While this round-up did not include av1_qsv due to not having an Intel ARC card, the results from av1_nvenc shows the encoder leading all hardware encoders so far, but being limited in bitrate and quality in the CQ mode, while still lagging behind libx265 in most instances.
The motivation for hardware encoding is still strong, because of the inordinate amount of CPU time required for a good AV1 encode. I have access to team red’s video encoding hardware, so let’s see how it compares with team green.
Encoding was performed on the Windows platform on the AMD Radeon 780M in my Lenovo Legion 5 Slim. The encoding options were as follows:
-rc 0 -qp_p [qp] -qp_i [qp] -quality [quality] -header_insertion_mode 1
Encoding is performed in Quantisation Parameter (QP) mode which is not as sophisticated or as efficient as constant quality/rate-factor bitrate control modes as AMF doesn’t appear to support CQ or CRF at this time. While many quality-based modes change QP dynamically based on picture type, I’ve opted to have the same QP for P and I frames. There appears to be no options for B-frames, suggesting perhaps no B-frame support. I did try the use of the -preanalysis option, but unfortunately, I received some errors (suggesting some kind of queue full overflow) during the encode and so I decided against using it. The header insertion mode flag is set to make the file seekable in a media player. Only QP values from 8 to 248 in steps of 8 have been tested.
One issue appears that the AMD video encoder operates on a fixed block size which results in 1920×1080 sources being encoded as 1920×1082 which may cause issues. Another thing to note is the [email protected] profile which is much higher than the [email protected] created by libaom_av1 and av1_nvenc.
The encoded files had the following properties in MediaInfo:
Video ID : 1 Format : AV1 Format/Info : AOMedia Video 1 Format profile : [email protected] Codec ID : V_AV1 Duration : 3 min 46 s Width : 1 920 pixels Height : 1 080 pixels Original height : 1 082 pixels Display aspect ratio : 16:9 Frame rate mode : Constant Frame rate : 23.976 (24000/1001) FPS Color space : YUV Chroma subsampling : 4:2:0 Bit depth : 8 bits Writing library : Lavc60.18.100 av1_amf Default : No Forced : No Color range : Limited
While it is noted the actual width and height are 1920×1080, ffmpeg appears to decode the full 1920×1082 from the file causing issues with video quality metrics comparisons. Therefore, I had to use a different command line to perform quality tests that would crop out the top 1920×1080 frame from the decoded result –
ffmpeg -i [distorted] -i [reference] -lavfi [0:v]crop=1920:1080:0:0[main]; [main][1:v]libvmaf=log_path=[log]:feature='name=float_ssim|name=psnr' -f null -
Video Quality vs QP at Various Quality
This section compares the encode result based on different quality settings, which is the AMF equivalent of preset or encoding speed.
It appears the PSNR has a breakpoint around QP=92, below which the PSNR increases almost exponentially, above which the PSNR decreases nearly linearly. A PSNR of 45dB is achieved at QP=56 and a PSNR of 40dB is achieved at QP=128. There is almost no difference with different quality settings.
The SSIM score seems to show a classical trajectory, with all quality settings being closely bunched together. An SSIM of 0.99 is achieved at QP=144, 0.995 at QP=112 and 0.999 at QP=32.
The VMAF score shows a similar trajectory, seemingly tailing off around QP=136. A VMAF of 99 is achieved at QP=88 and 99.5 at QP=56.
Best Video Quality vs QP
Encoding at quality 0 is considered “best”. The following results show the results for quality 0 only.
It seems that below QP=56, it seems the worst frame results improve faster than the best frame. Above QP=232, the best frame results take a dive, otherwise everything seems to scale somewhat proportionally.
The SSIM scores trend rather smoothly with QP, although the best frame SSIM does degrade visibly on the graph with higher QP, suggesting the high QP results may look rather poor.
Based on VMAF score, it seems that worst frames hit a “floor” at lower QPs, which improves slower below QP=104. Above that threshold, it seems worst frames tail off linearly. This suggests that perhaps QP=56 to QP=104 is a sane range to use.
