Designing The Gp2x's Successor - Hardware Selection Poll

Select one:

  • Option 1: ARM926EJ-S 266 MHz Processor, Floating Point/Vector Coprocessor, 2D/3D Accelerator

    Votes: 0 0.0%
  • Option 2: ARM920T 400 MHz Processor, No FPU/Vector Unit or Graphics Accelerators

    Votes: 0 0.0%
  • Option 3: These are both inadequate (post alternative)

    Votes: 0 0.0%

  • Total voters
    0

The speed of the processor isn't the important thing about option 1, it's the PowerVR accelerator which makes it so great, allowing for OpenGL and faster PSX, SNES and GBA emulation. It is listed in the first post that N64 could run at acceptable speeds with this chipset choice. Of course, who needs to read posts before they reply anyway? ;)
 
This notion that somehow a system with a 3D GPU will magically be able to emulate all 3D consoles perfectly is a bit false. It will help yes, but unless the 3D chip works EXACTLY like the one in the console you are emulating it still won't be perfect.

Emulating PSX isn't really practical anyway. The games take up way too much space. Filling a card with only 2 PSX games seems like a waste. Don't we have enough 3D out there anyway with all current consoles and handhelds being 3D?
 
Yono posted on Aug 25 2006 at 02:42 PM said:
The speed of the processor isn't the important thing about option 1, it's the PowerVR accelerator which makes it so great, allowing for OpenGL and faster PSX, SNES and GBA emulation. It is listed in the first post that N64 could run at acceptable speeds with this chipset choice. Of course, who needs to read posts before they reply anyway? ;)


If you believe that option 1's acceptable speeds are N64 and PSX at full speed FS0 or FS1 as I believe the next GP has to do to be worthy of upgrading to then good luck to you.

As for the sarcastic comment, I read the entire thread but I chose to believe the N64 and PSX comments were at best optimistic.

Open GL will be great for quake and quake 2 and opens up some nice possibilities for homebrew but personally I am of the opinion that a faster processor is more important than the GPU but I'd rather see both!
 
Last edited by a moderator:
BaDToaD posted on Aug 25 2006 at 10:42 AM said:
Yono posted on Aug 25 2006 at 02:42 PM said:
The speed of the processor isn't the important thing about option 1, it's the PowerVR accelerator which makes it so great, allowing for OpenGL and faster PSX, SNES and GBA emulation. It is listed in the first post that N64 could run at acceptable speeds with this chipset choice. Of course, who needs to read posts before they reply anyway? ;)


If you believe that option 1's acceptable speeds are N64 and PSX at full speed FS0 or FS1 as I believe the next GP has to do to be worthy of upgrading to then good luck to you.

As for the sarcastic comment, I read the entire thread but I chose to believe the N64 and PSX comments were at best optimistic.

Open GL will be great for quake and quake 2 and opens up some nice possibilities for homebrew but personally I am of the opinion that a faster processor is more important than the GPU but I'd rather see both!
You are right, the post is entirely speculation, but at this point, isn't that all we have? Ok, sorry about the sarcastic comment, it seemed as if you had just posted based on the last few comments. I'd rather see both also, but sadly I don't believe GPH has the budget for that. Maybe GP's next gen will utilize one chipset and GPH the other, stranger things have happened.
 
Last edited by a moderator:
BaDToaD posted on Aug 25 2006 at 06:13 AM said:
The next gen GP handheld needs to be able to emulate first gen 3D consoles at full speed with FS0 or at worst FS1 or we might as well stick with what we've got.
That's what I was saying, and got lightly flambayed. That should be the goal - no compromise psx/n64/gba.

but why stop there

I still think that if they put a HD in there, and made it a swell PMP in addition to a solid playstation / n64 / gba on-the-go box, you'd have a dream system worthy of a US $400 purchase price, and worthy of international recognition as the damn coolest thing ever.

