Background: Mathematics homework is highly prevalent in East Asia. Teachers and parents expect mathematics homework to improve students' performance; however, studies have not clearly defined the effectiveness of the assignment of different amounts of homework. Aims: This study analyses the differential effect of homework amount on various facets of Chinese students' mathematics achievement using a longitudinal design. It aims, thereby, to contribute to the existing knowledge of homework effectiveness and mathematics learning and to provide insights for student development. Sample: The sample included 2383 grade 9 students from a mid-sized city with a moderate level of economic and educational development in central China. Methods: We administered a student-reported questionnaire, collected school reports of mathematics achievements to track retrospective longitudinal variations in mathematics over half a year and analysed the differential effectiveness of homework on mathematics achievement at four hierarchical cognitive levels with Hierarchical Linear Model. Results: The results indicated that assigning more homework at the class level could enhance students' mathematics achievement at a low cognitive level, although this effectiveness tended to disappear when the set homework time reached about 1.5-2 h. We did not find evidence that assigning more homework enhances students' mathematics achievement at higher cognitive levels. Conclusions: Overall, this study provides evidence on the effectiveness of homework and offers implications for educational practice and future research.