VideoLogic is a name few people would now instantly associate with graphics accelerators. In fact, its parent company, Imagination Technologies, was present at the birth of the PC 3D accelerator field in June 1996 with its PowerVR card. This was a close competitor to the Voodoo from 3dfx, and while the 3dfx brand has become synonymous with graphics accelerators, PowerVR has taken something of a back seat.
The Vivid! is VideoLogics latest graphics accelerator, based around the PowerVR Series 3 KYRO chipset. This latest design is PowerVRs attempt to reassert itself into the graphics card market, but with huge competition from ATIs Radeon DDR and nVIDIA with its range of GeForce2-based cards, VideoLogic has a giant mountain to climb.
Prior to the series-3 KYRO chipset, VideoLogic had released a series-2 PowerVR-based chipset, which was used in the Neon 250 (reviewed issue 63, p174). Despite being a fundamentally solid graphics accelerator, the Neon 250 was released too late and fell foul when compared to rivals of the time, such as the TNT2 chipset from nVIDIA.
An initial glance at the specification list for the VideoLogic Vivid! evoked several feelings of deja vu. Some features were most notable by their absence, namely DDR memory and hardware Transformation and Lighting. The Vivid! also uses a relatively slow 115MHz chip clock speed, significantly slower than the 175MHz clock of the similarly priced Hercules 3D Prophet II MX (reviewed issue 37, p95).
On the plus side, the Vivid! features environment bump mapping, which has recently been gaining much needed support from software developers.
Given the relatively modest specification, I wasnt expecting the Vivid! to match high-end cards, such as the Radeon DDR. Performance benchmarking seemed to confirm this, with the Vivid! returning a 3DMark2000 Pro score of 3,289 at a resolution of 1,024 x 768 in 32-bit colour on a Pentium III/800 graphics test bed with 128Mb of RAM. However, for a low-priced card this is very impressive, slightly better than the score from the Hercules 3D Prophet II MX. Given that some 3DMark2000 Pro tests benefit from hardware T&L, the Vivid!s score is all the more impressive.
OpenGL performance was less exciting though, with the Vivid! obtaining 34.7fps and 38.6fps in Quake III at 1,024 x 768 in 32-bit colour and 16-bit colour respectively. Unfortunately, a problem with the display drivers caused texture corruption when running at maximum detail. Dropping to lower texture details corrected this, although the underlying problem, which may also affect performance, will need addressing in future driver releases.
This impressive performance could be down to the rather atypical way the PowerVR KYRO chipset renders 3D. The Vivid! splits a 3D scene into smaller segments, or tiles, which it then renders independently. Since each segment takes up less memory than the full scene, some 3D functions can be performed on each tile, on-chip, without frequently accessing the memory. This has the benefit of reducing required memory bandwidth and boosting performance. The process of Z-buffering which assigns a depth value to objects dictating whether theyre seen or obscured by another surface in the scene is one 3D function performed on-chip by Vivid!
Vivid! also employs a technique called deferred texturing, which provides additional potential performance gains. Deferred texturing is where the removal of obscured surfaces (as dictated by the Z-buffer) is completed before textures and shades are applied, reducing the amount of processing required and also saving on memory requirements.
In keeping with the Vivid!s efficient rendering techniques, the KYRO chipset introduces another performance enhancing feature called Internal True Colour. This allows 3D to be rendered in 16-bit colour on-screen without the visual artefacts that are often caused by dithering. This is achieved by performing all blending operations internally in true colour (32-bit) and then performing a
single dither routine to 16-bit before outputting to memory.
The notion that 65,000 (16-bit colour) can look as good 16.7 million colours (32-bit colour) may sound far-fetched, but it does appear to work quite well. Observing 3D applications running in 16-bit colour shows a noticeable reduction in the banding and speckling often associated with 16-bit rendering. However, there are still visible artefacts, especially in dark areas, and while the Vivid! is certainly better than most graphics accelerators
in 16-bit colour, true 32-bit rendering offers better image quality.
Of course, 3D alone doesnt make a graphics card, so it was good to see that the Vivid! also performed reasonably well in 2D. The Vivid! has the requisite DVD playback options while the 270MHz RAMDAC should be enough to keep things sharp and focused. The VideoMark2000 score of 1,427 is also quite impressive for a budget card, although the Matrox Millennium G450 is rated a better 2D choice.
It seems that VideoLogic has learnt from its mistakes in the past and has pitched correctly with Vivid! Its not a high-end card, but for $??? it offers excellent Direct3D performance and good 2D performance. The price to performance ratio makes the Vivid! an excellent choice for budget to mid-range systems, and its clever rendering methods gives good scalability, which means it should stand the test of time quite well. My only concern is driver stability, but if VideoLogic can correct this, the Vivid! is definitely worth a look.
This article appeared in the January, 2001 issue of PC Authority.
Be the first to comment on this article.