Bitrate vs QP at Various Quality
Unlike the av1_nvenc encode, it seems the bitrate does scale with QP value, being very similar across different encoding speed (quality) selections. The birate ranges from a healthy 92.1Mbit/s down to 0.5Mbit/s.
Video Quality vs Bitrate at Various Presets + Cross-Codec
The interesting part is to see how it compares across other codecs.
The PSNR result is not great, as it overlaps and tails libvpx-vp9, with even hevc_nvenc being quite comparable. The av1_nvenc results still lead by a notable margin.
The SSIM metrics seem to rate this codec even worse than hevc_nvenc and comparable to and occasionally worse than libx264. That’s not a good look.
Based on VMAF, at low bitrates, it is only a match for hevc_nvenc, while at higher bitrates, it is only a match for libx264. It consistently trails libvpx-vp9 and av1_nvenc. How unfortunate.
Image Comparison
Let’s take a look at some of the images.
- Frame 844
- At QP=40, it is clear that the texture/grain in the gradient has been smoothed out. At QP=104, some edge ringing and noise is apparent. Dark hair details seem crushed between QP=72~88.
- Frame 1347
- At QP=40, the background bokeh looks a bit splotchy. At QP=72, some background hair is lost and some blockiness/ringing is visible. Foreground hair detail is also somewhat clumped at QP=72.
- Frame 3733
- At QP=40, hair detail is noise reduced and softened, by QP=56 there is some sort of moire pattern emerging and by QP=72, it is very smooshed. Blockiness is emerging from QP=72~88.
- Frame 4415
- At QP=40, fine hair detail seems to have been smoothed out and texture in the background smoke is softened. At QP=72~88, the arm hair shape seems to have changed and smoothed almost as if partly shaved. Crown hair is also smoothed at QP=40, looking almost like comic-book plastic by QP=56.
The results are a bit interesting – there is quite a bit of smoothing and ringing/noise, but it seems perhaps QP=40 and QP=56 might be a good range for non-critical viewing with a good level of quality.
Conclusion
It would seem that av1_amf in QP mode is not as efficient as av1_nvenc in CQ mode and in fact, really only reaches hevc_nvenc and libx264 levels of compression efficiency depending on the metric. This level of performance, while better than some hardware encoders, is not really competitive with software encoders of even a previous generation which makes it unattractive for non-time-critical scenarions (e.g. offline transcoding).
Viewing of the images suggests that perhaps QP=40 (23.4Mbit/s) to QP=56 (16.2Mbit/s) is a good range for quality although there is quite a bit of smoothing in the images. At higher QPs, it seems that moire patterns, ringing on sharp edges becomes apparent.
Unfortunately, the encoder is producing files with 1920×1082 resolution that have 1920×1080 active area which may cause problems depending on the intended use as some devices or services may reject it.
Encode Stats (for this Part)
Total Number of Encodings: 124
Total Size: 43,627,667,440 bytes
Encode Stats (for all Parts up to now)
Total Number of Encodings: 5,619
Total Size: 1,973,359,126,157 bytes
This post is part of a series. See the main index at the bottom of Part 0.
