Enabling "threaded-me" resulting in a massive performance regression

[Akilan-Sivakumar]

BB Issue #1028
Original issue created by MTTF (bitbucket username :MTTF).
Created on Bitbucket: 2026-03-06
Last updated on Bitbucket: 2026-03-20

I do realize the feature just got implemented, but considering it is supposed to speed up certain operations by "up to 50%" there is clearly a major issue somewhere.

Arch Linux 6.19.6, AMD Zen 5 based 16C/32T processor.

1080, Medium-preset: -55%

Without threaded-me:

./x265 --input FHD.y4m --preset medium --crf 21.0 --output Output.h265 y4m  [info]: 1920x1080 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: Output.h265
x265 [info]: HEVC encoder version 4.1+234-e12088582
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2 AVX512
x265 [info]: Main 10 profile, Level-4 (Main tier)
x265 [info]: Thread pool created using 32 threads
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 5 / wpp(17 rows)
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : hex / 57 / 2 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00 
x265 [info]: Lookahead / bframes / badapt        : 20 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 3 / off / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-21.0 / 0.60
x265 [info]: tools: rd=3 psy-rd=2.00 early-skip rskip mode=1 signhide tmvp
x265 [info]: tools: b-intra strong-intra-smoothing lslices=6 deblock sao
x265 [info]: frame I:      3, Avg QP:21.98  kb/s: 35579.52                      
x265 [info]: frame P:     74, Avg QP:22.20  kb/s: 12535.63
x265 [info]: frame B:    276, Avg QP:28.01  kb/s: 1353.17 
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 3.66s (96.33 fps), 3988.25 kb/s, Avg QP:26.74
x265 [info]: CU: %12.10 time spent in motion estimation, averaging 8.999 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %50.52 %36.06 %43.47 %0.00
x265 [info]: CU: %03.03 time spent in intra analysis, averaging 8.709 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %170.85 %150.50
x265 [info]: CU: %15.54 time spent in inter RDO, measuring 25.750 inter/merge predictions per CTU
x265 [info]: CU: %10.11 time spent in intra RDO, measuring 22.416 intra predictions per CTU
x265 [info]: CU: %06.34 time spent in loop filters, average 0.540 ms per call
x265 [info]: CU: %00.16 time spent in weight analysis, average 1.071 ms per call
x265 [info]: CU: %40.38 time spent in slicetypeDecide (avg 32.547ms) and prelookahead (avg 4.206ms)
x265 [info]: CU: %12.34 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %20.69 %18.55 %60.76
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.52 %07.41 %90.07
x265 [info]: CU: Inter RDO time  per depth %22.60 %32.21 %21.23 %23.97
x265 [info]: CU: Inter RDO calls per depth %03.78 %14.31 %26.13 %55.79
x265 [info]: CU: 180030 64X64 CTUs compressed in 51.110 seconds, 3522.426 CTUs per worker-second
x265 [info]: CU: 13.947 average worker utilization, %43.58 of theoretical maximum utilization

With threaded-me:

./x265 --input FHD.y4m --preset medium --crf 21.0 --threaded-me --output Output.h265
y4m  [info]: 1920x1080 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: Output.h265
x265 [info]: HEVC encoder version 4.1+234-e12088582
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2 AVX512
x265 [info]: Main 10 profile, Level-4 (Main tier)
x265 [info]: Thread pool created using 25 threads
x265 [info]: Thread pool created using 7 threads
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 2 / wpp(17 rows)+threaded-me
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: ThreadedME: task block / buf rows   : 1 / 10
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : hex / 57 / 2 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00 
x265 [info]: Lookahead / bframes / badapt        : 20 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 3 / off / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-21.0 / 0.60
x265 [info]: tools: rd=3 psy-rd=2.00 early-skip rskip mode=1 signhide tmvp
x265 [info]: tools: b-intra strong-intra-smoothing lslices=6 deblock sao
x265 [info]: frame I:      3, Avg QP:21.98  kb/s: 35579.52                      
x265 [info]: frame P:     74, Avg QP:22.21  kb/s: 12669.02
x265 [info]: frame B:    276, Avg QP:27.96  kb/s: 1533.53 
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 8.19s (43.12 fps), 4157.23 kb/s, Avg QP:26.70
x265 [info]: CU: %74.82 time spent in motion estimation, averaging 53.960 CU inter modes per CTU
x265 [info]: CU: %01.01 time spent in intra analysis, averaging 9.208 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %164.98 %147.70
x265 [info]: CU: %05.19 time spent in inter RDO, measuring 25.458 inter/merge predictions per CTU
x265 [info]: CU: %03.46 time spent in intra RDO, measuring 22.915 intra predictions per CTU
x265 [info]: CU: %03.44 time spent in loop filters, average 0.974 ms per call
x265 [info]: CU: %00.04 time spent in weight analysis, average 0.999 ms per call
x265 [info]: CU: %08.18 time spent in slicetypeDecide (avg 95.183ms) and prelookahead (avg 4.056ms)
x265 [info]: CU: %03.85 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %22.18 %18.62 %59.20
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.68 %07.62 %89.70
x265 [info]: CU: Inter RDO time  per depth %23.82 %31.81 %20.83 %23.54
x265 [info]: CU: Inter RDO calls per depth %03.46 %13.37 %25.86 %57.32
x265 [info]: CU: 180030 64X64 CTUs compressed in 170.155 seconds, 1058.038 CTUs per worker-second
x265 [info]: CU: 13325 CTUs blocked by ThreadedME, %07.40 of total CTUs
x265 [info]: CU: 20.784 average worker utilization, %64.95 of theoretical maximum utilization

