The Bordoni relaxation in fcc metals is generally assumed to result from thermally activated kink-pair formation in dislocation segments aligned along Peierls valleys in crystallographic close-packed directions. Objections against this interpretation have been raised by pointing out, that, at the small stresses applied in the experiments, the Paré condition is not satisfied and that the Peierls stress derived from the experiments by far exceeds the experimental values for the flow stress extrapolated to 0 K. It is shown that, when the bow-out configuration of a dislocation segment is treated as a kink-chain, and when the splitting of dislocation into partials is accounted for, both objections cannot be maintained. Under the assumption that the kink mobility is high and using analytical expressions for the kink-chain configuration and the kink-chain enthalpy under stress an expression for the bow-out rate of dislocation segments is obtained. Trivial numerical integration, under the action of a periodic applied stress, leads to a phase-lag between strain and stress and hence to energy dissipation. The dependence of this energy loss on segment lengths, temperature and internal stress is derived explicitly.