I would agree with that. Amiga is a different species. It was targetted for games, real-time audio/video effects (which arcade-type games require), fast game port interface, etc. PCs gaming is like a delayed (after-thought) superficial imposition on the computer-- slower game port, effects have to be done through slower APIs rather than hardware standard (and not all hardware supports all API calls),
PC GPUs changes its micro-architecture nearly every generation. It’s like changing from PPC-to-X86-to-ARM-to-(yet another processing core) nearly every year.
For example
NV Geforce FX 5x00 VILW based architecture.
NV Geforce 6x00/7x00 SIMD/MIMD based architecture.
NV Geforce 8x00 (CUDA) Scalar based architecture.
AMD Radeon X1xxx SIMD/MIMD based architecture.
AMD Radeon HD 3xx0(CTM/CAL) VILW based architecture.
If the game was designed to “hit-the-metal” the hardware would become a “boat anchor” for any architecture changes.
I can’t run Fold@Home GPU2 (written on AMD's CAL) on NV CUDA hardware.
Note why you don’t see any commercial PC games being designed directly on NV's CUDA or AMD's CTM(aka "Close-To-Metal")/CAL. The abstraction layer enables rapid hardware architecture changes without completely killing software compatibility.