H.265 has been floating around for a long time in video surveillance. It promises greatly reduced bandwidth and decent storage savings which are attractive. So why hasn’t it taken off? And why is it still so problematic?
The math is mathing.
H.265 (HEVC) delivers ~40–50% bitrate reduction compared to H.264 (AVC) for the same perceptual quality, but that efficiency comes at a steep computational cost. The codec achieves compression gains through more complex transforms (up to 35 intra prediction modes vs 9 in H.264), larger coding tree units (CTUs up to 64×64 vs macroblocks at 16×16), variable block partitioning, and significantly heavier motion vector prediction math. For video management systems such as Milestone, Avigilon, Genetec, and Hanwha Wave, this translates directly into higher CPU cycles per frame during decode and analytics stages.
Whatever gains you might get in bandwidth and storage are much more than eaten up by the increase in processing requirements on both servers and workstations.
Analytics workloads—motion detection, object classification, ALPR, and facial recognition etc—typically require full frame decode to the pixel level. With H.265, the increased entropy decoding, de-blocking, and sample adaptive offset filtering introduce additional processing overhead before analytics can even begin. Real-world benchmarks show CPU utilization often increases by 40–70% when moving from H.264 to H.265 at equivalent stream counts and frame rates. And even the most expensive Nvidia GPUs don’t do h.265 decoding at the ASIC level, its still brute force math so GPUs are often even worse on h.265.
While H.265’s tighter compression can reduce storage requirements by up to 30–40%, in surveillance environments the cost curve is rarely favorable: storage is relatively cheap and easily scalable, but compute cycles are finite. The increased CPU and memory demand leads to higher-spec servers, more power draw, and reduced channel density per node. In practice, the processing penalty for H.265 decoding and analytics negates much of the savings from compression—especially when motion analytics are enabled—making H.264 still the more balanced and cost effective choice for most enterprise VMS deployments.
In video surveillance, H.265 streams require intense real-time decoding, which involves both integer and floating-point calculations for motion estimation, compression transforms, and analytics processing. High integer performance accelerates tasks like frame indexing, pixel manipulation, and metadata handling, while strong floating-point throughput enables efficient computation of discrete cosine transforms and motion vector calculations. When paired with high gigahertz CPU cores, these capabilities allow each core to process more H.265 streams simultaneously with lower latency, minimizing frame drops and enabling higher-resolution analytics without overloading the server. This combination directly translates to improved stream density, faster video analytics, and reduced hardware requirements.
