https://dwzz.cn/j8Il3A
The following inputs have been produced from TX Ready XDCAM-50 files:
Sample 1: ColorBars FULL HD H266 YV12 BT709 SDR 10bit.ts
Sample 2: Chicago Fire FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
Sample 3: Judas and the Black Messiah SD H266 Main10 3Mbits 25p YV12 BT601 SDR 10bit AC3 2ch 48000Hz.ts
Sample 4: The Creator FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
The decoding has been carried out using the following hardware:
A) Intel Xeon E5-2637 v4 3.50GHz dual socket 8c/16th, 64GB RAM
B) Intel i7 5930K 3.50GHz single socket 6c/12th 32GB RAM
C) Intel Xeon Gold 6238R 2.20GHz dual socket 56c/112th 128GB RAM
The tests have been conducted both using FFMpeg's new decoder and Fraunhofer's VVDec standalone decoder (i.e WITHOUT using the FFMpeg integration patch).
Colorbars have been skipped in the benchmark decoding test due to the fact that it's quite literally the same frame repeated all over again and it would yield no significant speed outcome.
Now, on to the FFMpeg decoder results:
Intel Xeon E5-2637 v4 3.50GHz dual socket 8c/16th, 64GB RAM
Sample 2: Chicago Fire FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU constantly at around 89-100%
fps= 35
Sample 3: Judas and the Black Messiah SD H266 Main10 3Mbits 25p YV12 BT601 SDR 10bit AC3 2ch 48000Hz.ts
CPU oscillated between 78% and 93% without ever touching full 100%.
fps=194
Sample 4: The Creator FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU oscillated between 41% to 76% without ever touching full 100%.
fps= 24
Intel i7 5930K 3.50GHz single socket 6c/12th 32GB RAM
Sample 2: Chicago Fire FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps= 34
Sample 3: Judas and the Black Messiah SD H266 Main10 3Mbits 25p YV12 BT601 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps=193
Sample 4: The Creator FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU oscillated between 65% and 93% with occasional very rare spikes to 100%
fps= 28
Intel Xeon Gold 6238R 2.20GHz dual socket 56c/112th 128GB RAM
Sample 2: Chicago Fire FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 23% overall with cores being used on one single socket
fps= 44
Sample 3: Judas and the Black Messiah SD H266 Main10 3Mbits 25p YV12 BT601 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 13% overall with cores being used on one single socket
fps=249
Sample 4: The Creator FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 18% overall with cores being used on one single socket
fps= 26
And now on to the VVDec results:
Intel Xeon E5-2637 v4 3.50GHz dual socket 8c/16th, 64GB RAM
Sample 2: Chicago Fire FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps=111
Sample 3: Judas and the Black Messiah SD H266 Main10 3Mbits 25p YV12 BT601 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps=618
Sample 4: The Creator FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps=119
Intel i7 5930K 3.50GHz single socket 6c/12th 32GB RAM
Sample 2: Chicago Fire FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps=108
Sample 3: Judas and the Black Messiah SD H266 Main10 3Mbits 25p YV12 BT601 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps=634
Sample 4: The Creator FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 100%
fps=115
Intel Xeon Gold 6238R 2.20GHz dual socket 56c/112th 128GB RAM
Sample 2: Chicago Fire FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 65% on one single socket
fps=226
Sample 3: Judas and the Black Messiah SD H266 Main10 3Mbits 25p YV12 BT601 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 55% on one single socket
fps=722
Sample 4: The Creator FULL HD H266 Main10 12Mbits 25p YV12 BT709 SDR 10bit AC3 2ch 48000Hz.ts
CPU steadily at 66% on one single socket
fps=211
This shows that not only VVDec is more optimized and makes use of the CPU resources in a better way (being able to scale even with CPUs that have a lot of cores, although not quite in every dual socket configuration) but the fps results tell the whole story, it's just faster. Before I made some claims about libav (i.e FFMpeg) being faster, however that was truly a mistake in my testing as it was just me not using VVDec directly but rather through an old unapproved patch via FFMpeg which was totally not optimized. Now that I compared libav via FFMpeg and VVDec using the native Fraunhofer executable, the whole story is completely different and VVDec appears to be miles ahead.
On top of that, upon further testing, I had to throw away some samples as they reported the following error while trying to decode them with libav/FFMpeg:
frame 78, P( 14, 1) failed with -1163346256
Intra Block CopyIntra Block Copy is not implemented.
your file has a feature which has not been implemented.
Decoding error: Not yet implemented in FFmpeg, patches welcome
yet, VVDec managed to decode even those files without the blink of an eye.
Nonetheless, here are the results of my tests:

ffmpeg的“原生”vvc解码实现和fraunhofer的vvdec的对比嘛，不是啥新鲜东西了
他们邮件列表还是啥的地方也有在讨论说为什么av1用外部解码库就可以，vvc的外部解码库ptach就不给人通过非要等人给他们写原生实现

结果回应就是（部分或某个）管理成员的“个人喜好”以及av1的原生解码“有计划做”