HPC-Class Simulations on Consumer Hardware
Because we now have the optimal k-value calculator that CASCADE helped us find and optimize, we can always find optimal simulation paths for any physics problem.
The Breakthrough
Using Non-Newtonian Calculus with the optimal k parameter, we achieve 15x fewer timesteps while maintaining accuracy. This translates to up to 13.7x faster simulations on the same hardware.
Results below were obtained on 5-year-old consumer hardware: Intel Core i5-10400F + NVIDIA RTX 2060 SUPER
CPU
Intel i5-10400F
6 cores, 12 threads @ 2.9GHz
GPU
RTX 2060 SUPER
8GB VRAM, ~$300 in 2020
Classification
Consumer Gaming PC
NOT a workstation or cluster
Benchmark Results: Lennard-Jones Molecular Dynamics
1 picosecond simulation at 300K with periodic boundaries
| Atoms | Classical | NNC (k=-1) | Speedup | Efficiency |
|---|---|---|---|---|
| 500 | 1.30s | 0.85s | 1.5x | 10% |
| 1,000 | 4.58s | 1.10s | 4.2x | 28% |
| 2,000 | 16.80s | 1.90s | 8.8x | 59% |
| 5,000 | 99.25s | 7.69s | 12.9x | 85% |
| 10,000 | 379.96s | 27.73s | 13.7x | 91% |
Wall Time Comparison (Log Scale)
Loading visualization...
What This Means for Real Simulations
| Simulation | Classical (HPC Cluster) | NNC (Your Gaming PC) |
|---|---|---|
| 1,000 atoms, 1 nanosecond | ~30 minutes | ~2 minutes |
| 10,000 atoms, 1 nanosecond | ~6 hours | ~28 minutes |
| 50,000 atoms, 100 picoseconds | ~2 days | ~3 hours |