Portable X-ray fluorescence (pXRF) analysers are increasingly popular tools for geoscientific applications, including mineral exploration. One promising application, illustrated in the companion paper, is to obtain high-spatial resolution down-hole geochemical profiles using pXRF on unprepared exploration drill-cores. However, the precision and accuracy of pXRF analysers on such samples is not well studied. We have tested three Olympus Innov-X analysers, both on a sediment standard (NIST 2702, ‘Inorganics in Marine Sediment’) and in-situ on unmineralized rock cores from volcanic and intrusive, mafic to felsic lithologies. We conclude that pXRF is quite precise for a number of elements, but not very accurate using factory calibrations. For example, the 1σ precision of one Delta Premium analyser tested on a basaltic core, in mining plus mode, with a 60 s integration time, is better than 5 % for Al, Ca, Fe, K, Mn, S, Si, Ti, Zn and Zr. The same analyser, tested on a range of volcanic and intrusive core samples, yielded the following average systematic errors: Al -23 %, Ca -4 %, Fe +1 %, K -9 %, Mg -17 %, Mn -15 %, P +218 %, Si +4 %, Ti -23 %, Cu +220 %, Zn +151 %, and Zr +17 %. These systematic errors can largely be removed by the application of correction factors, which are unique to each analyser and each project. Without such corrections, the three analysers tested, including two ‘identical’ Delta Premium models, yield different results on the same sample. Another important finding is that within 20 cm long core samples, the effect of mineralogical heterogeneity on in-situ pXRF data is much larger than that of the instrument precision. Finally, with the Delta analysers, both the ‘mining plus’ and the ‘soil’ modes are needed to determine as many elements as possible with the best data quality possible.