3Dlabs Permedia 3 Create!

Textures vs Polygons: What counts

Most performance gamers have heard the term texture at least once in their gaming "career" and the general rule of thumb is that a video card capable of processing textures faster results in faster gameplay. Unfortunately for the target market 3Dlabs is aiming for with the Permedia 3, texture fill rates are not the most important factor. Before we get into what the more important factor is, let’s compare the old Permedia 2 to what today’s gamers have on their wishlists.

For starters, the Permedia 2’s core was not capable of rendering multiple textures in a single pass, a technology rightfully dubbed, single pass multitexturing. The release of the Permedia 2 came at a time when single pass multitexturing was not an important issue, however with the release of Quake 2, and thus games based on the Quake 2 engine, single pass multitexturing became the buzz word that drove most gamers to upgrade to the next wave of 3D accelerators.

Professionals come to learn the exact opposite, that the key to a high performing system for the type of work they do (i.e. 3D rendering, etc…) isn’t an amazing fill rate, but a core capable of processing a lot of polygons. For high-end applications, the construction of the scene to be rendered is far more important than how the final product looks. The final production stage in any 3D animation takes place after the initial wireframes and the scene itself is constructed, for this reason a high fill rate isn’t as important as a high polygon throughput. In order to produce a cost effective product, a professional graphics card generally offers a more favorable polygon rate rather than a higher texture fill rate. On the flip side of the coin, a gamer’s solution generally offers a greater texture fill rate at the cost of a lower polygon rate. What this essentially means that the Achilles’ heel of all professional cards is that they can’t play games and likewise, the Achilles heel of all gaming cards is that they can’t be used very well in a professional environment.

Whereas fill rate depends on the speed of the graphics processor itself, polygon rate depends on the power of the CPU of the system the graphics card happens to be installed in. For the sake of simplicity we’re forgoing inclusion of all graphics accelerators with on-board geometry processors. On-board geometry processors, if you’ve ever heard the term before, essentially take the load of geometry processing off of the CPU and place it on a processor on the graphics card. These geometry processors are generally reserved for the most expensive of professional graphics accelerators, and thus drive the price of these accelerators beyond the cost of the most expensive desktop CPUs. For the professional by day, gamer by night market, a thousand dollar graphics card is usually out of the viable range, therefore making the more conventional geometry processor equipped solutions not in direct competition with cards such as the Permedia 3 which has no on-board geometry processor.

In order to better tailor to both gamers and professionals, the Permedia 3 attempts to establish a powerful balance between polygon throughput and fill rate. The Permedia 3 does support single pass multitexturing which is already a huge improvement over the original Permedia 2. By understanding the requirements for a professional card, or any graphics card for that matter, to excel in a high-end environment you’ll be prepared to accept the results of the high-end tests. Only in a high-end environment can ATI beat out 3dfx and NVIDIA with a year old Rage 128, sound like a stretch? It gets even better…