MeowPoW v2.0.0 — Maximum ASIC Resistance Benchmark
Prebuilt CUDA benchmarks for testing MeowPoW v2 hashrate on your GPU.
Downloads
- meowpow_bench.exe — RTX 50 series (sm_120)
- meowpow_bench_sm89.exe — RTX 40 series (sm_89)
Requirements
- NVIDIA GPU (RTX 40 or 50 series)
- Latest NVIDIA drivers installed
Algorithm Changes (vs MeowPoW v1)
| Parameter |
v1 |
v2 |
| Registers |
16 |
32 |
| L1 Cache |
16 KB |
128 KB |
| Cache Accesses |
6 |
14 |
| Math Operations |
9 |
24 |
| Math Op Types |
11 |
15 |
| Merge Functions |
4 |
6 |
| Period |
6 |
3 |
New GPU-Native Math Operations
byte_perm (CUDA __byte_perm)brev32 (CUDA __brev)funnelshift_l (CUDA __funnelshift_l)mad_lo32 (GPU MAD unit)
RTX 5090 Benchmark Results
| Batch Size |
Hashrate |
| 1,024 |
85.5 KH/s |
| 4,096 |
335.6 KH/s |
| 16,384 |
1.28 MH/s |
| 65,536 |
2.66 MH/s (peak) |
| 262,144 |
2.37 MH/s |
