Quantization has emerged to be an effective way to significantly boost the performance of deep neural networks (DNNs) by utilizing low-bit computations. Despite having lower numerical precision, quantized DNNs are able to reduce both memory bandwidth and computation cycles with little losses of accuracy. Integer GEMM (General Matrix Multiplication) is critical to running quantized DNN models efficiently, as GEMM operations often dominate the computations in these models. Various approaches have been developed by leveraging techniques such as vectorization and memory layout to improve the performance of integer GEMM. However, these existing approaches are not fast enough in certain scenarios. We developed NGEMM, a compiler-based GEMM implementation for accelerating lower-precision training and inference. NGEMM has better use of the vector units by avoiding unnecessary vector computation that is introduced during tree reduction. Our experimental results showed that NGEMM outperformed the state-of-art BLAS libraries such as MKL by an average of 1.4x. We have applied NGEMM to a number of production services in Microsoft