No time for a break after GamesCom

[EvilDragon]

Just some quick correction, as we just discussed about that in IRC:

fbturbo is NOT using NEON as accelleration.

I misread that.

From the description (which can be found here: https://github.com/ssvb/xf86-video-fbturbo)

any ARM based system should see better

performance thanks to some additional optimizations (the elimination of

ShadowFB layer, ARM NEON/VFP code for dealing with uncached framebuffer

reads, automatic backing store management for faster window moves).

 I misinterpreted that for: The ShadowFB layer had been eliminated and replaced with ARM NEON/VFP code.

But that's not the case, both are eliminated to speed up everything.

Sorry for the confusion.

 

[Illuminerdi]

But NO, God no burn it with fire DO NOT wrap the trigger around the lower/back edge of the case and the exposed corner again!

That is the biggest issue I've had with either one of my Pandoras is that corner getting bumped and jamming and in one case breaking the shoulder button.

DO NOT PUT BUTTONS ON THE EXPOSED CORNERS OF THE DEVICE.

Yes, I'm yelling because I see a train wreck in progress - and everyone appears to be climbing on board and forgetting the last go 'round with the exact same issue.

If done right I don't think this will be a problem.  The DS Lite and 3DS XL both use this style of shoulder buttons and I really prefer the feel of these.  Yes they can lead to fragility if done improperly, but with a strong enough post and thick enough collar around the post I see no reason this could not survive a reasonable amount of shock/damage.  If the button jams it shouldn't be due to the shape of the shoulder button.  My DS shoulder buttons never once jammed, even after MANY years of *heavy* use - so it is more than possible to make a shoulder button that wraps around the bottom without jamming.

Personally, I much prefer the bottom-wrap style, as I tend to rest my index fingers on the bottom corners of a handheld to provide a small bit of extra lift and stability - this frees up the rest of my grip for a slightly lighter touch, making it more nimble and able to shift around quicker.  Plus it makes one handed operation (with the stylus in the other hand) easier, since you your non-stylus hand can work the shoulder button and provide lift simultaneously, rather than having to choose between the two. So for me, a shoulder button that wraps the bottom is strongly preferred.  I'm not saying you don't have a right to your preference, but if your fears are rooted in wear-and-tear and not physical comfort, I think those are solvable problems.

Despite it's loooooong development time the Pandora didn't get nearly as much testing/QA time during the prototype phase that it needed.  ED has a good head on his shoulders and I have confidence that he will VERY thoroughly test every single aspect of the Pyra before he sends it off for production, and he also provides excellent after-sales support, so we can count on him to own and assist with any issues that do crop up long-term, within reason.

 

[Illuminerdi]

Is there any chance that the upgrade CPU could be 64 bit a year or two after the launch?  I noticed that nVidia's new K1 CPU comes in a 4 core 32 bit and a 2 core 64 bit variety.  I'm thinking that two years from now would be a great time to go 64 bit.

Also how would this work with port support going forward?

Also if this is going to be something that is can endure for a long time with a CPU upgrade then I think that the Pyra should use the new USB-C port.  I just think that it would be a shame to come out with the Pyra near the end of a port legacy.

An Nvidia chip is unlikely to ever be put into the Pyra for many reasons.  The biggest reason is the pin layout - the CPU's pin arrangement is one of the most crucial aspects of the PCB design, since there are so so so many traces that go to the CPU. Changing to an Nvidia CPU might necessitate a complete redesign of the entire Pyra PCB.  While my knowledge of the embedded CPU world is limited, I doubt there is a standard pinout for ARM CPU designs, meaning that it's just not possible. Then you have the power/thermal issues to consider - the Nvidia chips are much more powerful, but may generate more heat or draw more power than the existing CPU.  This means testing the ability of the chip to remain within safe operating temperatures.  Plus if it draws more power, it'll eat battery faster, and since battery space in the Pyra is incredibly limited, there's little chance that ED could just put in a bigger battery to compensate - meaning that an upgrade to any Nvidia chip would likely come at the cost of battery life.

Even if an Nvidia chip was, by some *miracle*, compatible and easily able to be swapped in for the existing CPU without needed a ton of hardware/software changes, Nvidia's mobile chips are among the most expensive on the market, and since the Pyra can't buy in the kind of bulk quantities that other companies can, they'd get hit even harder on that price, unless they can find some factory somewhere that has a bunch of spares they're looking to get rid of, but that's not a reliable enough thing to build a product around.

