View Single Post
Old 25 February 2017, 16:59   #35
Join Date: Jan 2010
Location: Kansas
Posts: 1,284
Originally Posted by Anakirob View Post
No 3D support? Are you sure about that? I'm pretty sure line drawing and flood filling are features of the Agnus chip, which would mean that there IS in fact 3D support features in the Amiga's chipset.

Ok, so it's nowhere near as complex as more modern 3D hardware, and textures are out. But polygons are polygons and 3D is 3D and the Amiga does in fact support this at the hardware level all the way back to the Amiga 1000.
Shall we say the Amiga custom chips had no 3D specific support then? Copper chunky was low resolution 2x2 or worse and improved with AGA but only compensated for part of the handicap of not having native chunky modes. The blitter could be used to fill 3D objects which was useful until texture mapping arrived. The limits of these techniques was reached somewhere between Frontier Elite and Wolfenstein.

Originally Posted by Anakirob View Post
And as for no chunky support. The CD32 does feature C2P conversion hardware, so obviously this feature was on Commodore's engineering team's radar. And if I'm reading the information correctly the Hombre would have had native chunky screen modes.
The CD32 C2P support required passing arrays of pixels through a register in hardware which is an extra step which is unnecessary with native chunky. It provided a speedup with a slow CPU (and no fast memory) but may have been a bottleneck with a fast CPU. It is lame support considering all the advantages of chunky and how much of the computer market was using it for higher color displays. It was not even that difficult to add as can be seen by the new FPGA Amiga chunky modes. From what I have read, the Hombre would have *only* had chunky modes. It was planned to be capable of being used as a gfx enhancement for the Amiga (Zorro III card?). Perhaps it would have overlayed the AGA or simply passed through AGA modes like the original CyberVision gfx card. The original 3DFX Voodoo 1 and 2 were 3D accelerators and did not include 2D support which came from the PC. Perhaps later it could have all been integrated but there needed to be plans to increase performance of both the 68k and 3D or they would have fallen further and further behind in competition with PC gaming (maybe nice embedded setup for set top boxes and kiosks though). We can already see that the 3D performance of a PA-7150 RISC CPU with SIMD was unimpressive. Double the clock speed of the 68060 and it could outperform the PA-7150 gfx accelerator. Add a nice SIMD and the performance could be doubled again. The 3D accelerator can turn into a 3D decelerator as happened with the S3 VIRGE (Cybervision 64/3D) and high performance PC CPUs.

Originally Posted by
When performing basic 3D-rendering with only texture mapping and no
other advanced features, ViRGE's pixel throughput was somewhat faster than the best software-optimized (host-based CPU) 3D-rendering of the era, and with better (16bpp) color fidelity. But when additional rendering operations were added to the polygon load (such as perspective-correction, Z-depth fogging, and bilinear filtering, rendering throughput dropped to the speed of software-based rendering on an entry-level CPU, which was unacceptable to most gamers. Due to this abysmal 3D-rendering performance, the ViRGE earned the dubious distinction of being the world's first "graphics decelerator."
The S3 Virge provides 3D acceleration on most classic Amigas because they have slow CPUs and it provided chunky (saving the C2P) while offloading the Amiga custom chips from stealing cycles from the CPU at higher resolutions and colour depths.
matthey is offline  
Page generated in 0.05571 seconds with 9 queries