GPU compute/ALU capacity growing faster than memory bandwidth made primitive-based SDFs increasingly competitive from 2007 onward

IQ identifies a specific hardware trend as the reason primitive-based SDFs became viable around 2007: GPUs were evolving their compute/ALU capabilities faster than their memory bandwidth. Primitive SDFs require many ALU operations (math evaluations per pixel) but almost no memory access (no textures, no geometry buffers). As GPUs grew more compute-dense relative to bandwidth, the trade-off increasingly favoured pure mathematical SDFs over baked representations. This principle generalises: shader techniques that are ALU-heavy and memory-light benefit from GPU architectural trends. IQ notes this trend ‘has only continued in the last 20 years’, making SDFs even more competitive over time.

Examples

A Shadertoy shader with 200 iterations of SDF evaluation per pixel is compute-heavy but touches no textures — well suited to modern GPUs where compute is cheap and memory bandwidth is the bottleneck.

Assessment

Explain the GPU hardware trend that made primitive-based SDFs increasingly viable from 2007, and describe what kind of shader workload benefits most from a high ALU-to-bandwidth ratio.