Originally posted by know-it-all-wanna-be
Does anybody knows the spec of the Noami 2?
I would like to compare it to ps2.
I think why arcade cost thousand of dollars is because they have lots of RAM memory...
CPU CPU: 200 MHz Hitachi SH-4 (SH7091)
CPU Memory: 32 MBytes (100 MHz SDRAM)
CPU Memory Data Path: 64-bits
Memory Bandwidth: (800 MBytes/sec)
Geometry Co-processor Geometry Co-processor: VideoLogic custom transformation & lighting (T&L) chip (Code named: ELAN)
Clock Rate: 100 MHz
Sustained Polygonal and Lighting Rate: 10 million polygons/sec with 6 light sources!Supported Lights: Ambient, parallel, point and spot
Vertex Support: Combined dynamic and static model processing
Geometry Memory: 32 MBytes
External Memory Data Path: unknown
Memory Bandwidth: unknown
Graphics Processing Unit (x2) GPU: Two PowerVR2 (CLX2)
Pixel Fill-Rate: 200 MPixels/sec (400 MPixels/sec to 600 MPixels/sec due to infinite plane architecture assuming depth overdraw complexity of 2 to 3 layers)
Graphics Memory: 2 x 32 MB (100 MHz SDRAM) External Data Path: 64-bits per GPU (128-bits total)
Memory Bandwidth: 800 MB/sec per GPU (1.6 GB/sec total)
Graphic Effects: Polygons/strips/fans engine, 16-bit and 24-bit color, multiple fog modes, super sampling for full scene anti-aliasing, specular highlighting, texture filtering: bilinear, trilinear, anistropic, MIP mapping, bump mapping, perspective correction, 8-bit alpha blending (256 levels of transparency), ARGB gouraud shading, general modifier volumes (GMV) for such effects as shadows, light, transpararency, etc.
--------------------------------------
Geometry Coprocessor --------------------------------------
Naomi 2 has a dedicated geometry coprocessor to handle transformations and lighting which is rated at 10 million polygons per second with 6 light sources. Note that the T&L processor is not limited to 6 lights, as a maximum of 16 lights per polygon can be achieved, but with a reduction in the polygon rate. The geometry chip will offload all T&L calculations previously performed by the 128-bit matrix math unit on the SH-4. The SH-4 will now be free to devote more of its resources for physics, artificial intelligence, collision detection and overall game code. The hardware T&L unit features combined dynamic and static model processing, and multiple light type support (ambient, parallel, point and spot).
Almost all T&L processors on the market never state what their polygon rate is with the number of light sources present per polygon, and the reason why, is because the polygon rate goes way down with more light sources, with the current T&L processors on the market. T&L should always be rated with number of polygons with number of light sources present. Note: the lighting information for a polygon does not have to be related to a light source, as light information can also be used to make an object look more realistic. Like trying to make plastic look like plastic in a game.
Dual Graphics Chips Two PowerVR2 (CLX2) GPU\'s with 32 MB of memory each, which is twice the amount that the PVR2 GPU had on the Naomi 1 board. Each chip renders half the screen (rectangular, stripes, and checker board options), so game textures have to be repeated in both local memory pools, but the display list (infinite plane) data covers only the area of the screen that each GPU has to render.
http://www.segatech.com/arcade/naomi2/index.html