Upgrading to a 64-bit version of the same CPU may be a similar story - I have no idea if upgrading to 64-bit version of the current chip would have a different pinout, but if it did you could guarantee the Pyra won't ever get one.  But more importantly - it doesn't need one, even if it IS possible.  The Pyra doesn't need 64-bit native functions.  The Pyra will not have more than 4GB of RAM, so there's no need to be able to address that much memory at once, and the software benefits of 64-bit processing are limited for a system like the Pyra.  A desktop system doing operations that involve crunching large numbers (encoding videos, compressing/decompressing large files, encryption/decryption, rendering software/CAD, etc) might make good use of 64-bit architecture, but the Pyra just doesn't *need* it, to be frank.  I doubt anyone's going to be doing CAD or 3d Modelling on the Pyra, and nobody is going to make a game natively for the Pyra that would benefit from a 64-bit CPU.

What IS much more feasible would be faster or better versions of the existing Pyra CPU being dropped into future models.  An increased clock-speed version of the CPU that ships with the Pyra could be dropped in with minimal changes - whether or not this ever happens is likely going to be based on how well the Pyra sells and if there is enough demand in a year or two for an upgraded model.  An increased version of the Pyra's CPU with more cores is a lesser known question - I don't know if they make a quad core version of the Pyra CPU and if they do it would need to be pin compatible (which I believe would be likely, I don't see how adding cores alone would necessitate change the pin layout of a CPU, but I could be wrong), but in theory we may someday see a Quad Core version of the Pyra, again this will likely be either if there is strong enough demand to garner enough pre-orders for ED to finance design, testing, and shipping this upgrade.

So really, if you want to see these things happen and see the Pyra improve, the best thing you can do is BUY ONE.  Or buy two.  Or ten.  Get your friends and family to buy them.  The more Pyras that ED sells, the more likely you are to see revisions and upgrades!

(Edits were to clarify a few ideas and fix typos)

 

[Granitehead]

Changing to an Nvidia CPU might necessitate a complete redesign of the entire Pyra PCB.

That's the assumption you got wrong: there is a modular design. You would only need to redesign a small cpu board which plugs into the mainboard. That's still a lot of work, but much easier than a complete redesign, and much cheaper for the customer who wants to upgrade.
I don't think it's likely there will be a completely pin compatible successor to the OMAP5, that's not how it tends to work.

 

[Illuminerdi]

Changing to an Nvidia CPU might necessitate a complete redesign of the entire Pyra PCB.

That's the assumption you got wrong: there is a modular design. You would only need to redesign a small cpu board which plugs into the mainboard. That's still a lot of work, but much easier than a complete redesign, and much cheaper for the customer who wants to upgrade.

I don't think it's likely there will be a completely pin compatible successor to the OMAP5, that's not how it tends to work.

That depends - yes the CPU does reside on the daughterboard, so yes in theory it would only require redesign of that PCB, but that's still an 8-layer PCB to redesign, no small feat.  Plus you have to factor in the position of the two connectors on the ends of the daughterboard and whether or not the pin layout of those is compatible with the theoretical Nvidia chip replacement.  Niklaus may be stretching that 8 layers worth of PCB to its limit, or he might have extra breathing room.

I'm not saying it's completely impossible - but it really boils down to the age old questions of time and ROI.  If it would take a bunch of time to redesign the hardware for a CPU change that only a few hundred people would want or could afford, then it wouldn't happen.  If the Pyra sells 10x more units than ED expects and he has enough profits and demand to make an Nvidia-chipped revision, then it might happen someday.  But pragmatically speaking, it's just plain unlikely.  The Pyra is an awesome system, but it's a niche system and it's not going to make ED rich.  It will hopefully sell enough to cover the development costs and make ED and his team *some* profits for their hard work, but realistically speaking he'll sell between 1000 and 10,000 Pyras in the entire life of the system.  Good numbers, for sure, but hardly large enough to cover the cost of a major hardware revision with one of the most expensive mobile CPUs in existence.

Again, I think a MUCH more likely scenario is an upgrade in clock speed or number of cores.  The modular design of the CPU means that in theory ED could sell this separately after the fact and make it a drop-in replacement for those of us adventurous enough to do so.  The prospect of this has me much more excited for the Pyra since this could extend its lifespan for several years compared to the need to purchase a whole new device with case/keys/screen/battery of similar specs, all to get that CPU upgrade.

 

[Illuminerdi]

I actually have a question for ED regarding the Pyra's LCD - could you post a more detailed breakdown of the specs on it?  I've searched the forums and I've been unable to get this answered.

