empirical
Four hosts. Three GPU families. Two CUDA toolchains. Two PyTorch wheels. 90 samples each. Same bytes.
count
hosts
count
samples / host
result
digest agreement
bit-identical across classes
Every row ran the same 90-sample workload. A single-byte drift would show here. None observed.
why it matters
If two honest miners on different GPUs produced different proofs for the same completion, mesh consensus would fail — not because anyone cheated, but because of numerical drift. Reliquary's proof envelope is sized to absorb long-sequence attention drift, while remaining tight enough that tampering is caught. The empirical result above shows the envelope has substantial headroom: the honest path across four distinct hardware + toolchain combinations produces byte-identical digests.
Full reproduction steps + per-sample digests are committed in reliquary-ledger/docs/audit/cross_gpu.