GeForce2

The GeForce2 (codenamed NV15) was the second generation of GeForce graphics cards by NVIDIA Corporation.

GeForce2 GTS

The first model, GeForce2 GTS (also known as NV15), was named for its texels rate of 1.6 billion per second - GigaTexel Shader. Due to the addition of a second TMU (texture map unit) to each of 4 pixel-pipelines, and a higher core-clock rate (200MHz vs 120MHz), the GTS's texel fillrate is 3.3 times higher than that of its predecessor, the GeForce 256 (480 Mtexel/sec.) Other hardware enhancements included an upgraded video-processing pipeline, called HDVP (high definition video processor). HDVP supported motion video-playback at HDTV-resolutions (MP@HL), although playback of high-resolution video still required a powerful CPU. The GeForce2 also introduced the NVIDIA Shading Rasterizer (NSR), which was actually a primitive form of what is known as Shaders today (The GeForce 256 also had this feature, but it was never publicly announced).

In 3D benchmarks and gaming applications, the GTS outperformed its predecessor (GeForce 256) by up to 40%. (The GTS was only 10% faster than the DDR-SDRAM GeForce.) In OpenGL games (such as Quake III), the GTS outperformed the ATI Radeon and 3DFX Voodoo 5 cards in both 16bpp and 32bpp (true-color) display modes. But in Direct3D games, the Radeon was sometimes able to take the lead in 32-bit colour modes.

As studies of the architecture used for GeForce 256 and GeForce2 progressed, it was determined that they were extremely memory bandwidth constrained. The chips wasted a lot of memory bandwidth and pixel fillrate on unoptimized z-buffer usage, drawing of hidden surfaces, and a relatively inefficient RAM controller. The main competition for these two chips, the ATI Radeon DDR, had effective optimizations (HyperZ) that combated these issues. Because of how wasteful these GeForce chips were they couldn't approach their theoretical performance potential, and Radeon, with its significantly less-endowed pipeline count, was able to offer strong competition simply due to greater efficiency. The later NV17 revision of the NV10 design, used for GeForce4 MX, was far more efficient and the MX 460 outperformed GeForce2 Ultra even though it was only a 2x2 pipeline design.

Additionally, the state of the PC gaming software at the time, and the newness of the DirectX 7 (API), likely limited the amount of game software able to take advantage hardware multitexturing capabilities (the most significant difference between GeForce256 and GeForce2). Most games emphasized single-layer texturing on surfaces, which would not benefit from multitexturing hardware found in the GeForce2 or Radeon.

GeForce2 cores

There were three more revisions on the GeForce2 GTS core - the first was the GeForce2 Ultra, launched in late 2000. Architecturally identical to the GTS, the Ultra was spec'd with higher core and memory clocks. It put a definite lead between it and the Radeon and Voodoo 5, and even outperformed the first GeForce 3 products. Geforce 3 initially had lower texture fillrate than GF2 Ultra, but GeForce 3 Ti500 in late 2001 finally overtook the GeForce2 Ultra. Some speculate the Ultra was intended to defeat 3DFX's Voodoo 5 6000. While later tests showed that the Ultra outperforming the Voodoo 5 6000, the Voodoo 5 6000 never reached the consumer market. The other two GTS revisions were the GeForce2 Pro and the GeForce2 Ti (for "titanium"). Both parts fell between the GTS and Ultra. They were positioned as cheaper, but less advanced (feature-wise), alternatives to the high-end GeForce 3, which lacked a mass-market version.

GeForce2 MX

Finally, the most successful GeForce2 part was budget-model GeForce2 MX. In terms of sales volume, no graphics processor before or since has matched the GeForce2 MX (and its variants), partly due to the fact that the GeForce 3 never had a budget model, meaning that the GeForce2 MX family was NVIDIA's mainstream processor for two and a half years, compared to the more usual one year. The MX retained the GTS's core 3D-architecture and feature-set, but removed two 3D pixel-pipelines and half of the GTS's memory bandwidth. NVIDIA also added true dual-display support for the MX. (The GTS and subsequent models could drive a separate TV-encoder, but this second-display was always tied to the primary desktop.)

The GeForce2's efficient design - 4 watts of power consumption versus the 8 watts for the GeForce2 GTS and 16 required for the old GeForce256 - enabled it to be easily adopted for the laptop market as the GeForce2 Go. ATI's competing Radeon VE (later Radeon 7000) did not offer hardware T&L while the Radeon SDR was released late and was still too expensive.

In addition to being released early and achieving the best price/performance ratio, the MX and the rest of the GeForce2 line were backed by a single reliable driver unlike ATI whose products suffered from unreliable drivers.

The MX performed well enough to make it a viable mainstream alternative to the GTS (and its later revisions.) Among the gamer community, the MX effectively replaced the older TNT2 cards. NVIDIA eventually split the MX product-line into "performance-oriented" (MX400) and "cost-oriented" (MX200) versions. The MX200 had a 64-bit SDR memory bus, greatly limiting its gaming potential. The MX400, like the original MX, had a 128-bit (SDR) memory bus which could also be configured as 64-bit DDR.

The GeForce2 MX's successor was the GeForce 4 MX. Although many disappointed enthusiasts described the GeForce 4 MX as a GeForce2 MX with a better (128-bit DDR) memory controller, the GeForce 4 MX also owes a good deal of its design heritage to NVIDIA's high-end CAD products. As a result, the GeForce 4 MX (with the exception of the MX420) efficiently achieves performance similar or better than the "brute-force" GeForce2 Ultra.

The GeForce2 MX was later used by NVIDIA as integrated graphics on its nForce line of motherboard chipsets for AMD Athlon and Duron processors.

Models

(Performance ranking, slowest to fastest) ^*

GeForce2 MX100 ^*
GeForce2 Go
GeForce2 MX200
GeForce2 MX
GeForce2 MX400
GeForce2 GTS
GeForce2 GTS (64 MiB) ^*
GeForce2 Pro
GeForce2 Ti
GeForce2 Ultra

Chip table

GeForce2 Chip	Triangles per second (Million)	Pixel Fillrate (Megapixels)	Texel Fillrate (Megatexels)	Memory Bandwidth (GB/s)
Ultra	31	1000	2000	7.3
Ti	31	1000	2000	6.4
PRO	25	800	1600	6.4
GTS	25	800	1600	5.3
MX DDR	20	350	700	2.3
MX400	20	400	800	2.7
MX200	20	350	700	1.35
MX100	< 20	286	572	0.68
Go	< 20	286	572	~1.35

External links

NVIDIA: GeForce2 leading-edge technology

Graphics cards | NVIDIA

GeForce2 | GeForce 2