H.264/AVC compression standard provides tools and solutions for an efficient coding of video sequences of various resolutions. Spatial redundancy in a video frame is removed by use of intra prediction algorithm. There are three block-wise types of intra prediction: 4×4, 8×8 and 16×16. This paper proposes an efficient, low-complexity architecture for intra 16×16 prediction that provides real-time processing of HD video sequences. All four prediction (V, H, DC, Plane) modes are supported in the implementation. The high-complexity plane mode computes a number of intermediate parameters required for creating prediction pixels. The local memory buffers are used for storing intermediate reconstructed data used as reference pixels in intra prediction process. The high throughput is achieved by 16-pixel parallelism and the proposed prediction process takes 48 cycles for processing one macroblock. The proposed architecture is synthesized and implemented on Kintex 705 -XC7K325T board and requires 94 MHz to encode a video sequence of HD 4k×2k (3840×2160) resolution at 60 fps in real time. This represents a significant improvement compared to the state of the art.