Detailed white-paper on our results from AVX512 acceleration

By Pradeep Ramachandran

A month or so back, we published a white-paper that gave high-level results on the encoder gains that we saw from our recently implemented AVX512 kernels in x265. We have now published a more detailed paper that tears the results down into more detail, including per-kernel gains for AVX512 over AVX2. The paper can be accessed from here. We hope you find the detailed analysis more informative to understand the reasons behind the results that you see on your hardware.

Do share your findings with us. As you have probably already noticed, AVX512 is part of the recently-released v2.8 of x265!

x265 Receives Significant Boost from Intel Xeon Scalable Processor Family

By Pradeep Ramachandran

Today Intel launched the next generation of Xeon processors, the Intel Xeon Scalable Processor Family (code-named “Purley”), based on the Skylake CPU architecture.  The Intel Xeon Scalable Processor Family is a powerful new generation of 14nm chips which provide significant improvements over the previous generation of Xeon processors (Xeon E5 v4 and E7 v4, code named “Broadwell”), including many fundamental  CPU architectural improvements, a much faster internal data transfer architecture (a mesh architecture with 2x the bandwidth instead of a ring architecture), AVX-512 vector processing, improved cache, and improved I/O architecture with six DDR4 memory channels and 48 PCIe lanes.

With x265 pushing the previous generation processors to the edge for memory bandwidth and threading, the benefits that these new Xeons provide for x265 users will be game changing. Our initial results with the latest build of x265 show a 67% average per-core gain for encoding using HEVC Main profile, and a 50% average gain with Main10 profile across different presets. In particularly, off-line encoding of 4K content is seeing tremendous benefits due to the higher memory bandwidth that the CPUs are able to utilize from cache and system memory. Intel’s Xeon Scalable Processor Family makes x265 and UHDkit the ideal option for a wider range of scenarios including both live and offline HEVC encoding, and they double the performance/cost you’ll get with our software-based encoding libraries.  We’re also seeing significant performance improvements with x264 – roughly 40% higher performance per core on average.

As we enhance x265 to take advantage of the new technologies that these new processors bring to the light, including AVX-512, we expect that users of x265 will love the benefits that they see with these new Xeons.  This even extends to the Core i9 (Skylake-X) consumer processor family, which are based on the same Purley architecture.  Give them a spin, and let us know what you think!