Pandora successor requirements / wishlist / ..


bsp

Very Active Member
Joined
Dec 2, 2006
Messages
314
There's been a lot of talk about which SOC should be chosen for a Pandora successor.

I am wondering: Shouldn't we first talk about the requirements a SOC has to fulfill to be considered as a successor ?

I do realize that, in regards to the SOC, there _currently_ are not that many options available anyway, so this is not only about the SOC itself but also covers e.g. peripherals / controls / casing.


From my point of view (in comparison to the (current) Pandora), these requirements (..you may call it my personal wishlist..) are:
 

  • battery life / power consumption: should be as good or better (at the same tasks) (highest priority for me, actually)
  • OS: I prefer Linux. Being able to run Android or Windows would be nice-to-have but definitely not a must.
  • full open source: preferably yes but I don't have a problem with e.g. a closed source GPU driver (unless it's a bug fest)
  • documentation: as a dev I want publically available, properly written/detailed hardware documentation
  • compatibility/ISA: there's a _lot_ of (ARM) SW available for the Pandora so being able to run (most of) that on a "P2" would be very nice. On the other hand, most of the current SW can probably be recompiled for e.g. an x86 based "P2" without much hassle. Personally I like the ARM ISA better than x86 but since I don't do much assembly coding these days I would not object to x86 _if_ a potential x86 SOC is actually better than (not only as good as) a potential ARM SOC
  • wifi: nice to have but I could live with USB wifi (these things are tiny and cheap) (+problem with japanese market)
  • bluetooth: nice to have but not too important to me (+problem with japanese market) (like WiFi, small+cheap USB dongles are available)
  • NAND: should be rather large (16gb+) if there is no 2nd sdcard slot. with 2nd sdcard 512-1gig (just for the firmware) is enough.
  • sdcard: normal size sdcards please. second sdcard slot is preferable to large NAND flash
  • HDMI: very important. max. supported resolution should be at least 1920x1200 so I can connect it to my beamer and PC monitor w/o rescaling
  • tv-output: not needed if HDMI is available
  • SATA: important. along with HDMI + external hd/mouse/kbd, the handheld can be used as a "workstation" (in that case, boot from SATA instead of NAND/sdcard).
  • RAM: don't go crazy there, the 512MB on my Pandora are more than enough. maybe 1GB to be more future-proof ?
  • CPU: almost any multi-core SOC today is at least 10 times faster than the Pandora. should rather watch for energy efficiency. the big/little concept (e.g. 2 high performance cores + 2 energy-saving cores, like on OMAP5) sounds good but how well is it supported by Linux ? Last but not least, in reality a lot of software is single-threaded and will not benefit from a multi-core processor so the actual performance gain is probably more like factor 2..3.
  • keyboard: very important. I totally dislike onscreen keyboards. should have real keys instead of rubber mat.
  • gaming controls: very important. 2*L/R shoulder buttons + 6-buttons would be nice but I can live with 1*L/R shoulder b. + 4-buttons. don't lose the d-pad and nubs
  • touchscreen: multitouch would be nice although I am hardly using the touchscreen on the Pandora (because of the clam-shell design ?). I'm fine with the WVGA resolution of the Pandora but it might make sense to increase that to e.g. 1280x800 ?
  • display controller: at least two graphics layers + hardware scaling support (very important) (Pandora has 4 + hw scaler)
  • video decoding: 1080p hw-accelerated, please (h264, mpeg*, ..)
  • audio output: should be as good or even better
  • audio input: at least via adapter cable.
  • speakers: I would pay a few bucks more for a decent set of speakers, the ones in the Pandora are not very good, IMHO
  • USB: onboard hub please. does not have to be 3.0. at least two type a (standard size) connectors (+mini USB for charging / USB network?).
  • accelerometer: seems to be cheap and is a nice alternative controller for games, would like to have one
  • LTE/UMTS: nice to have but don't really need it (it's expensive, right?) (can be added via USB, anyway)
  • camera: personally I do not need it. it could even be a disadvantage since some companies do not allow these things in meetings.
  • GPU: hard to say, definitely something beefier than the SGX530 but I do not need cutting edge 3d performance. would buy a Shield if I wanted that. Last but not least: what software would justify a high performance GPU ? PS2 emulation would be nice but I guess that's a pipe dream even with the best mobile phone tech available today (and the next few years). However, the GPU must be able to run a lag-free, smooth-scrolling, full HD desktop. And yeah, it's part of the SOC and can therefore not be freely chosen, anyway.
  • case: stick with the clam-shell design. I would pay extra for a nice alumnium shell (no issues with WiFi when external USB wifi dongle is used). maybe offer an alternative low cost plastic case ?
  • size: definitely not larger than the current Pandora but not much smaller, either. if possible, a bit thinner.
  • name: what's in a name ? I could get used to "Dragonbox", though.
okay that became a long list.

tl;dr: I want it to be a small Linux netbook that can, when connected to an external HD/monitor/mouse/keyboard, become a full workstation. The key features needed for that are HDMI + SATA. I also want a portable game console that fits into the pocket of my jacket.

(by "workstation" I mean something reasonably specced for (my) everyday tasks like web-browsing, watching videos, writing+debugging code, some audio editing/recording (nothing too fancy))


What do you think ? What would you expect from a Pandora successor ?
 
 
I'm sure "documentation" is one very important point but the other points are also worth to discuss, nice list, summs it up pretty well. :)
 
I mostly agree with all that, but would flag up wlan as being very important, and more than just a nice-to-have.
 
