Neutral-atom quantum computers are a promising platform for fault-tolerant quantum computation, but logical performance depends on systemic realistic noise factors during syndrome extraction. In dual-isotope Yb arrays, the roles of data and ancilla qu
Neutral-atom quantum computers are a promising platform for fault-tolerant quantum computation, but logical performance depends on systemic realistic noise factors during syndrome extraction. In dual-isotope Yb arrays, the roles of data and ancilla qubits are separated spectrally, allowing ancilla qubits to be measured in place without additional transport or shelving operations. Here we quantify the advantage of a dual-isotope Yb architecture for surface code memories. We develop an experimentally motivated Clifford-compatible noise model for dual-isotope 171Yb-174Yb systems using generalised Pauli twirling and implement it as a wrapper for Stim called DualYbSim, which has been packaged as an open source Python library. Simulations of rotated and XZZX surface codes show that a dual-isotope architecture with in-place measurement achieves the lowest logical error rates among the architectures considered, outperforming single-isotope schemes based on shelving or zoned measurement. Our error-budget analysis also identifies Rydberg-state decay as the dominant limitation, contributing to 74-80% of the logical error rate scaling, highlighting concrete experimental targets for improving FTQC performance.