The GeForce 2 series (NV15) is the second generation of Nvidia's GeForce line of graphics processing units (GPUs). Introduced in 2000, it is the successor to the GeForce 256.

The GeForce 2 family comprised a number of models. The GeForce 2 GTS, GeForce 2 Ultra, GeForce 2 Pro, and GeForce 2 Ti are based upon the original architecture (NV15), varying only by chip and memory clock speeds. For the low-end segment and OEMs, the GeForce 2 MX series (NV11) was created, from which the GeForce 2 Go was derived for laptops. In addition, the GeForce 2 architecture is used for the Quadro series on the Quadro 2 Pro, 2 MXR, and 2 EX cards with special drivers meant to accelerate computer-aided design applications.

The GeForce 2 architecture (NV15) is similar to the previous GeForce 256 line but with various improvements. Compared to the 220 nm GeForce 256, the GeForce 2 is built on a 180 nm manufacturing process, making the silicon more dense and allowing for more transistors and a higher clock speed. The most significant change for 3D acceleration is the addition of a second texture mapping unit to each of the four pixel pipelines. This doubles the texture fillrate per clock compared to the previous generation and is the reasoning behind the GeForce 2 GTS's naming suffix: GigaTexel Shader (GTS). The GeForce 2 also formally introduces the NSR (Nvidia Shading Rasterizer), a primitive type of programmable pixel pipeline that is somewhat similar to later pixel shaders. This functionality is also present in GeForce 256 but was unpublicized. Another hardware enhancement is an upgraded video processing pipeline, called HDVP (high definition video processor). HDVP supports motion video playback at HDTV-resolutions (MP@HL).

In 3D benchmarks and gaming applications, the GeForce 2 GTS outperforms its predecessor by up to 40%. In OpenGL games (such as Quake III), the card outperforms the ATI Radeon DDR and 3dfx Voodoo 5 5500 cards in both 16 bpp and 32 bpp display modes. However, in Direct3D games running 32 bpp, the Radeon DDR is sometimes able to take the lead.

The GeForce 2 (NV15) architecture is quite memory bandwidth constrained. The GPU wastes memory bandwidth and pixel fillrate due to unoptimized z-buffer usage, drawing of hidden surfaces, and a relatively inefficient RAM controller. The main competition for GeForce 2 GTS, the Radeon DDR (R100), has hardware functions (called HyperZ) that address these issues. Because of the inefficient nature of the GeForce 2 GPUs, they could not approach their theoretical performance potential and the Radeon, even with its significantly less powerful 3D architecture, offered strong competition. The later NV17 revision of the NV11 design used in the GeForce4 MX was more efficient.

The first models to arrive after the original GeForce 2 GTS was the GeForce 2 Ultra and GeForce2 MX, launched on September 7, 2000. On September 29, 2000, Nvidia started shipping graphics cards which had 16 and 32 MB of video memory size.

Architecturally identical to the GTS, the Ultra simply has higher core and memory clock rates. Meant to be a niche product, it was rumored that GeForce 2 Ultra was intended to prevent 3dfx taking the lead with their Voodoo 5 6000 that was ending up never released as 3dfx went bankrupt. The Ultra model actually outperforms the first GeForce 3 products in some cases, due to initial GeForce 3 cards having significantly lower fillrate. However, the Ultra loses its lead when anti-aliasing is enabled, because of the GeForce 3's new memory bandwidth/fillrate efficiency mechanisms; plus the GeForce 3 has a superior next-generation feature set with programmable vertex and pixel shaders for DirectX 8.0 games.

The GeForce 2 Pro, introduced shortly after the Ultra, was an alternative to the expensive top-line Ultra and is faster than the GTS.