Skip to content

GaiaChess 4.2.2

Latest

Choose a tag to compare

@github-actions github-actions released this 13 Aug 14:59
· 1 commit to main since this release

Highlights

  • First open-source release — GaiaChess is now open source under GPLv3, engine and NNUE trainer included. Source: https://github.com/jromang/gaiachess
  • Minor search improvements over 4.2.1 (two SPRT-validated heuristic refinements).

Download

Pick the binary matching your CPU. Windows binaries end in .exe, Linux and macOS have no extension.

Which binary should I use?

CPU Recommended binary
AMD Ryzen 9000 (Zen 5) znver5 or avx512vnni
AMD Ryzen 7000 (Zen 4) znver4 or avx512
AMD Ryzen 5000 (Zen 3) znver3 or bmi2
AMD Ryzen 1000–3000 (Zen 1/2) avx2
Intel 12th gen+ (Alder Lake+) avx512vnni
Intel 10th–11th gen avx512 or bmi2
Intel Haswell–Coffee Lake bmi2
Apple M1/M2/M3/M4 apple-silicon
Linux ARM64 (RPi 4+, Graviton) neon
Older CPUs (pre-2013) sse4-popcnt, ssse3, or x86-64

When in doubt, use bmi2 (x86) or neon (ARM).

Binary details

Suffix SIMD Eval Target CPUs
x86-64 PeSTO Any x86-64 CPU (2003+)
ssse3 SSSE3 PeSTO Core 2+ (2006+)
sse4-popcnt SSE4 + POPCNT PeSTO Nehalem+ (2008+)
avx2 AVX2 NNUE Haswell+, Zen 1/2 (no fast PEXT)
bmi2 AVX2 + PEXT NNUE Haswell+, Zen 3+
avx512 AVX-512 NNUE Skylake-X, Zen 4
avx512vnni AVX-512 + VNNI NNUE Ice Lake+, Zen 5
znver3 AVX2 + PEXT NNUE AMD Ryzen 5000 (tuned)
znver4 AVX-512 NNUE AMD Ryzen 7000 (tuned)
znver5 AVX-512 + VNNI NNUE AMD Ryzen 9000 (tuned)
neon NEON NNUE ARM64 Linux
apple-silicon NEON NNUE macOS Apple Silicon

All x86-64 binaries are compiled with PGO (Profile-Guided Optimization) for maximum performance.
The AVX-512 variants are profiled under Intel SDE, since no CI runner is guaranteed to support AVX-512.
NNUE binaries embed the neural network — no external file needed.
PeSTO binaries (x86-64, ssse3, sse4-popcnt) use a classical evaluation for older hardware.