RIVA TNT2
Nvidia's fifth graphics chip, succeeding the RIVA TNT.
The RIVA TNT2 is a graphics chip from Nvidia that hit the market in early 1999. Its internal codename is NV5, marking it as Nvidia’s fifth graphics design and the follow-up to the RIVA TNT (NV4). RIVA stands for Real-time Interactive Video and Animation accelerator, while TNT is short for TwiN Texel, referencing the chip’s ability to process two texels at once. Later in its life, Nvidia dropped the RIVA part of the name.
The TNT2 core keeps the same basic dual-pipeline setup as the RIVA TNT but adds several updates: support for a 32-bit Z-buffer and stencil, AGP 4X, up to 32MB of video memory, and a manufacturing process that shrank from 0.35 µm to 0.25 µm. That shrink allowed clock speeds to jump from 90 MHz to over 150 MHz, which is where most of the performance gain came from.
A cheaper version called the TNT2 M64 cut the memory interface from 128-bit down to 64-bit. These cards were sometimes labeled Vanta, a name that started with a budget RIVA TNT product. The M64 outperformed the older RIVA TNT while costing less to make, and it became popular in the OEM market because many buyers assumed all TNT2 cards were the same.
The TNT2 went up against the 3dfx Voodoo2, 3dfx Voodoo3, Matrox G400, and ATI Rage 128. Its main rival was the Voodoo3, which lacked 32-bit color output in 3D—a key advantage for the TNT2. The Voodoo3 was marketed on speed and game compatibility. At the time, 3dfx’s Glide API was still widely used and often ran faster with better image quality than Direct3D or OpenGL. Some games, like Wing Commander: Prophecy, had exclusive 3D features under Glide, and Unreal had a troubled history with Direct3D, missing details in that mode.
The Voodoo3 rendered internally at 32-bit precision but dithered that down to a 16-bit framebuffer, then applied a 2x2 box filter in the RAMDAC, which 3dfx marketed as “22-bit equivalent” output. This produced noticeably less dithering than the TNT2’s 16-bit mode, though it wasn’t true 32-bit. Because the effect happened after the framebuffer, screen captures didn’t show it, leading to mistaken claims that TNT2 16-bit quality matched Voodoo3. In many games, Voodoo3’s 16-bit output was actually closer to TNT2’s 32-bit quality. As games used more alpha blending and multipass effects, true 32-bit rendering became more important.
The two chips also differed in pipeline design.
Quick Facts
- Codename
- NV5
- Entry
- Vanta, M64
- Midrange
- TNT2, Pro
- Highend
- TNT2 Ultra
- Openglversion
- 1.2
- Predecessor
- RIVA TNT
- Successor
- GeForce 256
- Support Status
- Unsupported
Facts from the source article.
Lore & Background
The RIVA TNT2's competition included the 3dfx Voodoo2, 3dfx Voodoo3, the Matrox G400, and the ATI Rage 128. The main competitor was the Voodoo3, which lacked 32-bit color output in 3D, a distinguishing point for the TNT2. The Voodoo3 was marketed on superior speed and game compatibility. The 3dfx Glide API remained popular, frequently performing faster and with better image quality than Direct3D and OpenGL. Some games had exclusive 3D features when used with Glide, including Wing Commander: Prophecy, and Unreal had a troubled development history with Direct3D, plagued by missing details in that mode. Voodoo3 cards rendered internally in 32-bit precision color depth, dithered down for the 16-bit framebuffer, then postprocessed by a 2x2 box filter in the RAMDAC, marketed as '22-bit equivalent' output. This resulted in less dithering than TNT2's 16-bit output, but was not equivalent to real 32-bit output. The postprocessed nature meant framebuffer captures did not display the effect, leading to erroneous claims equating TNT2 16-bit quality to Voodoo3 when in many titles Voodoo3 16-bit quality was closer to TNT2 32-bit quality in practice. 32-bit rendering became more important with heavier use of alpha blending and multipass effects. The Voodoo3 has a single dual-texturing pipeline (1x2), while the TNT2 has two single-texturing pipelines (2x1). In games applying only a single texture per polygon face, the TNT2 could be more efficient and faster, though single-texturing was no longer used in most new games at launch.
Reader's Guide
The RIVA TNT2 was notable for its process shrink to 0.25 μm, which allowed higher clock speeds and substantial performance gains over its predecessor. Its support for 32-bit color output in 3D distinguished it from the competing Voodoo3, which relied on a dithered 16-bit framebuffer. The TNT2's dual single-texturing pipeline design offered efficiency advantages in older games, though by launch most new titles used multi-texturing. A low-cost version, the TNT2 M64 (sometimes labeled Vanta), reduced the memory interface to 64-bit, outperforming the older RIVA TNT while being less costly to produce, and proved popular in the OEM market as many consumers assumed all TNT2 cards were the same. ALi integrated the TNT2 core into the Aladdin TNT2 motherboard chipset, supporting both local frame buffer and unified memory mode, though low-cost boards without separate memory were known for low speed and stability issues. The TNT2's legacy includes its role in the transition to 32-bit rendering and its competition with 3dfx's Glide ecosystem, with some cards like the Creative 3D Blaster TNT2 Ultra including a Glide wrapper for compatibility.
Did You Know?
- The RIVA TNT2's process shrink from 0.35 μm to 0.25 μm enabled clock speeds from 90 MHz to 150+ MHz.
- A low-cost version, the TNT2 M64, reduced the memory interface from 128-bit to 64-bit and was sometimes labeled Vanta.
- The Falcon Northwest Special Edition Maxi Gamer Xentor 32 SE was a TNT2 Ultra card operating at a record-breaking 195 MHz core and 235 MHz RAM.
Frequently Asked Questions
What is the RIVA TNT2?
The RIVA TNT2 is a graphics processing chip designed by Nvidia and launched in early 1999. It carries the internal codename NV5, identifying it as the company's fifth-generation GPU architecture and the direct successor to the RIVA TNT.
What do the letters "RIVA" and "TNT" in the name actually stand for?
RIVA is an abbreviation for Real-time Interactive Video and Animation accelerator, while TNT stands for TwiN Texel. The TwiN Texel portion highlights the chip's dual-texture-pipeline design, which lets it handle two texels simultaneously.
How does the TNT2 improve on the original RIVA TNT?
While it retains the same dual-pipeline core layout as its predecessor, the TNT2 adds a 32-bit Z-buffer with stencil support, AGP 4X connectivity, and room for up to 32 MB of video memory. It also moved to a 0.25-micron manufacturing process and ran at core clocks of 150 MHz or higher.
What memory interfaces did the TNT2 support?
The standard TNT2 used a 128-bit memory bus, but Nvidia also offered a budget M64/Vanta variant that cut the interface down to 64 bits. Both versions shared the same NV5 core but targeted different price points.
Why did Nvidia eventually stop calling it "RIVA TNT2"?
As the product line matured, Nvidia phased out the "RIVA" branding entirely and simply referred to the chip as the TNT2. This marked a shift toward shorter, cleaner product names that carried forward into later GeForce-era GPUs.
More in Nvidia and AMD Graphics Processors 1-24
Spotted an error? Know more?
Reader corrections go straight into our review queue. Suggest an edit · How this site is sourced