LOSTCIRCUITS

Summary

3DLabs have been a leader in professional graphics for quite some time. Names like Oxygen and even the Permedia3 still turn some heads, not to mention the 3DLabs Wildcat series. Today's review focuses on the current flagship, that is the Wildcat VP990, equipped with no less than 512 MB on board DDR graphics memory.

Cinematic quality rendering has become a major part of the professional graphics market and it does require different hardware configurations than just a standard desktop card, even if it is modified to optimally fit the professional sector. Which part of the market sector does the 3DLabs Wildcat VP990 fit into and what, aside from adding extra memory has changed since the acquisition of 3DLabs by Creative Labs?

I believe, we have some really interesting findings to report

3DLabs Wildcat VP990; click for larger image.

The big guns in that particular market have historically been FireGL (now under the umbrella of ATI), nVidia with their Quadro series and 3Dlabs. In the last two years, nVidia has become a very dominating force in the professional sector, especially with OpenGL applications such as used in SPEC.org's ViewPERF benchmark suite. The countermeasure has been a limited alliance between ATI and 3Dlabs at least to the point of stopping some of nVidia's "tour de force" approaches like overrunning the market with proprietary shader languages like Cg.

The high-end professional graphics market is not a homogeneous entity either, there are OpenGL CAD applications that thrive on triangles / second which, combined with high precision of the rendered images in OpenGL is more the classic high-end market. A relative newcomer to that scene is the realtime rendering of large scale models for computer-generated images and cinematic real time rendering, which gains increasing foothold and importance in the movie industry. Contrary to some gaming applications, still following some stereotypical patterns, it is the absence of such patterns that really makes cinematic quality what it is.

Absence of patterns, in turn, requires a high number of independently functioning programmable subunits or shaders, rather than the somewhat limited flexibility of register combiners, a term used by 3Dlabs for the first generation pixel shaders developed by nVidia and ATI. That was awhile ago, though. Another requirement for independently rendered subunits is a huge memory space available in form of a unified frame buffer with flexible "on-the-fly" allocation of the overall memory space to hold vertex OR texture data as needed by the application.

So how are those differences in architecture put into real hardware? No worries, we won't go into gory detail, it'll be a brief description of the key features only in the following and then we'll concentrate on what is on every man's mind in the non-Freudian sense (even though "performance" has become a somewhat tainted word as of lately).

Next Page:    => The Paradoxical Term of "Visual Processing Unit" =>