As one of the key components in the non-volatile full adder (NV-FA), spin transfer torque assisted voltage controlled magnetic anisotropy magnetic tunnel junction (STT assisted VCMA-MTJ) will possess superior development prospects in internet of things, artificial intelligence and other fields due to its fast switching speed, low power consumption and good stability. However, with the downscaling of magnetic tunnel junction (MTJ) and the improvement of chip integration, the effects of process deviation on the performances of MTJ device as well as NV-FA circuit become more and more important. Based on the magnetization dynamics of STT assisted VCMA-MTJ, a new electrical model of STT assisted VCMA-MTJ, in which the effects of the film growth variation and the etching variation are taken into account, is established to study the effects of the above deviations on the performances of MTJ device and NV-FA circuit. It is shown that the MTJ state fails to be switched under the free layer thickness deviation γ_tf ≥ 6% or the oxide layer thickness deviation γ_tox ≥ 0.7%. The sensing margin (SM) is reduced by 17.5% as the tunnel magnetoresistance ratio deviation β increases to 30%. The writing error rate can be effectively reduced by increasing V_b1, and increasing V_b2 when writing ‘0’ or reducing V_b2 when writing ‘1’ in the NV-FA circuit. The output error rate can also be effectively reduced by increasing the driving voltage of logical operation V_dd.