Three-dimensional environmentally sustainable neuromorphic computing system based on natural organic honey-memristor is proposed in this paper. The experimental results indicates the proposed systems have high inference accuracy over 90 % with device variation and nonlinearity. What is more, four conductance drift scenarios, ADC (Analog-to-Digital Converter) quantization effects, and different algorithms (VGG8 and DenseNet-40) are considered to further verify the proposed systems.