Now if there were only a way to keep it running for 3 hours on 2 AA's :blink: :huh: :lol:
 
Last edited by a moderator:
I couple years back I got all excited from a couple of press releases from VIA Technologies as I was certain that they were about to build the greatest handheld gaming device ever. A couple years later it’s nothing more then another piece of vaporware, but I think the concept of this system is still quite valid despite it never becoming a reality.

http://www.via.com.tw/en//initiatives/spea.../grace/moma.jsp

The MOMA Mobile Gaming Console is a PC in the shape of a mobile gaming console. It’s as simple as that, and what this gives you is the largest library of software imaginable. No need to port games, you can just run them natively. Now given the “modest” specs of this system you’ll be limited to games of the “pre-millennial” era but that just sounds fine to me. Also given that this is an x86 processor all of those emulators that make heavy use of assembly will run just fine. So given all of this the obvious question is why didn’t this console ever get made? I really don’t have a clue, perhaps it was to ambitious a system to make, a design ahead of it’s time. We’ll a couple years have passed, and a couple more will pass before we’re ready for the GP2X’s successor, and by then maybe this will look reasonable.

I’ll go through the system specs for the MOMA Mobile Gaming Console, and describe how I would change them.

System
» 533 MHz Eden-N Processor
» 133MHz Bus
» SSE and MMX instruction Set support
» 128MB DDR266 SDRAM
» High-Performance 64-bit DDR SDRAM Controller
» 20GB Hard Drive - Data transfer rates of up to 133 MB/s

CPU & RAM look good, but the hard drive seems like an expensive, battery consuming and potentially breakable part. Unfortunately there are several obstacles must be overcome when eliminating the internal hard drive. All currently viable solid-state non-volatile storage media have a finite number of writes that make them unusable with a “stock” operating system that expects to be able to page/swap. MRAM, PRAM, and FRAM are all “neat” but no where near production ready in densities large enough to hold an OS, it’s doubtful that this will change soon enough for them to be a viable option. This unfortunately precludes booting a regular copy of Windows 98 on the device; however it does not eliminate Linux. Luckily with an x86 processor WINE is available, and thus the need to boot windows is greatly reduced. The best bet would be to use a ‘live-cd’ like approach that is not entirely dissimilar from how the GP2X currently works. That is to say you boot off a read-only device, and then load your OS entirely into RAM. However unlike the GP2X I would like to see this done without the use of internal flash memory, simply because this is what leads to bricks. The device should instead have two SD card slots, and it should boot off the first one, and allow the second one to be used as auxiliary storage/SDIO. If the user tries to boot the device without an SD card in it, how to remedy the situation is displayed on the screen. Have the device come with a CD that has an application for Windows, Linux, and OSX that will properly format an SD card and put the OS on it. The booting card itself should be partitioned into two parts. One part should be read-only and thus act as the live-cd, the other part should be read/write, and used to store user preferences, and installed applications. The end result of all of this is that the user decides how much storage they need, and if that storage ever breaks it’s easy to replace. However, because the entire OS will be loaded into RAM, the amount of RAM the system has would need to be increased. 256mb should be plenty, and ideally if this could be done using a SO-DIMM, it could be upgraded later if need be.

Graphics
» VIA CN400 Digital Media Chipset - Integrates S3 Graphics UniChrome Pro Graphics Core
» 200Mhz Graphics Engine Clock
» 128-bit 3D Graphics Engine o Pixel rate up to 200 million pixels per second, 2 textures each
- Triangle rate up to 4.5 million triangles per second
- Microsoft DirectX 7.0, 8.0, and 9.0 compatible
- Microsoft DirectX Texture Compression (DXTC, S3TC)
- OpenGL™ Support
- Z-bias, LOD-bias, Polygon Offset, Edge Anti-aliasing and Alpha Blending
- Specular Lighting
» MPEG2 & MPEG4 Hardware decoding

I see nothing wrong with any of that. UniChrome has good open source drivers, and while not a powerful 3d accelerated by today’s standards it should be fine for a handheld device meet to player older games.OpenGL support is incredibly important as this opens up a lot of games that wouldn’t otherwise be possible. There are plenty of good games for Linux that require OpenGL hardware support. (Neverball, Chromium BSU, and Stepmania for example) For emulation this puts the Playstation and the N64 within striking distance.

Audio
» VIA Vinyl Six-channel Audio
» Full Immerzio™ Gaming support
- EAX® 1.0, EAX® 2.0
- DirectSound3D®
- I3DL2™
- A3D® 1.0

Six Channel Audio… On a handheld… We’ll that’s just a side effect of this being a VIA based platform and Vinyl being the chip that they like to stick on things. Any of VIA’s on-board sound solutions would be fine.

