Previous studies on photo-realistic volume rendering have failed to optimize for performance with respect to the cache-hierarchy. With this thesis, we consider a tile-based approach for photo-realistic volume rendering, in an effort to improve cache performance and decrease overall execution time. We evaluated the algorithm compared to the traditional approach, with workloads of varying data sizes, resolutions, samples per pixel. Overall we ran 48 serial experiments, and found that the tile-based approach is consistently faster than the traditional approach, including speedups of up to 20%. Additionally we determine that the improvement does not carry forward directly to parallel platforms like Intel TBB.