Details Regarding The Xgp 2d/3d Accelerator

[Epicenter]

After a bit of research, I've found claims by Magiceyes of "1 million polys/sec." performance on the 2D/3D accelerator, as well as of course OpenGL-ES support. After a bit of research on common ARM platforms, I found the primary 2D/3D graphics acceleration avenue on the official ARM platform is Imagination Technologies' PowerVR MBX series. Checking the specifications of the MBX lineup, the chip in line with this render-speed claim would be the MBX RS or "MBX Lite".

I dug up some demonstration videos of some of its capabilities-- it's quite robust indeed. In fact, it's the same core utilized in the Samsung S3C2460 I've sung praises of (and suggested GP/GPH utilize in future handhelds, which GP made known to me they intend to do.)

3 demo videos of a torso model with some impressive environment mapping, a pan shot around a stonehenge-lookalike with a skybox and some shots of WW2-era planes flying around a high-poly environment and skybox, all at extremely high framerate:
http://www.imgtec.com/Downloads/DocumentDo.../FixedPoint.zip

A very high-framerate bump-mapping test with complex light and shadow n a high-poly model.
http://www.imgtec.com/Downloads/DocumentDo.../EXE/Virgin.zip

A heavy particles test with a very high framerate.
http://www.imgtec.com/Downloads/DocumentDo...E/Particles.zip

A test of rendering a sphere with 8 moving colored lights.
http://www.imgtec.com/Downloads/DocumentDo...XE/Lighting.zip

A demonstration of mip-mapped, trilinear and standard texture filtering on 3 moving spheres and a high-poly moving environment.
http://www.imgtec.com/Downloads/DocumentDo...E/Trilinear.zip

Judging by the very high framerate of these tests, the hardware is clearly not working anywhere near maximum capacity. They appear to be running on Dell Axim x50v hardware with the MBX RS/Lite. This platform is similar to the intended XGP hardware, with the same accelerator coupled with an XScale processor at 624 MHz, roughly equivalent to a pure ARM9 processor at 324 MHz. (CPU involvement in these tests is relatively minimal.)

Further research suggests slowdown in Quake III may not be fully graphically related as a highly floating-point heavy binary seems to be in use-- not a modifed fixed-point version. Tests of Quake III on PowerVR MBX Lite-based chipsets have reportedly achieved 30fps using bilinear texture filtering and relatively low graphical intensity. Quake I/II performance should be extremely high, I would expect around 60 FPS operation with a predominantly fixed-point binary.

Naturally, 2D performance should be extremely impressive given the GPU's apparent 3D prowess. Even meager 3D accelerators like that utilized by the original Playstation managed to render hundreds of thousands of sizable sprites per second, all in hardware-- no need to burden the main CPU. This is a task the ARM9 processor series on its own is not capable of at acceptable framerate, though what it can achieve on its own remains wholly impressive, what can be accomplished by relegating graphical tasks to the GPU and leaving the ARM9 chip free to process only game logic, music decoding, input, etc. will provide exceptional gains in functionality and performance for not only development of homebrew applications, but acceleration of emulators and ports of 2D and 3D games.

Definitely looking forward to these little machines...

 

[washo]

I remember viewinf those demos (virgin....) on a Dell Axim X50v
It looks great (on a 640*480 screen)

see you :lol:

 

[Epicenter]

It should be much higher-framerate, and use less RAM for the framebuffer at 320x240.

 

[Vimacs]

it uses the vrender3d chip of mse afaik.

 

[aapje89]

very impressive!

I look forward to playing QIII on a handheld

 

[Epicenter]

it uses the vrender3d chip of mse afaik.

Yep, VRENDER-3D. I get the impression that's not fully set in stone after talking with GP, and they've made statements before they are considering changing the "CPU and GPU" selection (which would of course entail purchasing an entirely different SoC.) I have a feeling they stick with the VRENDER-3D though since they've sort of advertised for Magiceyes already in their demonstrations.

 

[aapje89]

it uses the vrender3d chip of mse afaik.

Yep, VRENDER-3D. I get the impression that's not fully set in stone after talking with GP, and they've made statements before they are considering changing the "CPU and GPU" selection (which would of course entail purchasing an entirely different SoC.) I have a feeling they stick with the VRENDER-3D though since they've sort of advertised for Magiceyes already in their demonstrations.

Is that a good or a bad thing?

 

[Epicenter]

Well it's not a bad thing. There are better chips they could be using than that. Mostly because of MES' shitty customer service and proprietary design. They can take open hardware and shut out the means by which to utilize it to good-paying customers, and that's just fucked up.

