To detect how much genetic variance is accounted for by different genomic regions one first step is to work at the chromosomal level. We used a half sib data structure for two growth traits in sheep as a potentially powerful design to partition the genetic variance across chromosomes. Records for post weaning weight (PW) and scan C site back fat (CF) were used from 5239 merino sheep. The model of analysis accounted for population structure by fitting genetic group effects as well as the numerator relationship matrix or the first five principal components. Different approximations were compared fitting the genomic relationship matrix based on 48599 markers, or on SNPs of an individual chromosome. The correlation between chromosome length and variance explained was PW: 0.53 and CF: 0.70, however significant differences were found between chromosomes, ranging from 0-17.5%. Some chromosomes explained more variance and covariance than expected, under the assumption that it is proportional to the chromosome size; suggesting that some chromosomes harbor more QTL. Some regions show a covariance of opposite sign indicating these loci could be used in selection to ‘break’ an unfavourable correlation (e.g. chromosome 8). These results represent a powerful source of information for genomic selection