Half of that list is stuff that sits on the board and isn't part of the SoC

We have a lot of redundant threads lately :-/

I want an FM and IR transmitter
 
Last edited by a moderator:
I know I'm a minority here, but if we lost the regular TV out to HDMI I wouldn't care, but I would still like to have gpio outputs like the EXT connector has and perhaps add external access to the I2C as well.

I'd still like the successor to be hardware hackable friendly...
 
Last edited by a moderator:
"1280x800"

WTF? That's more than in my beamer.

Why would you need that if it's a pocket sized screen and no tablet?
 
I'm actually being won over more and more by DisplayPort.  I'm a bit confused by it though.

I saw a reference that going from Display Port to HDMI requires an active translation box.

However, I see passive dongles that go from Display Port to DVI and passive dongles that go from DVI to HDMI.  Yes, audio gets lost in that path.  Is there some reason why a person couldn't do video directly via a passive dongle/cable from Display Port to HDMI?
 
However, I see passive dongles that go from Display Port to DVI and passive dongles that go from DVI to HDMI
Basic DisplayPort can't do passive conversion to DVI/HDMI, it requires an active converter. There's an add-on standard, "Dual-Mode DisplayPort" which transmits both DisplayPort and DVI signals and the passive adapter simply chooses the right signal. I'm not sure if Dual-Mode has become so ubiquitous that it is just called "DisplayPort" now, or if there's a lot of disappointed people buying something they can't use.
 
As far as I can tell, Dual-Mode is pretty much the norm. If we go for DP, I'd certainly make sure that it is a dual-mode one. HDMI is not going to die any time soon, and having a passive adapter is a big advantage.
 
Great list bsp!  I almost agree :)

I would want an additional usb port if wlan was not included. Seems a bit strange not to include though.

Size: love to see a larger screen than current screen - the larger the better =< 7". So imo it is very ok if that makes P2 larger - larger but preferably thinner. Haven't made up my mind if resistive or capacitive is better.

Keyboard - must have

Compatible - More than nice. If it is not compatible I would probably wait to buy until the software repository is similar to the current Pandora.

The rest of the list I agree or don't really care.
 
Compatible - More than nice. If it is not compatible I would probably wait to buy until the software repository is similar to the current Pandora.
Wouldn't most of the stuff on the repo have to be recompiled anyway?
 
 I mostly agree with all that, but would flag up wlan as being very important, and more than just a nice-to-have.
WiFi is very important to me, too but since it is an issue with the japanese market (and maybe a design challenge with an aluminium case, as other threads suggested), and USB WiFi dongles are very small and cheap, I would not mind if WiFi was not "on board".

I would still like to have gpio outputs like the EXT connector has and perhaps add external access to the I2C as well.

I'd still like the successor to be hardware hackable friendly...
yay for hacker friendly, wouldn't hurt to have a few GPIOs although this is not the kind of HW I would use for experiments. A physically decoupled (e.g. optocoupled) interface to the outside world would be nice, though.

"1280x800"

WTF? That's more than in my beamer.

Why would you need that if it's a pocket sized screen and no tablet?
I did not say I needed that, actually I'm fine with WVGA but we are talking about a device years away from now so a slight resolution bump (e.g. to display 720p video) might be reasonable. (and my two years old <1k EUR beamer's native resolution is 1920x1200)

Basic DisplayPort can't do passive conversion to DVI/HDMI, it requires an active converter. There's an add-on standard, "Dual-Mode DisplayPort" which transmits both DisplayPort and DVI signals and the passive adapter simply chooses the right signal. I'm not sure if Dual-Mode has become so ubiquitous that it is just called "DisplayPort" now, or if there's a lot of disappointed people buying something they can't use.
As far as I can tell, Dual-Mode is pretty much the norm. If we go for DP, I'd certainly make sure that it is a dual-mode one. HDMI is not going to die any time soon, and having a passive adapter is a big advantage.
if that's true then we could go for DisplayLink (don't mind a passive adapter. Actually, one of these arrived for my new office PC just today (well, yesterday))

Wouldn't most of the stuff on the repo have to be recompiled anyway?
No idea. Will everything "just run" as long as it's ARM?
If properly done, most of it will, yes.
let's not kid ourselves. SW will have to be recompiled and adjusted to the new hardware (e.g. display size). A compatibility layer would just be a temporary solution. It's more about _how much_ effort it will be to port existing SW to a new device. ISA (ARM) compatibility would make things much easier, IMHO.

Size: love to see a larger screen than current screen - the larger the better =< 7"
Personally I do not think so. If I wanted a tablet I would buy one, I want a pocket computer.
 
Last edited by a moderator:
SW will have to be recompiled and adjusted to the new hardware (e.g. display size)
The Pandora's GPU has a hardware scaler, you can request basically any resolution and it scales it to the screen. SDL handles this entirely automatically, the current screen could be swapped out for something twice the resolution and the software wouldn't have any idea, they'd continue to request the same size they always did and it would work just fine.The "compatibility" layer isn't temporary, it already exists in the form of libraries. By using OpenGLES, SDL, ALSA, and a slew of other standard libraries for handling IO, we can change the hardware and rebuild the libraries and all applications should continue to work without any need to recompile.

Some exceptions exist, of course, but those will be in the minority.
 
that's all true but you know how pixel graphics look like when they are being scaled to another (arbitrary) resolution. that's why Apple for example was very careful to increase screen estate in integer multiples (IIRC).

EDIT(later today):

My statement about the pixel graphics scaling is not really an argument, of course (it was late, I was tired..)

Running Pandora apps on an ARM based "P2" would very well be possible without recompilation.

Some things might be a bit tricky, e.g. video-scaling apps that use the video scaler themselves, but it's definitely not impossible to work around this.
 
Last edited by a moderator:
Back
Top