 

[Vimacs]

The good thing is that it will probably be easy to get gp2x stuff running, the "render video as texture" effect of the vrender3d is also quite nice.

 

[Epicenter]

Especially if they stick with Linux, SDL applications could be brought over with extreme ease, my own works included. I'm encouraging them to do just that.. not too thrilled if they offer only a 'GPOS' (G.. POS?)

 

[Vimacs]

let them offer what they want, it shouldnt be to hard to port the gp2x linux over.

 

[DaveC]

Especially if they stick with Linux, SDL applications could be brought over with extreme ease, my own works included. I'm encouraging them to do just that.. not too thrilled if they offer only a 'GPOS' (G.. POS?)

Then convince them to put in a d-pad intead of that crappy analog stick. It seems that everyone is so interested in ditching the GP2X ASAP for this widescreen wacky resolution screened thing for some reason. emus will be shit with an analog. It will be bad enough that they will be all stretched to shit (ICK) but using an analog controller will just make bad things worse.

 

[JaqMs]

DaveC, I promise to stay out of the GP2X forum if you stay out of this forum.

 

[Alex.]

DaveC, I promise to stay out of the GP2X forum if you stay out of this forum.

Hahaha, this made my day Cheers

- Alex

 

[Dzz]

Especially if they stick with Linux, SDL applications could be brought over with extreme ease, my own works included. I'm encouraging them to do just that.. not too thrilled if they offer only a 'GPOS' (G.. POS?)

Then convince them to put in a d-pad intead of that crappy analog stick. It seems that everyone is so interested in ditching the GP2X ASAP for this widescreen wacky resolution screened thing for some reason. emus will be shit with an analog. It will be bad enough that they will be all stretched to shit (ICK) but using an analog controller will just make bad things worse.

To be fair, I don't think anybody has said they are going to "ditch" their gp2x asap for the xgp (which isn't the same thing as "everybody"). I personally am interested in all machines, and the possibilities of an open platform with a nice graphics accelerator are worth having a look at. You're probably right that some emulators won't be any better on that machine, but not everybody is too concerned by that. Emulators could just not use the sides of the display if desired -- the fact that old TV shows are in 4:3 doesn't stop people from making do with widescreen televisions -- they stretch it if they don't mind stretching and leave some of the screen space wasted if they do. Yes, the 272 pixel tall resolution also introduces 32 pixels of new dead space for those systems that are naturally 240 pixels tall. Those things mean that certain emulators will have smaller displays on the xgp than on the gp2x, which is certainly a disadvantage. However, i think movies are likely to be better on the xgp (assuming they work well from a software perspective), and the bigger screen might be better for homebrew games of the type I like to make.

Having said that, it does not mean i'm ditching the gp2x asap, just curious and open minded. I think the market potential of this machine is rather dismal but I could be wrong.

 

[Epicenter]

I wouldn't be ditching the GP2X at all; most of my emulators would look pretty crappy skewed to fit a 16:9 screen correctly. However the XGP is still a n exciting development platform-- DaveC, you'd understand the desire to see your hard work's true potential on more powerful hardware if you'd done any sort of programming. New platforms and hardware let you get away with newer and more amazing things, building on what the GP2X already can do.

Also, yes, I am trying to convince GP to replace the full XGP's control scheme with a d-pad.

 

[DaveC]

Also, yes, I am trying to convince GP to replace the full XGP's control scheme with a d-pad.

After you get them to ditch the analog see if you can get them to get rid of the 16:9 weird 480 x 272 rsolution screen and the proprietary battery.

 

[JaqMs]

Ok, so say GP changed the screen to 320x240 and the batteries to AA, what are you going to do? For YOU (disregarding any technology that does not relate to emulation), it would pretty much be like buying a GP2X again but with a D-pad.

 

[OMars]

Ok, so say GP changed the screen to 320x240 and the batteries to AA, what are you going to do? For YOU (disregarding any technology that does not relate to emulation), it would pretty much be like buying a GP2X again but with a D-pad.

I would definitely be in!

 

[DaveC]

Ok, so say GP changed the screen to 320x240 and the batteries to AA, what are you going to do? For YOU (disregarding any technology that does not relate to emulation), it would pretty much be like buying a GP2X again but with a D-pad.

That is the idea. Then devs will be leaving the Gp2X for a clone of the GP2X. So this way I can't loose.
If many stop developing for the Gp2X to go to this now and it has the fucked up screen, analog, and non replaceable battery then I would be forced to play new emus on a shitty stretched out (or small emu image) screen with a sloppy analog. Since you can't d-pad mod an analog that would suck.