Appendix: Raw Data
QP,Quality,Bitrate (Mbit/s),psnr_y_min,psnr_y_max,psnr_y_mean,psnr_y_hmean,float_ssim_min,float_ssim_max,float_ssim_mean,float_ssim_hmean,vmaf_min,vmaf_max,vmaf_mean,vmaf_hmean 8,0,91.19838532,51.145647,59.729353,52.199957,52.19308,0.999329,0.99997,0.99977,0.99977,95.822305,100,99.93567,99.934734 16,0,55.56006343,48.560406,58.460542,49.869833,49.851229,0.998396,0.999935,0.999523,0.999523,95.270446,100,99.892426,99.890636 24,0,39.02517361,46.095996,57.770056,48.428542,48.391062,0.997642,0.999884,0.999277,0.999277,94.702116,100,99.838733,99.835698 32,0,28.72117504,44.249227,57.07534,47.275667,47.213937,0.997235,0.999857,0.999009,0.999009,94.280299,100,99.767696,99.762675 40,0,22.45001171,42.894573,56.511581,46.366746,46.281454,0.996955,0.999769,0.99873,0.998729,93.734237,100,99.690514,99.682909 48,0,18.2725049,41.600482,56.15837,45.627066,45.520444,0.996604,0.999705,0.998438,0.998438,93.252175,100,99.603465,99.592392 56,0,15.46740024,40.601471,55.680244,45.052235,44.929722,0.99601,0.999644,0.998169,0.998169,92.685907,100,99.519864,99.505073 64,0,13.04297424,39.600407,55.292626,44.484711,44.345567,0.995181,0.999572,0.99785,0.99785,91.895134,100,99.384537,99.363499 72,0,11.44938343,38.847378,55.04533,44.070544,43.917802,0.994501,0.999526,0.99757,0.997569,91.522583,100,99.276557,99.250283 80,0,10.06671667,38.121094,54.718522,43.617087,43.45048,0.993585,0.999452,0.997238,0.997238,90.648298,100,99.136044,99.102796 88,0,9.088651247,37.547502,54.538566,43.24758,43.070412,0.992587,0.999399,0.996933,0.996932,90.235386,100,99.000523,98.960073 96,0,8.15473313,36.954965,54.270759,42.840035,42.651107,0.991672,0.99934,0.996559,0.996558,89.821942,100,98.829512,98.779635 104,0,6.859467478,36.137738,53.865897,42.176661,41.970352,0.989461,0.999187,0.995842,0.995841,86.974021,100,98.48491,98.414514 112,0,5.97129645,35.462098,53.502695,41.59153,41.369237,0.987598,0.999077,0.995108,0.995107,86.698069,100,98.088245,97.992685 120,0,5.296564528,34.859451,53.122889,41.070185,40.833909,0.985592,0.998956,0.994343,0.994342,85.542947,100,97.653517,97.529058 128,0,4.53498521,34.102916,52.534316,40.382039,40.128642,0.982373,0.99883,0.99313,0.993128,83.438299,100,96.819227,96.636605 136,0,4.02039399,33.463385,52.553473,39.846073,39.57636,0.979989,0.998689,0.991963,0.99196,81.578899,100,96.031388,95.792782 144,0,3.458229914,32.716938,52.293983,39.162697,38.875779,0.975551,0.998494,0.990284,0.99028,77.673783,100,94.758913,94.438337 152,0,3.005923854,32.012177,51.886936,38.529904,38.227511,0.971132,0.998289,0.988384,0.988378,73.755875,100,93.226079,92.816738 160,0,2.612595942,31.2191,51.75824,37.882932,37.566275,0.965665,0.998043,0.986094,0.986084,69.286655,100,91.292269,90.782043 168,0,2.268324819,30.357776,51.36647,37.228236,36.897298,0.958883,0.997681,0.983377,0.983361,64.084264,100,88.952195,88.33469 176,0,1.972374792,29.494743,50.78169,36.564643,36.221618,0.951374,0.997301,0.980056,0.980031,59.285849,100,86.135858,85.412892 184,0,1.709412268,28.661608,50.326307,35.895666,35.542368,0.942838,0.996856,0.976132,0.976094,53.765289,100,82.876937,82.066777 192,0,1.484928402,27.875049,49.896887,35.225235,34.861962,0.933086,0.99623,0.971451,0.971393,48.618631,100,78.980586,78.090391 