1080, Slow -preset: -69%

Without threaded-me:

./x265 --input FHD.y4m --preset slow --crf 21.0 --output Output.h265
y4m  [info]: 1920x1080 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: Output.h265
x265 [info]: HEVC encoder version 4.1+234-e12088582
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2 AVX512
x265 [info]: Main 10 profile, Level-4 (Main tier)
x265 [info]: Thread pool created using 32 threads
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 5 / wpp(17 rows)
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : star / 57 / 3 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00 
x265 [info]: Lookahead / bframes / badapt        : 25 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 4 / on / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-21.0 / 0.60
x265 [info]: tools: rect limit-modes rd=4 psy-rd=2.00 rdoq=2 psy-rdoq=1.00
x265 [info]: tools: rskip mode=1 signhide tmvp strong-intra-smoothing lslices=4
x265 [info]: tools: deblock sao
x265 [info]: frame I:      3, Avg QP:21.92  kb/s: 33589.44                      
x265 [info]: frame P:     73, Avg QP:22.07  kb/s: 14977.62
x265 [info]: frame B:    277, Avg QP:27.84  kb/s: 1667.21 
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 9.50s (37.15 fps), 4691.08 kb/s, Avg QP:26.60
x265 [info]: CU: %50.45 time spent in motion estimation, averaging 45.673 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %62.83 %62.97 %63.55 %59.53
x265 [info]: CU: %01.00 time spent in intra analysis, averaging 6.832 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %133.50 %120.75
x265 [info]: CU: %18.12 time spent in inter RDO, measuring 34.844 inter/merge predictions per CTU
x265 [info]: CU: %06.31 time spent in intra RDO, measuring 21.483 intra predictions per CTU
x265 [info]: CU: %00.98 time spent in loop filters, average 0.202 ms per call
x265 [info]: CU: %00.06 time spent in weight analysis, average 1.001 ms per call
x265 [info]: CU: %15.97 time spent in slicetypeDecide (avg 33.792ms) and prelookahead (avg 4.514ms)
x265 [info]: CU: %07.11 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %20.83 %20.44 %58.72
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.06 %06.98 %90.95
x265 [info]: CU: Inter RDO time  per depth %26.56 %31.18 %20.91 %21.35
x265 [info]: CU: Inter RDO calls per depth %05.15 %15.42 %25.79 %53.63
x265 [info]: CU: 180030 64X64 CTUs compressed in 124.500 seconds, 1446.026 CTUs per worker-second
x265 [info]: CU: 13.103 average worker utilization, %40.95 of theoretical maximum utilization

With threaded-me:

