|
|||||||||||
|
|||||||||||
|
|||||||||||
| This board,
featuring a unique ASIC design, struggled mightily when
used with our slower Pentium 100 machine, but improved
significantly in our Pentium 200. The Apocalypse 3D's architecture relies on a technique known as "hidden surface removal," which means that, for each frame, the objects hidden behind other objects are not drawn (like when a monster suddenly comes up to you and blocks, say, the wall behind him). Although other boards also use this technique, Apocalypse 3D's method is unique. With the standard z-buffer and hidden surface removal present in other ASICS, the chip has to go out to memory for each pixel in order to find out whether or not to place it in the frame buffer. Basically, the other boards look at each polygon, then check the z-information of each pixel within that polygon. With PowerVR, this process is basically reversed. By looking at the pixel itself first, then evaluating the depth of the polygons, the PowerVR chip doesn't need extra memory for a z-buffer nor extra memory accesses for z calculations. This cuts down on memory bandwidth, as a z-buffer is not necessary (saving 1 bit for each pixel). Unfortunately, fancier triangle set-up (a job done by the main system's processor) is necessary as the PowerVR ASIC simply does not take care of this whole process by itself. This means a slower computer just won't perform as well, even on a relative scale. During our tests, this problem was drastically apparent. When using our Pentium 100 (the minimum system requirements VideoLogic recommends), the board performed very poorly, particularly with Direct3D operations, and specifically with the fill portion of our Direct3D testing. Running Monster Truck Madness was pretty awful as well, with the board clearly not able to keep pace with the graphics updates necessary for smooth play. |
|||||||||||
| Even the games ported directly to the board, such as MechWarrior 2 and Ultimate Race, fared only a bit better. Only WipeOut, a game we set to have very little texturing (but a great number of polygons) played smoothly. | |||||||||||
| Using a
Pentium 200 was an entirely different story. To be sure,
all the boards perform better with twice the system
processing. But the difference with the Apocalypse 3D was
far more dramatic. Most clearly, the fill rates rocketed.
For all other boards, the doubling of the system processor has hardly any impact on fill rates, as that is a board-intensive operation. Basically, the Apocalypse 3D went from the bottom of the barrel, in terms of overall Direct3D performance, to being competitive with our very best products. Naturally, gameplay improved dramatically right along with it. |
|||||||||||
| Like the Voodoo-based products, the Apocalypse 3D is an add-in board, so if you're happy with your current graphics card and want to save $100 off the retail price (when compared to the Diamond and Orchid Voodoo-based products) - and most importantly, you have at least a 166-MHz Pentium - this board is an excellent choice. Slower CPU owners, stay away. | |||||||||||
|
|||||||||||
|
| |||||||||||