Fusarium head blight (FHB) is one of the most destructive diseases of durum wheat. Therefore, reducing losses in yield, quality and preventing mycotoxin contaminations are of major importance, as durum wheat is mostly used for human consumption. While growing resistant varieties is the most effective approach for managing this fungal disease, enhancing FHB resistance in durum wheat is hampered by the limited genetic variation in the elite gene pool and difficulties in efficiently combining the numerous small-effect resistance quantitative trait loci (QTL) in the same breeding line. We evaluated an international collection of 228 genotyped durum cultivars and breeding lines from Northern America, the Mediterranean, Central Europe, Australia and CIMMYT. The elite durum lines were phenotyped for FHB resistance over three years to investigate the genetic architecture and the potential of genomic-assisted breeding. In addition, to broaden the genetic basis for FHB resistance in durum, we introgressed resistance alleles from wild and cultivated relatives (T. aestivum, T. dicoccoides, T. dicoccum) and developed 750 multi-parental lines. These pre-breeding lines were phenotyped over three years for FHB resistance and genotyped for the introduced resistance QTL. Although a lack of highly resistant lines was evident for both collections, broad variation was found, including many moderately resistant pre-breeding lines. Plant height strongly influenced FHB resistance levels and led to co-localization of plant height and resistance QTL. Nevertheless, a major QTL on chromosome 3B independent of plant height was identified in the elite durum wheat gene pool. Interestingly the QTL mapped in the same chromosomal interval as reported for the prominent hexaploid resistance QTL Fhb1, though haplotype analysis highlighted the distinctiveness of both QTL. In the durum prebreeding material markers indicative for ‘exotic’ resistance QTL were significantly associated with FHB resistance confirming their successful introgression. Comparison between phenotypic and genomic selection for FHB resistance in the elite germplasm revealed a superior prediction ability of the former. However, simulated selection experiments resulted in higher selection responses when using genomic breeding values for early generation selection. An earlier identification of the most promising lines was furthermore feasible with a genomic selection index, which suggested a much faster short-term population improvement than previously possible in durum. In the long-term, exotic germplasm can broaden the genetic base for FHB resistance beyond the capabilities of elite material for achieving higher levels of resistance.