Appling Malmquist index method to calculate total factor productivity (TFP) with a sample of 29 provinces in China from 1995 to 2008, based on following two scenarios: taking CO2 emissions as an undesirable output or not. The efficiency of this process is calculated by Data Envelopment Analysis (DEA) formulations, related to GDP and CO2 as outputs, capital and labor as inputs. Comparing the results of the two scenarios, we analyze the differences of TFP, its determinants and growth mode among the 29 provinces. Results show that: significant regional differences in TFP exist, and under the two scenarios, the eastern China has the highest TFP, followed by middle China and western China. Moreover, the slow-down of the TFP growth in scenario 2 is mainly because that the technology progress is discounted by the environmental factors. Especially, the growth method of western China illustrates “environmental-extensive”. As a result, China’s governments must take into full consideration regional differences of TFP and carbon reduction potentials, so as to realize low-cost or no incremental cost reduction.