200,0,1.286114657,27.063146,49.079862,34.557637,34.185294,0.921527,0.995413,0.965863,0.965776,42.77175,100,74.999774,74.019476 208,0,1.111643528,26.284447,48.785663,33.903064,33.518824,0.909715,0.994742,0.959348,0.959217,37.956812,100,70.829261,69.748358 216,0,0.961316743,25.478638,48.260123,33.211025,32.819365,0.897913,0.994569,0.95156,0.951367,33.902533,100,66.454097,65.254066 224,0,0.825834974,24.637232,47.506995,32.443094,32.047608,0.87958,0.994261,0.941656,0.941365,31.233355,100,61.493968,60.125468 232,0,0.693567832,23.792384,46.50287,31.661553,31.258204,0.855903,0.993209,0.928995,0.928557,28.381024,100,55.928468,54.325696 240,0,0.574282017,22.843195,45.271327,30.725442,30.321986,0.828728,0.99243,0.912029,0.911335,25.216908,100,49.593229,47.656014 248,0,0.456708943,21.720692,43.067141,29.674776,29.271572,0.787561,0.990677,0.888339,0.887201,16.97415,100,42.26571,39.813873 8,30,92.0962412,51.036203,59.719589,52.123036,52.115781,0.999293,0.999969,0.999764,0.999764,95.882564,100,99.933831,99.932861 16,30,56.17449041,48.373745,58.460565,49.782864,49.763249,0.998284,0.999933,0.999507,0.999507,95.187938,100,99.888767,99.886903 24,30,39.36837813,45.892166,57.772755,48.337611,48.298416,0.99751,0.999881,0.999253,0.999253,94.738633,100,99.833024,99.829843 32,30,28.87582706,44.062435,57.07723,47.174932,47.110631,0.997157,0.999857,0.998976,0.998976,94.409358,100,99.759348,99.75407 40,30,22.55705363,42.75217,56.509648,46.271701,46.183196,0.99689,0.999769,0.998692,0.998692,93.656253,100,99.680762,99.672802 48,30,18.34399014,41.423455,56.146922,45.531617,45.42118,0.996558,0.999704,0.998395,0.998395,93.259285,100,99.591551,99.579985 56,30,15.52021717,40.393247,55.683581,44.959897,44.83336,0.996029,0.999646,0.998123,0.998123,92.685907,100,99.507366,99.492021 64,30,13.07314302,39.390962,55.301067,44.398162,44.254871,0.99517,0.99957,0.997799,0.997799,91.895134,100,99.366234,99.344438 72,30,11.47243381,38.632764,55.053572,43.98304,43.826137,0.994493,0.999524,0.997516,0.997515,91.522583,100,99.255798,99.228614 80,30,10.09139811,37.936083,54.714247,43.534768,43.363888,0.993579,0.999452,0.997179,0.997179,90.648298,100,99.106414,99.071782 88,30,9.110814068,37.313153,54.500496,43.165759,42.984162,0.992826,0.9994,0.99687,0.99687,90.235386,100,98.970048,98.928317 96,30,8.172331092,36.752637,54.271775,42.75785,42.564483,0.991656,0.999333,0.99649,0.99649,89.821942,100,98.790868,98.739435 104,30,6.868757369,35.967877,53.875776,42.095711,41.884838,0.989463,0.999183,0.995767,0.995766,87.1123,100,98.435719,98.362892 112,30,5.980205559,35.330607,53.499629,41.510886,41.283878,0.987591,0.999073,0.995022,0.995022,86.702798,100,98.030013,97.931637 120,30,5.301896712,34.739153,53.111203,40.988701,40.74751,0.985593,0.998966,0.994245,0.994244,85.539688,100,97.574454,97.445926 128,30,4.537318038,33.982399,52.531609,40.300872,40.042585,0.982395,0.998848,0.993002,0.993,83.438299,100,96.698503,96.509012 136,30,4.021937475,33.360763,52.556948,39.766522,39.491996,0.979848,0.9987,0.991816,0.991814,81.589461,100,95.870493,95.623262 144,30,3.457377943,32.610666,52.287144,39.082128,38.790381,0.975474,0.998482,0.990106,0.990101,77.37527,100,94.545949,94.215636 152,30,3.001972084,31.89336,51.937065,38.450928,38.143403,0.971207,0.998299,0.988181,0.988174,73.717281,100,92.982199,92.561132 