The GeForce 256 is the product that gave the GPU its name. NVIDIA’s own corporate timeline places the invention of “the GPU, the graphics processing unit” in 1999, and the GeForce 256 was the chip that carried the claim. NVIDIA marketed it as the world’s first GPU and defined the category as a single-chip processor with integrated transform, lighting, triangle setup and clipping, and rendering engines, capable of processing a minimum threshold of polygons per second.
The headline technical advance was hardware transform and lighting (T&L). In earlier 3D accelerators, the geometry stage, the math that positions, rotates, and projects each vertex and computes its lighting, ran on the host CPU; the accelerator handled only the later rasterization stage. The GeForce 256 pulled that geometry stage onto the graphics chip. NVIDIA’s own transform-and-lighting technical material described dedicated, separate transform and lighting engines on the GPU so that each could run at maximum efficiency, freeing the CPU and removing what had become a major bottleneck for richer 3D scenes.
This was a structural change to where the rendering pipeline ran, not merely a faster version of the same thing. By integrating transform, lighting, triangle setup, clipping, and rendering on one chip, the GeForce 256 implemented the bulk of the fixed-function 3D pipeline in silicon. The frame buffer, depth buffer, and texture units were all driven by a processor that now also owned geometry processing, which is precisely what distinguished a GPU from the graphics accelerators that preceded it.
The architectural lineage from this part runs directly to the programmable GPUs that followed. NVIDIA’s GPU Gems 2 description of the later GeForce 6 Series lays out the full pipeline that the GeForce 256 began to consolidate, command parsing, vertex fetching and processing, primitive grouping, culling and clipping, rasterization, fragment processing, depth testing and blending, and output to the frame buffer, with the fixed-function vertex stage of 1999 eventually replaced by fully programmable vertex and fragment processors.
A note on dates: NVIDIA announced the GeForce 256 in 1999 and this entry uses the August 31, 1999 announcement date. Retail availability of the SDR card followed later in the autumn of 1999, with a DDR-memory variant shortly after; the precise ship dates are reported by secondary sources and are not asserted here beyond the 1999 launch that NVIDIA’s own timeline records.