LCD Display
The MOMA Mobile Gaming Console integrates a 4", 640x480 TFT LCD screen. In supporting up to 640x480 resolutions, the MOMA Console can easily play all PC games natively.

640x40 is a necessity. This is the lowest resolution many PC games support, and so this is where we need to be. Running a game designed for 640x480 at 320x200 makes the text look absolutely hideous. Adding a touch screen that could be used to mimic a mouse would be a really nice addition.

Hard Drive
The MOMA Console includes a single 1.8" 20GB hard drive for storing the OS and the game data. This is sufficient to store a large library of games and digital media content. This is transferred to the system either via wireless LAN or by USB2.0 pass-through to a PC.

I’ve already gone over what they need to do about the hard drive – get rid of it.

Batteries
Prismatic Lithium-Ion batteries are the current favorite choice for the console. Two packs of batteries are provided so that the batteries can be hot-swapped for continuous, uninterrupted play.

Lithium-Ion batteries are good, proprietary Lithium-Ion batteries are bad however. I believe the device should use RCR-V3 batteries. (http://en.wikipedia.org/wiki/RCR-V3), and it should ALSO be capable of using AA’s as well. Both NiMH’s and RCR-V3 are good options, and there’s no reason that we can’t have both of them.

Port List
The ports that are user-accessible on the MOMA Console include:
» Power jack (DC barrel type)
» 2x USB 2..0 jacks (Type A host style)
» 1x Compact FLASH type II slot
» 2x battery slots (as specified as handgrips)
» 1x smartcard card slot (15mm x 1mm slot)
» 2x 1/8” stereo jack for headphones out & microphone in
» 1x 1/8” stereo jack for composite video out
» 802.11b wireless LAN

Ditch the smartcard slot, and the Compact FLASH type II. However I’d like to keep the wireless and bump it up to 802.11g. Since we have a microphone jack the device would make for an excellent wifi phone. As I wrote earlier, we’ll need two SD card slots so those will have to be added.

Button List
» D-pad (WASD)
» L/R triggers
» Cluster of 4 action buttons
» Line of 6 auxiliary function buttons

We defiantly need a D-pad, and we’ll need some software that allows us to map the keys/buttons however we want.

So that’s what I want, the MOMA Mobile Gaming Console with a few tweaks made to it.
 
Diji, I love that system, but I don't think either GPH or Gamepark could finance such a system unless the GP2X sells enough units to produce an enormous amount of them, further reducing cost per unit to where we could afford such a device. If I bought it, I would have to either eat it, live in it or wear it as clothing, just from a rough mental guess at what the price would be.
 
DijiTao posted on Aug 25 2006 at 07:51 PM said:
I figured with the changes I'm listing it would cost around $400. The "core" of what I want is something based on VIA Eden-N:

http://www.via.com.tw/en/products/processors/eden-n/

Perhaps my maths is bad but @ 533mhz it would use considerably more juice than a gp2x and that's just for the processor.
I don't think they are really viable for portable systems. The whole MOMA thing is probably just a marketing exercise to generate buzz around their products not a product intended for sale.
 
Last edited by a moderator:
DaveC posted on Aug 25 2006 at 08:08 AM said:
This notion that somehow a system with a 3D GPU will magically be able to emulate all 3D consoles perfectly is a bit false. It will help yes, but unless the 3D chip works EXACTLY like the one in the console you are emulating it still won't be perfect.

Emulating PSX isn't really practical anyway. The games take up way too much space. Filling a card with only 2 PSX games seems like a waste. Don't we have enough 3D out there anyway with all current consoles and handhelds being 3D?

... It isn't magic. It's called OpenGL rendering. Have you ever used a PSX emulator for a PC? All the most successful ones that provide the best performance give you the option of drawing 3D with an actual hardware accelerator through OpenGL or Direct3D APIs. These are standard 3D APIs that operate on any compliant hardware. Such rendering engines would also work on the OpenGL-compliant PowerVR MBX Lite chipset in the S3C2460, or the OpenGL-ES compatible 2D/3D Accelerator in the XGP/XGP Mini.