160,30,2.605856519,31.097374,51.676985,37.800791,37.479144,0.965655,0.998005,0.985856,0.985846,69.278155,100,91.000242,90.475997 168,30,2.259704778,30.189308,51.142792,37.141829,36.805685,0.958752,0.997608,0.983067,0.983051,63.954918,100,88.601138,87.966688 176,30,1.962821492,29.346181,50.83617,36.480327,36.131626,0.951378,0.997272,0.979691,0.979665,59.190292,100,85.734762,84.994879 184,30,1.698594266,28.523062,50.289033,35.805756,35.447074,0.942923,0.996879,0.975675,0.975635,54.160318,100,82.413284,81.583368 192,30,1.472254733,27.710653,49.859677,35.132893,34.763863,0.932962,0.996252,0.970886,0.970824,48.547038,100,78.449863,77.536503 200,30,1.272683312,26.933971,49.21957,34.461425,34.083224,0.921543,0.995382,0.965168,0.965076,42.716768,100,74.428039,73.419488 208,30,1.098010271,26.139984,48.581345,33.799585,33.409263,0.909717,0.994773,0.958472,0.958333,37.900979,100,70.164679,69.045991 216,30,0.948504319,25.357215,47.992238,33.10654,32.708736,0.89794,0.994303,0.950525,0.950319,33.886181,100,65.752482,64.505572 224,30,0.813970935,24.513907,47.673364,32.338266,31.936643,0.879324,0.994321,0.940436,0.940127,31.249007,100,60.734718,59.314306 232,30,0.683727727,23.648144,46.857029,31.564131,31.15475,0.85575,0.993305,0.927629,0.927165,28.360038,100,55.240071,53.575847 240,30,0.567081761,22.725132,45.118955,30.638608,30.230102,0.826325,0.992398,0.910648,0.909922,23.764968,100,49.01329,47.028701 248,30,0.452426472,21.679101,42.881305,29.611824,29.204622,0.783497,0.990829,0.887158,0.885984,16.392362,100,41.888245,39.363021 8,70,92.13425711,50.92106,59.58193,51.978869,51.971372,0.999196,0.999968,0.999742,0.999742,95.702354,100,99.929692,99.928653 16,70,56.66397808,48.195595,58.431446,49.712168,49.691728,0.998283,0.999932,0.9995,0.9995,95.316399,100,99.885137,99.883198 24,70,40.0180961,45.865774,57.75236,48.303092,48.262927,0.997616,0.999877,0.999258,0.999258,94.813489,100,99.830802,99.82755 32,70,29.6332024,44.104353,57.056983,47.152955,47.088158,0.997192,0.999858,0.998987,0.998987,94.300091,100,99.758683,99.75338 40,70,23.3606172,42.741312,56.50171,46.267272,46.178797,0.996974,0.999762,0.998723,0.998723,93.901905,100,99.684641,99.67687 48,70,19.05180627,41.454123,56.115363,45.524272,45.414636,0.996656,0.999707,0.998429,0.998429,93.212479,100,99.595628,99.584266 56,70,16.18793191,40.486197,55.677934,44.944073,44.818532,0.996073,0.999644,0.998156,0.998155,92.684308,100,99.510698,99.495565 64,70,13.63901868,39.514647,55.249823,44.393344,44.251395,0.995158,0.999571,0.99784,0.99784,91.840478,100,99.371119,99.349486 72,70,11.9567563,38.787193,55.042511,43.978065,43.822641,0.994451,0.99952,0.997552,0.997552,91.404602,100,99.261242,99.234304 80,70,10.50762058,38.071209,54.694256,43.518851,43.349026,0.993547,0.999456,0.99721,0.997209,90.560329,100,99.110275,99.075828 88,70,9.476054102,37.491994,54.504892,43.150048,42.96919,0.992784,0.999401,0.996907,0.996907,90.179467,100,98.97017,98.928363 96,70,8.48998081,36.902073,54.28006,42.737889,42.54517,0.991771,0.999332,0.996522,0.996522,89.573602,100,98.79532,98.743951 104,70,7.149586354,36.069249,53.854443,42.077252,41.866599,0.989534,0.999186,0.995813,0.995812,88.038893,100,98.450396,98.378759 