12 Graphics Card Generations Compared by the Performance Jump Each Delivered
9. The Architecture Wars - Fermi vs. VLIW5 Optimization (2010-2012)
The ninth generation was characterized by competing architectural philosophies, with NVIDIA's Fermi architecture (GTX 580) emphasizing compute performance and ATI's VLIW5 design (HD 6970) focusing on graphics optimization. Performance improvements were relatively modest at 1.5-2x, but this generation established important precedents for future GPU development. The GTX 580's Fermi architecture featured improved shader efficiency, better branch handling, and enhanced double-precision floating-point performance that benefited both graphics and compute applications. The HD 6970's VLIW5 architecture maximized graphics performance through highly optimized shader units specifically designed for typical graphics workloads, often delivering superior frame rates in traditional gaming scenarios. This generation highlighted the growing importance of software optimization, with games increasingly showing significant performance differences based on how well they utilized each architecture's strengths. The introduction of advanced anti-aliasing techniques like FXAA and improved temporal anti-aliasing provided better image quality with lower performance penalties than traditional MSAA. Power efficiency improvements were incremental but important, with both architectures delivering better performance per watt than their predecessors through architectural refinements rather than process improvements. The generation also saw the maturation of GPU computing applications, with CUDA and OpenCL enabling graphics cards to accelerate everything from video encoding to scientific simulations, expanding the market beyond traditional gaming applications.
