An anniversary celebrated a little early
In a news article published on October 12, VideoCardz reports that NVIDIA is celebrating the GeForce 8800 GTX as part of GeForce Day 2026. A video hosted by Jacob Freeman revisits the card and its role in the early days of CUDA.
The timing does need some clarification: the 8800 GTX launched on November 8, 2006. Its official twentieth anniversary is therefore still a few weeks away. It was offered at $599, alongside an 8800 GTS announced at $449: those were US launch prices, not current prices.
Why G80 changed how the GPU works
The key innovation lay in how the compute units were organised. With its unified architecture, the 8800 GTX has 128 stream processors that can be assigned to different shader tasks as needed. Instead of reserving separate units for vertices and pixels, the GPU allocates its resources more flexibly.
NVIDIA’s technical documentation also describes support for DirectX 10 and Shader Model 4.0. This evolution notably paved the way for geometry shaders. At the time, using DirectX 10 required Windows Vista: upgrading the graphics card alone was not enough.
Specifications that reflect their era
The GeForce 8800 GTX pairs a GPU clocked at 575 MHz with shader processors running at 1.35 GHz. It carries 768 MB of GDDR3 on a 384-bit bus, delivering 86.4 GB/s of bandwidth. Its outputs include two dual-link DVI ports and an HDTV output.
Its other legacy reaches beyond gaming. CUDA makes it possible to use GPU units for general-purpose computing, and NVIDIA’s technical presentations from that period already mention physics, matrix operations and biological sequence analysis. The 8800 GTX thus illustrates the convergence of graphics acceleration and parallel computing.
Key takeaways
Above all, this anniversary recalls an architectural shift: with G80, the graphics card became more flexible at rendering games and a programmable computing resource beyond 3D rendering.