By Andrey Vorobiev

This chip is presented by Hercules Dynamite TNT2Ultra 32 MBytes, AGP videocard (legendary last product of Hercules Computer - Dynamite TNT2 Ultra that works on higher clock rates than others of the type. Hercules Computer went bankrupt and was bought by Guillemot due to production of this card).

Features:

• NVIDIA Riva TNT2Ultra chip, clock rate lowered down to 125 MHz;
• 32 MBytes SDRAM memory in 16 Hyundai 5.5ns microchips, clock rate lowered down to 150 MHz, 128-bit bus;
• Peak speed of scene drawing in multitexturing mode is 125 megapixels/sec and 250 megatexels/sec.

Wasn't tweaked due to official version that works on higher clock rates (NVIDIA Riva TNT2 Ultra).

Exactly 2 years ago in hard business battles between NVIDIA, 3dfx and Matrox, Riva TNT2 based cards have appeared. Actually they were just tweked Riva TNTs as the architechture had remained the same: two rendering pipelines with 1 texture module per each, so actually 1 pipeline with 2 TMUs (pipelines combine in the multitexturing mode). This chip is rather old, it doesn't support all new features such as anisotropy, cubic and relief texturing, etc. It also lacks the real trilinear filtering (it can be enabled for the price of half the perfomance). There are almost no cards based on this chip as they were replaced by Riva TNT2Pro and Riva TNT2Ultra.

For the 20th of March '2001 the latest drivers versions from NVIDIA are 6.88, 7.56, 10.70, 10.80 and 11.00.

As you see old 2.* drivers work very good with this card. Pay attention to falloffs with 6.* (6.11, 6.16, 6.18, 6.20) versions. I talked about speed boost with 6.26 version drivers but there was just a bug in other versions. Now it's fixed. I want readers to note 7.* drivers, they are just "raw" and can't functions properly even on beta level. Quake3 and other OpenGL applications doesn't run at all, and D3D games show big speed falloff. 10.* version showed slight perfomance falloffs in Direct3D. I chose 7.56 version for further testing, estimating of 3D-quality and also for calculating usability coefficients.

There are not so many problems with 3D graphics in old games, but new ones like Giants cause bunches of artefacts to show up:

And trilinear approximation instead of true trilinear filtering lowers the card's rating (you can see this approximation on the shots above - take a look at the sandy soil). Below are shots taken on this card (to the left) and on NVIDIA GeForce3 (to the right).

Quake3:

Unreal:

Unreal Tournament:

No One Lives Forever:

Serious Sam:

Colin McRAE Rally2:

American McGee's Alice:

Mercedes Benz Truck Racing:

Sacrifice:

FAKK2:

Need For Speed 5 (Porshe 2000):

Real MYST:

Blade of Darkness:

Giants:

3DMark2001 Pro, GAME1:

3DMark2001 Pro, GAME2:

3DMark2001 Pro, GAME3:

3DMark2001 Pro, Vertex Shaders:

This card has fair to good 2D quality. There are "soaping" effects in resolutions of and higher than 1280x1024. There are many videocards based on TNT2/TNT2-A/TNT2Ultra chipsets with not so fair 2D quality therefore given material can't prove the opinion that all Riva TNT2 based cards have good 2D. [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive] [an error occurred while processing this directive]