We report novel techniques allowing the measurement of Nd-isotope ratios with unprecedented accuracy and precision by multi-collector inductively coupled plasma mass spectrometry. Using the new protocol, we have measured the Nd-isotopic composition of rock and synthetic Nd standards as well as that of the Allende carbonaceous chondrite. Analyses of BCR-2, BHVO-2 and GSP-2 rock standards yield mass-independent compositions identical to the JNdi-1 Nd-reference standard, with an external reproducibility of 2.4, 1.6, 1.6 and 3.5 ppm respectively, on μ142Nd, μ145Nd, μ146Nd and μ150Nd (μ representing the ppm-deviation of the ratios from JNdi-1) using 148Nd/144Nd for internal normalization. This represents an improvement in precision by a factor of 2, 7 and 9 respectively for μ142Nd, μ145Nd and μ150Nd. Near-quantitative recovery from purification chemistry and sample-standard bracketing allow for the determination of mass-dependent Nd-isotopic composition of samples. Synthetic standards, namely La Jolla and AMES, record mass-dependent variability of up to 1.2 ε per atomic mass unit and mass-independent compositions resolvable by up to 3 ppm for μ142Nd and 8 ppm for μ150Nd, relative to JNdi-1. The mass-independent compositions are consistent with equilibrium mass fractionation during purification. The terrestrial rock standards define a uniform stable ε145Nd of -0.24 ± 0.19 (2SD) relative to JNdi-1, indistinguishable from the mean Allende ε145Nd of -0.19 ± 0.09. We consider this value to represent the mass-dependent Nd-isotope composition of Bulk Silicate Earth (BSE). The modest mass-dependent fractionation of JNdi-1 relative to BSE results in potential effects on mass-independent composition that cannot be resolved within the reproducibility of our analyses when correcting for natural and instrumental mass fractionation by kinetic law, making it a suitable reference standard for analysis of unknowns. Analysis of Allende (CV3) carbonaceous chondrite returns an average μ142Nd deficit of -30.1 ± 3.7 ppm in agreement with previous studies. The apparent deficit is, however, lowered to -23.8 ± 4.0 ppm while normalizing to 148Nd/144Nd instead of 146Nd/144Nd. We interpret this as the effect of a possible nucleosynthetic anomaly of -6.3 ± 0.5 ppm in μ146Nd. As 142Nd and 146Nd are both s-process-dominated nuclides, this hints at the possibility that terrestrial μ142Nd excess may not reflect 146Sm decay as widely accepted.