112,70,6.224670049,35.385719,53.421685,41.479587,41.25196,0.987589,0.999079,0.99506,0.995059,86.548433,100,98.046063,97.949076 120,70,5.514588214,34.772859,53.000765,40.957597,40.716161,0.985551,0.998975,0.994289,0.994288,85.273099,100,97.601119,97.474497 128,70,4.718884066,34.015067,52.54284,40.268586,40.009915,0.982199,0.998839,0.993052,0.99305,83.10187,100,96.733137,96.546164 136,70,4.177366929,33.403923,52.459907,39.734508,39.460024,0.979844,0.998713,0.991875,0.991872,80.924807,100,95.919984,95.676861 144,70,3.594205137,32.65867,52.084696,39.04157,38.749751,0.975335,0.998495,0.990162,0.990157,76.913054,100,94.593642,94.268264 152,70,3.116118015,31.942904,51.833133,38.411343,38.103945,0.97083,0.998285,0.988243,0.988237,72.743408,100,93.025758,92.611023 160,70,2.706232099,31.050213,51.566863,37.766101,37.444385,0.965177,0.998015,0.985967,0.985956,68.770348,100,91.068737,90.555664 168,70,2.342073904,30.161052,51.046536,37.10594,36.770159,0.958445,0.997673,0.983216,0.983199,63.025381,100,88.676854,88.056732 176,70,2.037712039,29.320349,50.829977,36.446295,36.098671,0.951128,0.997273,0.979883,0.979857,58.872073,100,85.802806,85.07917 184,70,1.761105938,28.495093,50.294472,35.77895,35.421817,0.942362,0.996923,0.975936,0.975898,53.488314,100,82.49814,81.693179 192,70,1.525044768,27.717665,49.727543,35.110721,34.743697,0.932333,0.996218,0.971181,0.971122,48.020932,100,78.499925,77.619698 200,70,1.316947247,26.917065,49.052454,34.445512,34.070505,0.920444,0.995294,0.965568,0.96548,42.499926,100,74.495604,73.530253 208,70,1.13337213,26.114956,48.425896,33.783922,33.398601,0.908476,0.99434,0.958951,0.95882,38.173163,100,70.201003,69.137361 216,70,0.97752694,25.315366,47.829896,33.093283,32.70165,0.896362,0.994127,0.95107,0.950877,33.960278,100,65.76387,64.576364 224,70,0.838986016,24.494639,47.151308,32.33031,31.937034,0.874989,0.993516,0.940989,0.9407,31.060712,100,60.741519,59.397211 232,70,0.704697132,23.635498,46.587834,31.566934,31.167096,0.851899,0.993021,0.928338,0.927908,28.294251,100,55.255473,53.691785 240,70,0.584931682,22.715953,45.101173,30.652696,30.256112,0.824828,0.991914,0.911517,0.910847,25.008545,100,49.062693,47.189603 248,70,0.466199227,21.67321,42.762394,29.62917,29.236327,0.78991,0.99069,0.888156,0.887078,16.770514,100,41.99427,39.631775 8,100,92.13425711,50.92106,59.58193,51.978869,51.971372,0.999196,0.999968,0.999742,0.999742,95.702354,100,99.929692,99.928653 16,100,56.66397808,48.195595,58.431446,49.712168,49.691728,0.998283,0.999932,0.9995,0.9995,95.316399,100,99.885137,99.883198 24,100,40.0180961,45.865774,57.75236,48.303092,48.262927,0.997616,0.999877,0.999258,0.999258,94.813489,100,99.830802,99.82755 32,100,29.6332024,44.104353,57.056983,47.152955,47.088158,0.997192,0.999858,0.998987,0.998987,94.300091,100,99.758683,99.75338 40,100,23.3606172,42.741312,56.50171,46.267272,46.178797,0.996974,0.999762,0.998723,0.998723,93.901905,100,99.684641,99.67687 48,100,19.05180627,41.454123,56.115363,45.524272,45.414636,0.996656,0.999707,0.998429,0.998429,93.212479,100,99.595628,99.584266 56,100,16.18793191,40.486197,55.677934,44.944073,44.818532,0.996073,0.999644,0.998156,0.998155,92.684308,100,99.510698,99.495565 64,100,13.63901868,39.514647,55.249823,44.393344,44.251395,0.995158,0.999571,0.99784,0.99784,91.840478,100,99.371119,99.349486 