SVT-VP9 Is Intel's Latest Open-Source Video Encoder Yielding High Performance VP9

[phoronix]

Phoronix: SVT-VP9 Is Intel's Latest Open-Source Video Encoder Yielding High Performance VP9

At the start of the month Intel open-sourced SVT-AV1 aiming for high-performance AV1 video encoding on CPUs. That complemented their existing SVT-HEVC encoder for H.265 content and already SVT-AV1 has been seeing nice performance improvements. Intel now has released SVT-VP9 as a speedy open-source VP9 video encoder...

http://www.phoronix.com/scan.php?page=news_item&px=SVT-VP9-Open-Source

[microcode]

It'd be interesting to see SVT-AV1 on Epyc.

[xiando]

I'd love to see a follow-up / addition to this which goes beyond just benchmarking the pure encoding speed. A visual quality comparison would be nice. This is of course a bit hard to do since it's somewhat subjective and showing frames from the encoded files by means of compressed jpg images isn't ideal. Still, I'd love to get an idea of how the results compare. As an example, the hardware HEVC encoder in current AMD GPUs will encode files really fast compared to x265 but the output files are twice as big with half the visual quality (that's just my opinion after doing some testing) which makes it a non-option.

On a completely different note, what's with the pretty insane (compared to libvpx) memory requirements for Intel's encoders? Tried their AV1 encoder and it ate like 15 GB RAM encoding a 1080p file.

[M@yeulC]

Nice to see some progress on that front. However, I can't scratch the feeling that it could be faster on AMD... Well, given it's an Intel library, I won't blame them for optimizing it for their products first, but there are probably a few low hanging fruits to be caught on AMD's side.

[RussianNeuroMancer]

there are probably a few low hanging fruits to be caught on AMD's side.

I wonder what would they do if AMD create push request which would make it work on AMD much faster than on Intel...

[dibal]

I wonder what would they do if AMD create push request which would make it work on AMD much faster than on Intel...

Maybe it is AVX-128 vs. AVX-512. Then the patch are expanded vector units.

[andreano]

There goes the business case for Two Orioles' EVE encoder, which was only affordable to big money makers...

... Finally a usable encoder that is also usable for the masses!

[milkylainen]

ICC vs GCC on an Intel written library maybe?
It'd be interesting to see if ICC is much better at wide AVX vectorization.

[Michael]

ICC vs GCC on an Intel written library maybe?
It'd be interesting to see if ICC is much better at wide AVX vectorization.

Tests were built from source on same compiler (GCC).

[milkylainen]

Tests were built from source on same compiler (GCC).

Yes, I understood that. It was meant as a comment on an interesting thing to test.