My main concern is this: is the LCD a TN (Twisted Nematic) or IPS (In Plane Switching)?  IPS panels always have much better viewing angles over TN panels. In my opinion, viewing angles are absolutely crucial on a portable device.  IPS panels also have greatly increased color reproduction (colors are brighter, crisper, and have a lot more "pop" to them) but, to me, this is less important than view angle.  I have used several portable game consoles from JXD and GPD and the TN-panel ones just look like garbage, and are uncomfortable since I have to hold them absolutely perfect to keep the colors adequate. They're borderline unusable - granted these are low-quality Chinese tablets, so the LCDs are the worst of the worst even for TN, but even the best TN panels are still worse than most IPS panels.

You won't find a single smartphone currently on the market today from a major manufacturer that uses any form of TN panel.  The way you hold a smartphone and use it at various angles necessitates the best viewing angles, and I don't see why the Pyra wouldn't feel the same about this - any portable device needs the widest viewing angle it can get, and as I said, even the absolute best TN panels have more narrow viewing angles than almost every IPS panel around.

Yeah, these will cost more, but you can get a 1280x760 IPS panel for around the prices you've quoted (40 - 50 EUR) and they will make a world of difference.  Plus, ED has mentioned things like using GIMP on the Pyra - even a really good TN would harm this because it would not do the colors justice, and t will look worse than it did on your Pyra the moment you view it elsewhere - phone, tablet, desktop, tv, etc.

 

[Granitehead]

That depends - yes the CPU does reside on the daughterboard, so yes in theory it would only require redesign of that PCB, but that's still an 8-layer PCB to redesign, no small feat.  Plus you have to factor in the position of the two connectors on the ends of the daughterboard and whether or not the pin layout of those is compatible with the theoretical Nvidia chip replacement.  Niklaus may be stretching that 8 layers worth of PCB to its limit, or he might have extra breathing room.

Again, I think a MUCH more likely scenario is an upgrade in clock speed or number of cores.  The modular design of the CPU means that in theory ED could sell this separately after the fact and make it a drop-in replacement for those of us adventurous enough to do so.

First: Those 8 layers aren't fixed. You could also use 10 layers, or even more, it just becomes more expensive (or less expensive, if you need less).Second: I think it's highly unlikely the position of the connectors would only work for a specific SoC.

Third: I wasn't specifically talking about Nvidia chips.

Fourth: Find me a series of SoC chips which stayed pin compatible over the generations. I don't think there are any. The most you are likely to get is a small increase in speed, like the different versions of the Pandora CPU, if you are /very/ lucky maybe more cores. I'm not sure something like this is good, it just makes it more difficult to support because there are incompatibilities. It would be much more desireable to wait like 5 years and then do a jump to a completely new chip. If the interface is done smartly (now), then you could probably still use the old mainboard with the new module. Looking at the example of USB, the new chip would probably support more than one USB 3.0 port. But if there are spare lines in the interface, and you put the addittional USB 3.0 lines there and leave the other lines where they were, they would fit with the USB 2.0 lines of the old mainboard and could power them (i.e. the USB 3.0 port of the SoC would work as a USB 2.0 port). A new mainboard revision would also route the additional USB 3.0 lines to the plug, so it would actually work as a USB 3.0 port. I'm pretty sure it would work the same way for those new faster SD cards: they are inherently backwards compatible, they just add additional lines.

That or an x86 version (done by someone else, but working together with ED) seem like much likelier use cases of the modularity to me.

 

[Silent-Hunter]

The 5.3" Samsung AMS529HA01 is 800x1280, but thats pentile matrix oled.

What's wrong with pentile matrix OLED?

 

[Illuminerdi]

First: Those 8 layers aren't fixed. You could also use 10 layers, or even more, it just becomes more expensive (or less expensive, if you need less).

Second: I think it's highly unlikely the position of the connectors would only work for a specific SoC.
Third: I wasn't specifically talking about Nvidia chips.


