Folding Scaling Benchmark
Comparing $O(N^2)$ Transformer-based Attention vs. NRC $O(N)$ Lattice Resonance.
Residues (N)
Standard GPU Time
NRC Lattice Time
Efficiency Delta
100
450ms
12ms
37.5x
500
11.2s
58ms
193x
2000
184s
230ms
800x
8000
Out of VRAM
940ms
∞