OMars posted on Sep 7 2006 at 06:28 PM said:
With these specs:
-320x240 res screen(atleast 3")
-250+mhz processor with gpu
-64MB RAM
-SD slot
-USB host
-uses two AA batteries
-lots of NAND memory
...
EDIT: I have taken the acceleration part out because craigix believes it would drive the cost up.
OMars posted on Sep 7 2006 at 06:51 PM said:
craigix posted on Sep 7 2006 at 06:47 PM said:
What would you consider cheap bearing in mind that 3d hardware will push the price up beyond PSP prices as GP will have to make a profit on the hardware unlike sony etc.
Ok, take away the "2D/3D acceleration" part. This would still be the best system.
Without 2D or 3D acceleration, it's just a GP2X with some mild overclocking. Even with a 300/400 MHz processor, as Craig and I discussed, not that much will really improve. Sure, you'll get better performance on some tasks, but most of the things the GP2X cannot do well require extreme 2D or 3D graphics operations that just don't go over well on an integer-only ARM9 processor which wasn't designed for such work.
What kind of things are those, you ask? Look at SquidgeSNES. It has no difficulty emulating the CPU and the audio is little problem; but the graphics core is very very slow. Rewriting it entirely in ARM9 ASM would be very, very difficult and take an extremely long amount of time. If the work of handling say, transparences, blitting of tiles and potentially even scaling and rotation were offloaded to a 2D GPU, performance would be dramatically higher.
PS1 emulation. The GP2X already has all the horsepower needed to emulate the PS1 CPU, which is the meat-and-potatoes of the emulation. However, it cannot by any means handle rendering a full frame with as many polys as the PS1's game library throws at it. PC emulators in the days of sub-600 MHz processors would never expect the main CPU to software-render the scene AND emulate the main CPU, that's quite a task-- instead, the most successful ones used OpenGL rendering on the graphics hardware of the machine to natively render a scene, interpreting the original graphical data from the PS1 game and translating to simple drawing instructions a PC's GPU understood. This is the same principle which would be utilized emulating the PS1 with a true 3D GPU, and simple to do since the PowerVR RS ("MBX Lite") in the MagicEyes VRENDER-3D or the Samsung S3C2460, whichever Gamepark decides on, has full support for the commonplace OpenGL-ES standard, and is over 7 times as powerful as the PS1's 3D hardware, as well as offering multiple features the PS1's polygon rendering engine didn't offer. There is no logical reason
Further benefits are that, also through the OpenGL-ES API, homebrew game developers can draw not a couple of hundred tops, to THOUSANDS of sprites on the display in a second or potentially in a FRAME, for incredibly immersive games without slowdown, as the main ARM9 core is not being burdened.
My reason for railing so strongly for the Samsung S3C2460 is that it offers not just the main ARM9 CPU, but also an AFP9 Floating Point / Vector processor that is nearly as powerful as the GP2X's ARM920T! Using a mix of integer and floating point math in a program equitably, this pair of processors can operate as quickly as a chip of nearly twice the clockrate without writing any special code like that to utilize the 940T, working in the same memory space, transparent to the user or even the programmer. Floating point support will supplement 3D acceleration in making ports of PC games like Quake I/II much faster (Quake I contains nearly NO integer-based code!) run far smoother-- 60 FPS would be no problem at all-- After speaking with Zodttd, he also stated he could utilize the AFP9 for emulation of the Geometry Transformation Engine (GTE).
The performance level of a ~500 Mhz ARM9 with proper utilization of integer and floating point processing units, plus powerful graphical acceleration is the only logical course of action, and makes a single high-clocked ARM9 (especially just 250 Mhz) and no acceleration a fairly ridiculous proposition-- especially when such chipsets do not cost nearly as much as they used to. Broadening the horizons of what can be done with a machine like the GP2X is much more important than putting the same old thing in a new case with a little incremental speed boost and calling it a day.
(My apologies for all who read this far, my little rant ran a little longer than originally anticipated. :lol: )