There is no issue of being 'compatible' with the PSX-- the emulator interpets instructions intended for the PSX's 3D rendering hardware and outputs corresponding instructions to render the same scene on the system's OpenGL/Direct3D hardware. This hardware uses the instructions to assemble a scene that looks just like it would on the actual console.

If you are still not convinced it would work; take into consideration that the PSX's 3D hardware is NOT EMULATED, it is SIMULATED. Instructions from the PSX titles that would normally go to the PSX's 3D hardware are interpreted, converted into equivalents that will run in software on an ARM9 processor, and executed to draw a scene that imitates what the PSX would draw. The ONLY difference is that in the case of the S3C2460 or XGP hardware, a more specialized microprocessor designed for large rendering workloads would be used instead of the ARM9 which is a spectacularly terrible architecture to use for any sort of realtime 3D rendering job.

And more of this railng against 3D .. I don't really care about 3D for the most part, either, as I've expounded on many times. However, the PSX games I do want to play are the Megaman X series, which are 2D, Raiden DX and Raiden Fighters which is 2D, and the Klonoa series which are 3D turned to a 2D perspective (a bit of a gray area.) None of these titles will ever run acceptably enough for me on the GP2X due to the CPU being terribly suited for software rendering-- it doesn't matter how unbelievably fast PSX4ALL gets, it won't get close to full speed on an ARM9 with no custom processor to handle rendering.

If you can't believe ME, I imagine you can believe Zodttd, who has expressed interest in porting an existing OpenGL rendering plugin from a PC emulator, or writing his own. It's perfectly doable.

BaDToaD posted on Aug 25 2006 at 08:42 AM said:
If you believe that option 1's acceptable speeds are N64 and PSX at full speed FS0 or FS1 as I believe the next GP has to do to be worthy of upgrading to then good luck to you.

As for the sarcastic comment, I read the entire thread but I chose to believe the N64 and PSX comments were at best optimistic.

N64 emulation would be feasible but the framerate would be probably about where the PSX is now. It'd most assuredly not reach nearly full speed; but it'd be a lot better than 4 or 5 FPS. :) PSX emulation is much more preactical. If ~20+ FPS is achievable with software rendering, unburdening the ARM9 CPU of all the rendering workload and putting this task on the GPU, also doing vector processing on the AFP9 unit as Zodttd had said he would like to do, there is no logical reason 60 FPS is not attainable in most games.
 
Last edited by a moderator:
DijiTao posted on Aug 26 2006 at 04:51 AM said:
Back when the thing was first announched it was planned to cost "less than $500"

http://www.extremetech.com/article2/0,1558...2129TX1K0000532

I figured with the changes I'm listing it would cost around $400. The "core" of what I want is something based on VIA Eden-N:

http://www.via.com.tw/en/products/processors/eden-n/

In case you're interested, this is poping up again as a chipset with "john" technology.

http://www.mini-itx.com/2006/06/29/vias-john-technology

You'll see it in some ultra mobile PCs soon.

If you want more info on the video, it's basically Savage4 for 3D and Savage2000 for 2D. (This makes it a very good 2D chipset)

I don't think it will be so cheap. The system is much more complicated then what you would see in a GP2X.
 
Last edited by a moderator:
There's no way anyone in their right mind would pay $400 on a handheld, no matter how performant it would be.

The main selling point of the GP2X is that it can do so much at a relatively low cost. We wouldn't even have 5% of the current community if it would cost as much as $400.. When you're talking that kind of money, I'd rather just save a little more and buy a decent notebook (and keep my trusty GP2X :)).

- Alex
 
Alex. posted on Aug 27 2006 at 11:02 AM said:
All I'm gonna say is that there's no fucking way anyone in their right mind would pay $400 on a bloody handheld, no matter how "cool". Goddamn it you're talking about twice the cost of the current system, you wouldn't even have 5% of the current community if it was that expensive. When you're talking that kind of money, I'd rather save a little more and buy a decent notebook instead.

- Alex

Totally agree, if you're that rich, samsung will use the chipset in its new Q1, replacing the intel celeron with intel graphics. 1 gig of ram and 80GB HDD, but it will probably cost more then $800. Honestly, that is what you expect, by the time the hardware is molded and branded. (ofcourse there is always the OS licence.)

