POWER9 + ARM64 Performance For Dav1d 0.8 AV1 Decoding
With last week's release of dav1d 0.8 for CPU-based AV1 video decoding we provided a number of x86_64 benchmarks while questions were raised around the ARM64 and POWER9 performance. Here are such benchmarks for those wondering about the AV1 video decoding speed on those architectures.
To satisfy readers wondering around them I ran some follow-up tests on the Talos II 2P server with 44 cores / 176 threads and then for ARM64 was the Altra in its 160 core 2P configuration. For these quick reference tests are also performance numbers for the Xeon Platinum 8280 2P and EPYC 7742 2P. Just some quick reference tests for those curious how things are looking for dav1d decoding on other CPU architectures...
1080p with these high core count servers tend to be not too interesting and difficult to scale well but at least for the Summer Nature on EPYC and Altra were some nice improvements while the Intel Cascade Lake and IBM POWER9 were flat with the new release.
For all but POWER9 were nice 4K uplift with the new dav1d 0.8 series.
The 10-bit content encode performance was flat but interesting to note ARM64 performing much greater than the rest.
More dav1d benchmark results for various CPUs can be found over on OpenBenchmarking.org.
To satisfy readers wondering around them I ran some follow-up tests on the Talos II 2P server with 44 cores / 176 threads and then for ARM64 was the Altra in its 160 core 2P configuration. For these quick reference tests are also performance numbers for the Xeon Platinum 8280 2P and EPYC 7742 2P. Just some quick reference tests for those curious how things are looking for dav1d decoding on other CPU architectures...
1080p with these high core count servers tend to be not too interesting and difficult to scale well but at least for the Summer Nature on EPYC and Altra were some nice improvements while the Intel Cascade Lake and IBM POWER9 were flat with the new release.
For all but POWER9 were nice 4K uplift with the new dav1d 0.8 series.
The 10-bit content encode performance was flat but interesting to note ARM64 performing much greater than the rest.
More dav1d benchmark results for various CPUs can be found over on OpenBenchmarking.org.
13 Comments