Fourth: Find me a series of SoC chips which stayed pin compatible over the generations. I don't think there are any. The most you are likely to get is a small increase in speed, like the different versions of the Pandora CPU, if you are /very/ lucky maybe more cores. I'm not sure something like this is good, it just makes it more difficult to support because there are incompatibilities. It would be much more desireable to wait like 5 years and then do a jump to a completely new chip. If the interface is done smartly (now), then you could probably still use the old mainboard with the new module. Looking at the example of USB, the new chip would probably support more than one USB 3.0 port. But if there are spare lines in the interface, and you put the addittional USB 3.0 lines there and leave the other lines where they were, they would fit with the USB 2.0 lines of the old mainboard and could power them (i.e. the USB 3.0 port of the SoC would work as a USB 2.0 port). A new mainboard revision would also route the additional USB 3.0 lines to the plug, so it would actually work as a USB 3.0 port. I'm pretty sure it would work the same way for those new faster SD cards: they are inherently backwards compatible, they just add additional lines.


That or an x86 version (done by someone else, but working together with ED) seem like much likelier use cases of the modularity to me.

I'm not trying to start a long debate on this, so I'll try to keep this brief.

True, the number of PCB layers isn't fixed and adding layers is fairly trivial and as long as you have more pins on the connectors than you have total number of traces to/from the main PCB, there should never be a problem in that regard, but redesigning an 8-layer PCB, even a small one, is still no small task, but this modular design does very smartly at least make CPU changes much more possible than a CPU that is mounted directly to the main PCB, which is more likely to need redesign of the PCB in the even of a CPU change.

As far as the CPU goes though, I *highly* doubt we will ever see someone make an x86 chip for the Pyra. The time and money required to design and manufacture a PCB with an x86 chip for a system that will sell a few thousand units total, with maybe a fraction of those people wanting to spend the additional money to move to x86.  Plus doing so would mean that someone would have to make major changes to the Pyra's version of Linux.  They'd have to compile a whole new core based on x86, would have to swap all of the drivers for various components (CPU, GPU, Sound, Touchscreen, LCD, etc etc) for x86 drivers (if they even exist for all the parts in the Pyra), test and ship.  THEN various developers would need to port their apps from ARM to x86.

And frankly, if you want x86 on the Pyra so that you can do something like run Windows, why not just get a netbook or something?  They'll cost less and probably be more powerful and have more features. While x86 will still be relevant throughout the life of the Pyra, it is incredibly unlikely that anyone will want or make an x86 CPU module for the Pyra.

The OMAP 5 in the Pyra is one of the first OMAP5's in existence, so if TI does make future revisions that are die shrinks I see no reason that ED won't at least consider putting the new chips into the Pyra or offering them as an upgrade.  If TI never releases a revision of the OMAP 5, however, it's far less likely we'll see any other CPUs available for the Pyra.

 

[WizardStan]

What's wrong with pentile matrix OLED?

The way the subpixels are arranged it is impossible to make a perfect blue or red line, pixels all contain green but alternate red and blue so some resampling is required. How much of a difference the human eye can actually detect in different situations and how often such situations come up is a matter of controversy.

 

[Silent-Hunter]

Why the hell would anyone design it like that?

 

[TeDaDeS]

The OMAP 5 in the Pyra is one of the first OMAP5's in existence, so if TI does make future revisions that are die shrinks I see no reason that ED won't at least consider putting the new chips into the Pyra or offering them as an upgrade.  If TI never releases a revision of the OMAP 5, however, it's far less likely we'll see any other CPUs available for the Pyra.

I think there is a possibly a newer board will be created with more RAM, maybe pin-compatible omap5 upgrades to slightly faster speeds.

The x86 option would still be nice, netbooks are indeed cheap and small. But umpc still exist, they don't sell large quantity and are really expensive.

The Pyra with x86 would be cheap alternative I think, but it might not be worth the investment from ED's side (x86 PCB is more difficult to develop).

 

[Null]

Why the hell would anyone design it like that?

It uses less power.

 

[Illuminerdi]

Why the hell would anyone design it like that?

Why else?  Reductions in price and/or power.

 

[Grench]

But NO, God no burn it with fire DO NOT wrap the trigger around the lower/back edge of the case and the exposed corner again!

That is the biggest issue I've had with either one of my Pandoras is that corner getting bumped and jamming and in one case breaking the shoulder button.

DO NOT PUT BUTTONS ON THE EXPOSED CORNERS OF THE DEVICE.

Yes, I'm yelling because I see a train wreck in progress - and everyone appears to be climbing on board and forgetting the last go 'round with the exact same issue.

If done right I don't think this will be a problem.  The DS Lite and 3DS XL both use this style of shoulder buttons and I really prefer the feel of these.  Yes they can lead to fragility if done improperly, but with a strong enough post and thick enough collar around the post I see no reason this could not survive a reasonable amount of shock/damage.  If the button jams it shouldn't be due to the shape of the shoulder button.  My DS shoulder buttons never once jammed, even after MANY years of *heavy* use - so it is more than possible to make a shoulder button that wraps around the bottom without jamming.

