In this study we investigate the inter-instrument variation of the alunite 1,480-nm OH (overall range ~1,475–1,498 nm) and the white mica Al-OH (overall range ~2,186–2,225 nm) spectral absorption features by measuring the same sets of samples using two devices of each of the three major types of portable SWIR instruments: portable infrared mineral analyzer (PIMA), TerraSpec, and portable near-infrared spectrometer (PNIRS). All instruments have good short-term reproducibility. The ARC Centre of Excellence in Ore Deposits (CODES) PIMA was assessed for long-term reproducibility and it was found to be stable, with 0- to 1-nm variations over at least seven years. There are differences in the results measured by different instruments, even of the same type. The PIMA instrument from the Economic Geology Research Unit generates alunite results 1 to 2 nm lower than those of the CODES PIMA. The two TerraSpec instruments used in this study produced results very similar to each other, mostly with 0- to 1-nm differences in alunite 1,480-nm spectral absorption measurements, but the results are 1 to 3 nm lower than those of the CODES PIMA. The two PNIRS devices generated results that differed from each other by 1 to 4 nm, mostly with 2- to 3-nm differences in alunite measurements. Compared with the PIMA results, one PNIRS dataset is very similar, with 0- to 1-nm differences, whereas the other PNIRS dataset is 0 to 5 nm off, mostly with 3-nm differences, in alunite measurements. The white mica Al-OH spectral absorption feature wavelength measurements are more similar among the instruments, with the maximum difference being 2 nm. Despite the differences in absolute values, the ranking of the samples in those datasets for the two tested parameters is largely the same, which indicates that all the instruments will produce datasets showing the same trends. The implication is that one project should use the same instrument for all the measurements. If another instrument has to be used, or the original instrument is repaired or tuned, it is suggested that a set of samples covering most of the variation range should be measured with both instruments to evaluate the inter-instrument variation and to establish a correction formula to keep consistency in the database.