Based on an extensive comparison of test methods for abrasion resistance from home and abroad, a new technique for abrasion resistance at elevated temperature has been developed with some improvements. Two connected airtight chambers are designed to prevent compressed air from entering the sample chamber, so that heat loss of the sample chamber can be reduced and the chamber temperature is less disturbed. The sample surface temperature can reach stable within 5 min and the temperature fluctuation on sample surface can be less than 20°C in the course of inleting the compressed air at working temperatures up to1400°C. The repeatability was tested using float-glass plate as reference sample at ambient temperature and using high alumina bricks for elevated temperature. A variation coefficient under 7% at ambient temperature has been achieved. Comparison of abrasion resistance at elevated temperature was tested respectively on a high alumina brick and on a silicon nitride bonded silicon carbide brick and the results could be significantly distinguished. Nitrogen can be blown into sample chamber during heating the furnace to prevent nonoxide bearing samples from being oxidized.