I'd be interested to see what kind of a UMPC can be made using arm and a linux os, in a similar vein to what GPH has done. A UMPC at less then $400 would certainly be a sensation.
 
Last edited by a moderator:
Epicenter posted on Aug 26 2006 at 09:23 PM said:
... It isn't magic. It's called OpenGL rendering. Have you ever used a PSX emulator for a PC? All the most successful ones that provide the best performance give you the option of drawing 3D with an actual hardware accelerator through OpenGL or Direct3D APIs. These are standard 3D APIs that operate on any compliant hardware. Such rendering engines would also work on the OpenGL-compliant PowerVR MBX Lite chipset in the S3C2460, or the OpenGL-ES compatible 2D/3D Accelerator in the XGP/XGP Mini.

Well yeah but on the PC you have super fast graphics cards and a super fast CPU to interpret these 3D instructions. The XGP etc is very weak compared to PC graphics cards, even bad ones.

But yes having a 3D GPU will definetly help alot but as I said it won't give perfect emulation still. I don't really care anyway. I am not going to play old PSX games on a handheld. It is impractical to me to put 2 games on a card and fill it. Also if I am going to play 3D I will go to the Xbox 360 or PS3 when it comes, that is just me I suppose. To me these handhelds are for emulation of old 2D games and homebrew.
 
Last edited by a moderator:
DaveC posted on Aug 27 2006 at 01:09 AM said:
Well yeah but on the PC you have super fast graphics cards and a super fast CPU to interpret these 3D instructions. The XGP etc is very weak compared to PC graphics cards, even bad ones.

But yes having a 3D GPU will definetly help alot but as I said it won't give perfect emulation still. I don't really care anyway. I am not going to play old PSX games on a handheld. It is impractical to me to put 2 games on a card and fill it. Also if I am going to play 3D I will go to the Xbox 360 or PS3 when it comes, that is just me I suppose. To me these handhelds are for emulation of old 2D games and homebrew.

You really don't seem to understand, yet. The CPU in the GP2X is already interpreting all of them. Drawing the whole scenes in wireframe would be no problem. But things like fill-rate, texturing, all of these more stressful 3D tasks are what drag down GP2X emulation of the Playstation. Translation of PSX graphics instructions to OpenGL equivalents is already being handled just fine; and the GPU isn't the component to do it. It didn't take a powerful GPU to aid emulation of the PSX, either-- the PSX's 3D rendering hardware is weak. Extremely weak-- it was an 'acceptable' rendering solution for 1995. The XGP has already demonstrated its ability to rank with at least 1998's midrange class of PC 3D accelerator with its Quake III performance (Seems a bit behind the 3DFX Voodoo 2), which, while lackluster, demonstrates acceptable framerate in a render job that far exceeds the capability of the PSX. Even Quake II struggled to run acceptably on the PSX, and it was scaled back a LOT to even be accomodated on that platform. (By the way, PSX emulators with OpenGL acceleration have run great on hardware far, far older than what we use today. I've used PSX emulators at the full 60 FPS using mediocre accelerators over 6 years old.)

Now, that's discussing the XGP's GPU's capabilities. Let's look at the S3C2460's graphical accelerator, PowerVR's MBX Lite's specs.

- Fill Rate on the Playstation 1: 2.4 Million pixels/sec.
- Fill Rate on the PowerVR MBX Lite: >135 million pixels/sec.
- Shaded/Textured Polygon rendering capability on Playstation 1: 360,000 polygons/sec
- Shaded/Textured Polygon rendering capability on PowerVR MBX Lite: 1,700,000 polygons/sec

If you're still going to be dense about this, well, I doubt there's much more I can say. The facts are in front of you; decide for yourself.
 
Last edited by a moderator:
All in all, core emulation is CPU intensive. Going for a 66 MHz increase on the same arm9 architecture isn't going to prove much other then making PSX more playable and other systems look more pretty.

N64 will never be playable as it has to emulate a 94 MHz MIPS R4300i-series 64-bit RISC CPU. This isn't even taking into account the video card.

