Anchor-based multi-view clustering (MVC) has received wide attention due to its effectiveness and scalability. However, most existing works explore prior structure knowledge in the spatial domain for anchors or embeddings. Ignorance of the underlying frequency information guidance may lead to inferior performance. In this paper, we found that it is beneficial for MVC in both clustering accuracy and computation efficiency, utilizing frequency information to guide spatial clustering. Specifically, we transform the multi-view data from the original to the embedding feature space. Subsequently, we construct a third-order tensor and conduct the fast Fourier transform (FFT) along the sample dimension. Using the proposed saliency measurement of the frequency bands, we identify the most influential bands and transform them back to the spatial feature space via inverse FFT for further clustering. Additionally, the subspace learning and partitioning processes are jointly modeled within a unified framework, ensuring mutual enhancement of subspace learning and clustering. We conduct comprehensive experiments on twelve large multi-view datasets. The results verify that the proposed approach exhibits superior or highly competitive performance compared to nine state-of-the-art methods. Furthermore, the efficiency of the proposed methodology is validated, demonstrating its scalability and practicability for large-scale datasets. The code is available at https://github.com/crab8/EFFMVC .