The biosynthesis of terpenoids in heterologous hosts has become increasingly popular. Isopentenyl diphosphate (IPP) is the central precursor of all isoprenoids, and the synthesis can proceed via two separate pathways in different organisms: The 1-deoxylulose 5-phosphate (DXP) pathway and the mevalonate (MVA) pathway. In this study, an in silico comparison was made between the maximum theoretical IPP yields and the thermodynamic properties of the DXP and MVA pathways using different hosts and carbon sources. We found that Escherichia coli and its DXP pathway have the most potential for IPP production. Consequently, codon usage redesign, and combinations of chromosomal engineering and various strains were considered for optimizing taxadiene biosynthesis through the endogenic DXP pathway. A high production strain yielding $876{pm}60$ mg/L taxadiene, with an overall volumetric productivity of 8.9 mg/($L{ imes}h$), was successfully obtained by combining the chromosomal engineered upstream DXP pathway and the downstream taxadiene biosynthesis pathway. This is the highest yield thus far reported for taxadiene production in a heterologous host. These results indicate that genetic manipulation of the DXP pathway has great potential to be used for production of terpenoids, and that chromosomal engineering is a powerful tool for heterologous biosynthesis of natural products.