More Graphics

What does the video RAM do?

Video RAM is important for storing textures and for the frame buffer. So, the more memory you have, the more textures you can store in memory, and the higher resolutions you can use. While 32MB of RAM is becoming popular on PC 3D accelerators, the Dreamcast is left with a fixed 8MB of RAM. However, as you probably guessed the Dreamcast has a few tricks up its sleeve.

Vector-Quantization Texture Compression

The Dreamcast offers VQ Texture Compression which has similar compression performance to S3TC. The 4:1 average compression ratio of VQ texture compression allows Dreamcast developers to use approximately 32MB of texture RAM, on par with a PC. This is another case of the console development environment saving the hardware. Besides S3's entire current product line-up, only ATI's Rage 128 Pro and Nvidia's GeForce 256 supports DirectX Texture Compression. With so many other platforms on the PC that do not support DXTC still on the market such as the venerable TNT2 and G400, most developers are not using this feature in their games. Unreal Tournament for example, has S3TC textures on the second CD, but not DXTC. If PC games started using DXTC, then PC 3D cards would be able to hold over 100 MB of textures.

Lower resolutions on the Dreamcast

A triple-buffered 1024x768 frame buffer with 32-bit color depth takes up more than 8MB of RAM. In other words, even if performance were cast aside, the Dreamcast cannot output at that resolution. However, as noted earlier, the Dreamcast's target monitor is an NTSC television and so the DC only has to worry about 640x480 resolution. At 480 lines of resolution and with a SVHS cable, a Dreamcast can boast similar image definition to DVD.

What else does the DC do to "cheat"?

Random Fact: Some games like Messiah dynamically adjust the polygon detail based upon system speed. In other words, if your system is fast enough, you'll be seeing things the developers have never seen themselves.

Lower Color Depths on the Dreamcast

The Dreamcast does not need to run at 32-bit color and can successfully get by with 16-bit color again thanks to the TV and a bit of PowerVR trickery. By now, half of you are ready to send a colorful email my way calling me blinded by 3dfx marketing. Don't get me wrong, I'd never give up 32-bit color depths on the PC, and before you send your email, please keep reading. 16-bit artifacting shows up in two different cases, first when significantly multi-texturing and in the "cheese-cloth" dithering that seems to cover the whole screen.

The biggest problem with 16-bit color depths is undoubtedly the multi-texturing artifacts. Each time another texture is blended onto the surface, the errors are magnified in the same way a VHS tape would look if you made a copy of a copy of a copy. The magnitude multiplication of the multi-texturing artifacts happens each time to write to the frame-buffer and here the PowerVR's unconventional architecture is the Dreamcast's saving grace. As a scene-based renderer, the PowerVR only writes to the frame-buffer once - the very end; all of the texture blends are done at the same time internally. Thus, the artifacts you get with quad-texturing are no more than the artifacts you get if you were only single texturing.

The "general" haze associated with 16-bit color is avoided on the Dreamcast because you sit further away from a TV than you do with a monitor and so the TV phosphors blend the colors together.

What's the verdict?

If we were comparing the PowerVR Series2 chip against a GeForce 256 or Rage Fury Maxx, then the PowerVR would lose…by a lot. However, we are considering the Dreamcast as a unit as a whole and in this case, the Dreamcast is capable of the same equivalent fill rate performance as today's PC graphics accelerator. The PC may have more RAM available to the video card, but again the Dreamcast enjoys comparable performance thanks to a few special circumstances.