will be more work for him than OMAP5), then I think Snapdragon is the best choice at the moment. If there is a choice, that is.
 
Last edited by a moderator:
Three things I like about Snapdragon compared to OMAP 5432:

  • supports LPDDR3 (14.9 GB/s), versus only LPDDR2 (8.5 GB/s) on OMAP (or DDR3, but not low power) -- so it can have lower power and/or faster RAM
Actually the OMAP 5 version with LPDDR2 has been discontinued/never actually produced, as far as I know.
 
Three things I like about Snapdragon compared to OMAP 5432:

  • it can have built-in wifi/bluetooth/3G/GPS, which probably saves PCB space and guarantees that everything works well together, and simplifies design. It also means that wifi etc are using 28nm nodes, which means they're probably lower power than external chips which tend to use older processes.
Nope, that's the most horrible thing to do.

3G will be a binary blob, and you have NO idea what data it transfers and what backdoors there are.
 
Three things I like about Snapdragon compared to OMAP 5432:

  • it can have built-in wifi/bluetooth/3G/GPS, which probably saves PCB space and guarantees that everything works well together, and simplifies design. It also means that wifi etc are using 28nm nodes, which means they're probably lower power than external chips which tend to use older processes.
We can't use these functions even if they are on the chip for several reasons:

1. we don't have the design tools to develop and test the low-level RF circuitry needed for such a design (this is NOT on SoC)

2. we then must get full certification for the whole software/hardware

3. we get more firmware blobs where we don't know how long we can get updates and how compatible they are to future kernels

If we use precertified modules connected through USB/SDIO on one end and an antenna on the other end, this saves a really, really big amount of money.
 
Last edited by a moderator:
Three things I like about Snapdragon compared to OMAP 5432:

  • supports LPDDR3 (14.9 GB/s), versus only LPDDR2 (8.5 GB/s) on OMAP (or DDR3, but not low power) -- so it can have lower power and/or faster RAM
Actually the OMAP 5 version with LPDDR2 has been discontinued/never actually produced, as far as I know.
Yes the OMAP5430 was planned to allow for Package on Package (PoP) memory to save space. But it appears that it did and will not go into production (but who knows about TI's decisions?).
 
Three things I like about Snapdragon compared to OMAP 5432:

  • supports LPDDR3 (14.9 GB/s), versus only LPDDR2 (8.5 GB/s) on OMAP (or DDR3, but not low power) -- so it can have lower power and/or faster RAM
Actually the OMAP 5 version with LPDDR2 has been discontinued/never actually produced, as far as I know.
The way I understand things, OMAP 5430 supported LPDDR2, while OMAP 5432 supports both LPDDR2 and (non-LP) DDR3.

Three things I like about Snapdragon compared to OMAP 5432:

  • it can have built-in wifi/bluetooth/3G/GPS, which probably saves PCB space and guarantees that everything works well together, and simplifies design. It also means that wifi etc are using 28nm nodes, which means they're probably lower power than external chips which tend to use older processes.
Nope, that's the most horrible thing to do.

3G will be a binary blob, and you have NO idea what data it transfers and what backdoors there are.
Will it not be a binary blob driver anyway?

You could always use a Snapdragon without 3G, you still get the wifi advantage. Or use two models, one with and one without 3G. The privacy paranoids can get one without 3G. If there are backdoors.

Three things I like about Snapdragon compared to OMAP 5432:

  • it can have built-in wifi/bluetooth/3G/GPS, which probably saves PCB space and guarantees that everything works well together, and simplifies design. It also means that wifi etc are using 28nm nodes, which means they're probably lower power than external chips which tend to use older processes.
We can't use these functions even if they are on the chip for several reasons:

1. we don't have the design tools to develop and test the low-level RF circuitry needed for such a design (this is NOT on SoC)

2. we then must get full certification for the whole software/hardware

3. we get more firmware blobs where we don't know how long we can get updates and how compatible they are to future kernels

If we use precertified modules connected through USB/SDIO on one end and an antenna on the other end, this saves a really, really big amount of money.
OK, those are good reasons to not take that route. I'll scratch that "advantage".
 
The Snapdragon 805 is announced in Q2/2014, so that would throw us back about 6 - 8 months... not feasable, I think.
 
Design for Snapdragon 800 and let's do the upgrade thing again in a few jears.
If it does come to that point, we would probably be looking at the Snapdragon 1000. :lol:
 
That's what I was hoping for :) .
Why are you so fixed on the Snapdragon?


Unless Qualcomm can provide Linux drivers and a 2D accelerator (which seems unlikely, since they're concentrating on Android), we'll most probably not use it.
Because I think Intel is the worst way to go.

If performance is the issue then Snapdragon seems the choice over the OMAP.

All my Snapdragon arguments are against Intel, not TI.

But to be honest:

Let's stop it right now.

Take the OMAP 5.

Be happy.

It's a nice chip and important: We know what we get.

I still don't trust those other SoCs.
 
Yeah I think the back and forth on SoCs should stop now.  Let ED and his team decide behind closed doors now that you all had your say.  Otherwise we won't have a Pyra for another Two-MonthsTM
 
I would be happy with either OMAP5 or Snapdragon 800. It is of course up to ED and gta04 and notaz etc (lets call them the DBPT, for DragonBox Pyra Team) to decide. I'm just saying that in my opinion, the Snapdragon is (somewhat) better in terms of specs -- but of course specs are not everything.
 
Well, we need to write down all the specs and pros and cons.
Yes, exaclty what I thought. Collect all needed infos and compare each Cipset in relation to it's usefulness for the Pyra. Don't forget to also count in the price, how long the chip is available, Power Consumption etc.

If only all of these Chips would be Pin-compatible, so we could simply switch the Chipset on the fly *dream*... ^^
 
I do agree that we should stop arguing without getting any Snapdragon confirmation. :lol:
 
Last edited by a moderator:
This actually is pretty absurd. Especially the Android part. Intel wants companies to actually use their chips (yes, also on Android not only Windows), so they have to supply working drivers.
Sometimes the reality differs from your expectations.

There is no android devices on z3* atoms and there's not very much windows tablets either comparing to the last year although z2* and z3* atoms were announced at the same time of year. It was not the greatest launch and I would not be surprised if Intel decides to silently skip this generation in favor of next-gen atoms, focusing on their support first (+broadwell)

Well, we need to write down all the specs and pros and cons.
Unless we're given specific timeframe this discussion will continue forever.

But it doesn't matter actually. Unless someone decides to create another pocketable linux computer with good hardware keyboard people who want such device (including the ones flaming here) will buy pyra even if they hate SoC, wi-fi chip, yourself or anything that can be hated about this device.
 
Last edited by a moderator:
Back
Top