72,100,11.9567563,38.787193,55.042511,43.978065,43.822641,0.994451,0.99952,0.997552,0.997552,91.404602,100,99.261242,99.234304 80,100,10.50762058,38.071209,54.694256,43.518851,43.349026,0.993547,0.999456,0.99721,0.997209,90.560329,100,99.110275,99.075828 88,100,9.476054102,37.491994,54.504892,43.150048,42.96919,0.992784,0.999401,0.996907,0.996907,90.179467,100,98.97017,98.928363 96,100,8.48998081,36.902073,54.28006,42.737889,42.54517,0.991771,0.999332,0.996522,0.996522,89.573602,100,98.79532,98.743951 104,100,7.149586354,36.069249,53.854443,42.077252,41.866599,0.989534,0.999186,0.995813,0.995812,88.038893,100,98.450396,98.378759 112,100,6.224670049,35.385719,53.421685,41.479587,41.25196,0.987589,0.999079,0.99506,0.995059,86.548433,100,98.046063,97.949076 120,100,5.514588214,34.772859,53.000765,40.957597,40.716161,0.985551,0.998975,0.994289,0.994288,85.273099,100,97.601119,97.474497 128,100,4.718884066,34.015067,52.54284,40.268586,40.009915,0.982199,0.998839,0.993052,0.99305,83.10187,100,96.733137,96.546164 136,100,4.177366929,33.403923,52.459907,39.734508,39.460024,0.979844,0.998713,0.991875,0.991872,80.924807,100,95.919984,95.676861 144,100,3.594205137,32.65867,52.084696,39.04157,38.749751,0.975335,0.998495,0.990162,0.990157,76.913054,100,94.593642,94.268264 152,100,3.116118015,31.942904,51.833133,38.411343,38.103945,0.97083,0.998285,0.988243,0.988237,72.743408,100,93.025758,92.611023 160,100,2.706232099,31.050213,51.566863,37.766101,37.444385,0.965177,0.998015,0.985967,0.985956,68.770348,100,91.068737,90.555664 168,100,2.342073904,30.161052,51.046536,37.10594,36.770159,0.958445,0.997673,0.983216,0.983199,63.025381,100,88.676854,88.056732 176,100,2.037712039,29.320349,50.829977,36.446295,36.098671,0.951128,0.997273,0.979883,0.979857,58.872073,100,85.802806,85.07917 184,100,1.761105938,28.495093,50.294472,35.77895,35.421817,0.942362,0.996923,0.975936,0.975898,53.488314,100,82.49814,81.693179 192,100,1.525044768,27.717665,49.727543,35.110721,34.743697,0.932333,0.996218,0.971181,0.971122,48.020932,100,78.499925,77.619698 200,100,1.316947247,26.917065,49.052454,34.445512,34.070505,0.920444,0.995294,0.965568,0.96548,42.499926,100,74.495604,73.530253 208,100,1.13337213,26.114956,48.425896,33.783922,33.398601,0.908476,0.99434,0.958951,0.95882,38.173163,100,70.201003,69.137361 216,100,0.97752694,25.315366,47.829896,33.093283,32.70165,0.896362,0.994127,0.95107,0.950877,33.960278,100,65.76387,64.576364 224,100,0.838986016,24.494639,47.151308,32.33031,31.937034,0.874989,0.993516,0.940989,0.9407,31.060712,100,60.741519,59.397211 232,100,0.704697132,23.635498,46.587834,31.566934,31.167096,0.851899,0.993021,0.928338,0.927908,28.294251,100,55.255473,53.691785 240,100,0.584931682,22.715953,45.101173,30.652696,30.256112,0.824828,0.991914,0.911517,0.910847,25.008545,100,49.062693,47.189603 248,100,0.466199227,21.67321,42.762394,29.62917,29.236327,0.78991,0.99069,0.888156,0.887078,16.770514,100,41.99427,39.631775
are the axis swapped? h264_nvenc needing the lowest bitrate for a high VMAF score?
No. Graph is correct – libaom-av1 is the leader (grey, unfilled, short dash). h264_nvenc is grey, unfilled, LONG dash and is somewhere middle of the pack.
– Gough