Personally, I much prefer the bottom-wrap style, as I tend to rest my index fingers on the bottom corners of a handheld to provide a small bit of extra lift and stability - this frees up the rest of my grip for a slightly lighter touch, making it more nimble and able to shift around quicker.  Plus it makes one handed operation (with the stylus in the other hand) easier, since you your non-stylus hand can work the shoulder button and provide lift simultaneously, rather than having to choose between the two. So for me, a shoulder button that wraps the bottom is strongly preferred.  I'm not saying you don't have a right to your preference, but if your fears are rooted in wear-and-tear and not physical comfort, I think those are solvable problems.

Despite it's loooooong development time the Pandora didn't get nearly as much testing/QA time during the prototype phase that it needed.  ED has a good head on his shoulders and I have confidence that he will VERY thoroughly test every single aspect of the Pyra before he sends it off for production, and he also provides excellent after-sales support, so we can count on him to own and assist with any issues that do crop up long-term, within reason.

Wrap the side OR wrap the bottom - that is fine.  Get the button OFF of the corner though.

I don't have and don't want a DS.  I -do- want the Pyra to be an improvement over the Pandora.  Having a button on a device corner is a 100% mistake regardless of the method of construction.  

There is no need to put moving parts in the two positions most likely to get bumped regularly.  Those two back corners get bumped every time you set the device down.

There is talk in the keyboard threads around using the shoulder buttons for music playback control while pocketed.  The corner-wrapped shoulder buttons will miss-click (butt-click) in side pockets from just the fabric stretching over the device corner.

Make the corner thick plastic and not a button.  ED's designs got this right - lets not regress please?

 

[Silent-Hunter]

Is it really that noticeable? The red/blue line thing I mean.

 

[Illuminerdi]

The OMAP 5 in the Pyra is one of the first OMAP5's in existence, so if TI does make future revisions that are die shrinks I see no reason that ED won't at least consider putting the new chips into the Pyra or offering them as an upgrade.  If TI never releases a revision of the OMAP 5, however, it's far less likely we'll see any other CPUs available for the Pyra.

I think there is a possibly a newer board will be created with more RAM, maybe pin-compatible omap5 upgrades to slightly faster speeds.

The x86 option would still be nice, netbooks are indeed cheap and small. But umpc still exist, they don't sell large quantity and are really expensive.

The Pyra with x86 would be cheap alternative I think, but it might not be worth the investment from ED's side (x86 PCB is more difficult to develop).

ED has said in the original post that the CPU board being swappable is likely to enable CPU upgrades, so it seems very likely that we'll see new CPU board(s) for the Pyra within a year or two of its release.

Again though, I highly, highly doubt we'll ever see an x86 CPU board.  ED probably won't ever release an x86 board for the Pyra - doing so would mean splitting the entire user base between ARM-based and x86-based OSes/apps.  He'd have to find people to make the x86 versions of Linux (and support them), and people to make and support x86 app depots. Plus, devs (most of whom do this for free, mind you) would have to port their apps to x86 and maintain 2 different, simultaneous versions of their apps.  Or ED could make an x86 board and either nobody or everybody bought it, and whichever architecture "lost" would stop seeing development/active support.

Is it really that noticeable? The red/blue line thing I mean.

I haven't seen it myself, but I'd guess it's one of those things you can only really see if you're "looking for it" - it might factor into overall color quality though and result in an overall (slightly) poorer image onscreen, I'm not sure.

 

[Granitehead]

Is it really that noticeable? The red/blue line thing I mean.

If there is controversy on the subject of whether it's visible or not, then it's probably not visible, but some people swear they see it (as long as they aren't being tested on their claims). Same as for the magic audio guys who hear much more frequencies and stuff.
Edit: @Illuminerdi: I don't agree, but I don't think we need to discuss this.

 

[Silent-Hunter]

Or someone else could make one, and people would have to choose which community they want to be in.

Is it really that noticeable? The red/blue line thing I mean.

If there is controversy on the subject of whether it's visible or not, then it's probably not visible, but some people swear they see it (as long as they aren't being tested on their claims). Same as for the magic audio guys who hear much more frequencies and stuff.

Then why don't we use that screen?

 

[Null]

Then why don't we use that screen?

It depends on whether ED can get hold of one or not.