Raphael Canut
2018-Dec-11 16:57 UTC
[theora] NHW Project - speed measurement 2 - very high compression
Hello, If it could be of interest, here are new speed measurement for -l13 very high compression setting. For x265 (HEVC), I use BPG codec, and this time I use -m 0 setting which uses the fastest encoding setting of x265.I run the .exe 10x times and I pick the best scores. So here are the timings at -l13 very high compression: Encode time: x265 (BPG -m 0): 80ms NHW: 20ms Decode time: x265 (BPG): 90ms NHW: 10ms So with an equivalent level of optimization (x4-5, multithreading, SIMD,...), the NHW Project is 80/20x5=20x faster to encode and 90/10x5=45x faster to decode than x265 (optimized HEVC)! Yes, 45x faster to decode than x265 at very high compression!, this would really save on battery life for mobile devices! Else also, with the new improvements I finally find the NHW Project competitive with x265 (HEVC) on nearly 90%-95% of the images I have tested at -l13 very high compression, because it has more neatness. I will continue to go down to very high compression, but it is not that easy, anyway you can have a first idea with -l13 very high compression setting on my demo page ( http://nhwcodec.blogspot.com/ ). Many thanks! Cheers, Raphael -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.xiph.org/pipermail/theora/attachments/20181211/58ac2b56/attachment.html>
Possibly Parallel Threads
- NHW Project - good improvement of very high compression (-l8 to -l13 quality settings)
- NHW Project - good quality improvement & encoding speed improvement
- NHW Project - new -l13 very high compression quality setting
- NHW Project - more test with x265 (HEVC) codec
- NHW Project - Image comparison