I would prefer a more substantial CPU increase with a simple FPU to ensure that Amiga can be playable. NeoGeo can be played without overclocking and all the other pre 3D systems systems work fine.

ARM11 looks really good (with FPU), and it has been around for a long time. I'd be happy to wait for a good chipset to be released at a reasonable price and then have GPH jump all over it.

Both options available now are flawed for emulation in my opinion. I mean, the 3D one looks fun. The powerVR is the same architecture as the Kyro, or the GPU used in the dreamcast. But OpenGL and OpenGL ES are completely different, so its not like we will be porting old OpenGL classics in no time. However, it would be great for homebrew.
 
icurafu posted on Aug 27 2006 at 06:47 AM said:
All in all, core emulation is CPU intensive. Going for a 66 MHz increase on the same arm9 architecture isn't going to prove much other then making PSX more playable and other systems look more pretty.
The point is that CPU emulation is covered already in PSX Emulation; that's no longer the bottleneck. The huge slowdown results from trying to do hardware rendering on a CPU that's not suited for it, with no hardware support whatsoever.

icurafu posted on Aug 27 2006 at 06:47 AM said:
N64 will never be playable as it has to emulate a 94 MHz MIPS R4300i-series 64-bit RISC CPU. This isn't even taking into account the video card.
The idea is, free up the CPU from the 3D rendering tasks by offloading that to the GPU; removing the video 'card' from the equation. Being able to emulate the CPU at full speed is unlikely, but playable speeds could probably be obtained. I would expect them to get about as far as PSX will on the GP2X (Decent but not really playable in most instances.)

icurafu posted on Aug 27 2006 at 06:47 AM said:
ARM11 looks really good (with FPU), and it has been around for a long time. I'd be happy to wait for a good chipset to be released at a reasonable price and then have GPH jump all over it.
ARM11s are available, e.g. in TI's high-end OMAP2 series. But they're still very expensive. At any rate, the 'option 1' SoC provides an FPU and Vector processing unit (AFP9).

icurafu posted on Aug 27 2006 at 06:47 AM said:
I would prefer a more substantial CPU increase with a simple FPU to ensure that Amiga can be playable.

From what I recall, Amiga emulation is slow due to the complexity of the video subsystem; also existing emulators like UAE4ALL implement this functionality in slow C code, not ARM9 ASM. An FPU would probably help, but it won't ensure playability on all games.
 
Last edited by a moderator:
Thanks for taking the time to answer; I'm aware of TIs OMAP2. They are used in Sony's P990 (and w990... or whatever the new walkman variant is called) and a few Nokia phones using the latest Symbion OS, but I'm not 100% on that one. I notice it has the ability to execute java bytecode.

Actually, ARM10 would be good enough. Have you seen any reference boards for ARM1020E? I've seen some benchmarks and they seem to perform extremely well compared to their power consumption. I know Samsung do them on 0.13, but there are no specific details.

A 350 MHz - 500 MHz ARM10 would be a much more suitable board for option 2. But I guess that's down to availability. I don't think anyone wants to go through the non-FPU route again.

I'm happy to go with option 1 (I even voted for it); I think it opens up a lot of interesting options. But mixing a PowerVR with a 266 ARM9 CPU won't allow you to take full advantage of a video card you’ve probably paid a lot of money to have included in the solution.
 
all I know is it seems that GPH needs to address these issues:

1. performance to fully support 32 bit system emulators like psx
2. better battery performance. maybe a way to choose between double As or a lithium ion pack. that would be really cool. I think sticking with double As is the best bet here. lith ion should only be optional.
3. there absolutely must be a good interface for connecting gp2xs together. some kind of common way to communicate. the best way would be through a usb port I suppose.
4. more than one usb port.
5. price needs to be no higher than 250. imo.
6. a touchscreen would be amazing. just as long as it didnt shoot the price up too much.
7. more than one sd slot would be cool.

I dont know about a HDD.... heavy wieght comes to mind. dont want hand fatigue due to the overall wieght. I really like the sd thing it has going.

so thats it in a nutshell. better hardware performance, better battery performance, and some good usb ports so we can throw in some nice accesories like wireless usb sticks :D wireless gp2x play between 4 players all playing smash bros now that sounds like Im dreaming here. heh ^_^
 
Back
Top