 ./x265 --input FHD.y4m --preset slow --crf 21.0 --threaded-me --output Output.h265
y4m  [info]: 1920x1080 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: Output.h265
x265 [info]: HEVC encoder version 4.1+234-e12088582
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2 AVX512
x265 [info]: Main 10 profile, Level-4 (Main tier)
x265 [info]: Thread pool created using 25 threads
x265 [info]: Thread pool created using 7 threads
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 2 / wpp(17 rows)+threaded-me
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: ThreadedME: task block / buf rows   : 1 / 10
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : star / 57 / 3 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00 
x265 [info]: Lookahead / bframes / badapt        : 25 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 4 / on / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-21.0 / 0.60
x265 [info]: tools: rect limit-modes rd=4 psy-rd=2.00 rdoq=2 psy-rdoq=1.00
x265 [info]: tools: rskip mode=1 signhide tmvp strong-intra-smoothing lslices=4
x265 [info]: tools: deblock sao
x265 [info]: frame I:      3, Avg QP:21.92  kb/s: 33589.44                      
x265 [info]: frame P:     73, Avg QP:22.07  kb/s: 15329.04
x265 [info]: frame B:    277, Avg QP:27.67  kb/s: 1793.78 
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 30.67s (11.51 fps), 4863.07 kb/s, Avg QP:26.46
x265 [info]: CU: %91.19 time spent in motion estimation, averaging 173.199 CU inter modes per CTU
x265 [info]: CU: %00.19 time spent in intra analysis, averaging 6.888 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %124.28 %114.08
x265 [info]: CU: %03.51 time spent in inter RDO, measuring 31.776 inter/merge predictions per CTU
x265 [info]: CU: %01.31 time spent in intra RDO, measuring 21.539 intra predictions per CTU
x265 [info]: CU: %00.52 time spent in loop filters, average 0.726 ms per call
x265 [info]: CU: %00.01 time spent in weight analysis, average 1.410 ms per call
x265 [info]: CU: %02.06 time spent in slicetypeDecide (avg 86.118ms) and prelookahead (avg 4.259ms)
x265 [info]: CU: %01.19 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %21.97 %20.45 %57.58
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.13 %07.04 %90.83
x265 [info]: CU: Inter RDO time  per depth %24.68 %31.36 %20.96 %23.00
x265 [info]: CU: Inter RDO calls per depth %04.12 %14.08 %25.84 %55.96
x265 [info]: CU: 180030 64X64 CTUs compressed in 835.014 seconds, 215.601 CTUs per worker-second
x265 [info]: CU: 38752 CTUs blocked by ThreadedME, %21.53 of total CTUs
x265 [info]: CU: 27.225 average worker utilization, %85.08 of theoretical maximum utilization

[Akilan-Sivakumar]

Original comment by divya_m (Bitbucket username :divya_m)
Posted on Bitbucket: 2026-03-11

we tried reproducing the issue in 32 thread machine in power save mode and we could observe performance improvement in powersave mode of the machine but the same thread is giving degradation in performance mode of the machine which we are currently debugging.

Note: 32 thread is the minimum thread count of machine where we see improvement with this feature.

[Akilan-Sivakumar]

Original comment by divya_m (Bitbucket username :divya_m)
Posted on Bitbucket: 2026-03-11

One doubt: Can you please clarify whether you are running in performance mode of the machine or not? And also its good to know about exact processor details

[Akilan-Sivakumar]

Original comment by MTTF (Bitbucket username :MTTF)
Posted on Bitbucket: 2026-03-11

The exact CPU model is AMD Ryzen 9 9950X3D.

The selected frequency scaling setting (power-saver, balanced or performance) made no difference whatsoever to the performance, as is to be expected.

Tested the scaling by limiting core utilization through pools (on a05eebb):

./x265 --input FHD.y4m --preset medium --crf 21.0 --output-depth 10 --output Output.h265

Pools=2, without threaded-me

encoded 353 frames in 18.40s (19.19 fps), 3943.94 kb/s, Avg QP:26.77
x265 [info]: CU: %15.16 time spent in motion estimation, averaging 8.984 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %50.76 %36.11 %43.52 %0.00
x265 [info]: CU: %04.02 time spent in intra analysis, averaging 8.688 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %171.17 %150.88
x265 [info]: CU: %19.97 time spent in inter RDO, measuring 25.688 inter/merge predictions per CTU
x265 [info]: CU: %12.99 time spent in intra RDO, measuring 22.371 intra predictions per CTU
x265 [info]: CU: %04.75 time spent in loop filters, average 0.248 ms per call
x265 [info]: CU: %00.24 time spent in weight analysis, average 1.011 ms per call
x265 [info]: CU: %27.61 time spent in slicetypeDecide (avg 94.079ms) and prelookahead (avg 2.283ms)
x265 [info]: CU: %15.25 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %20.67 %18.44 %60.89
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.53 %07.42 %90.06
x265 [info]: CU: Inter RDO time  per depth %22.39 %31.82 %21.44 %24.35
x265 [info]: CU: Inter RDO calls per depth %03.79 %14.28 %26.12 %55.81
x265 [info]: CU: 180030 64X64 CTUs compressed in 31.343 seconds, 5743.957 CTUs per worker-second
x265 [info]: CU: 1.703 average worker utilization, %85.17 of theoretical maximum utilization

Pools=2, with threaded-me (-77.6%)

encoded 353 frames in 82.15s (4.30 fps), 4107.94 kb/s, Avg QP:26.74
x265 [info]: CU: %73.50 time spent in motion estimation, averaging 98.823 CU inter modes per CTU
x265 [info]: CU: %01.40 time spent in intra analysis, averaging 9.206 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %166.00 %147.98
x265 [info]: CU: %06.26 time spent in inter RDO, measuring 25.443 inter/merge predictions per CTU
x265 [info]: CU: %04.52 time spent in intra RDO, measuring 22.889 intra predictions per CTU
x265 [info]: CU: %00.66 time spent in loop filters, average 0.108 ms per call
x265 [info]: CU: %00.07 time spent in weight analysis, average 0.966 ms per call
x265 [info]: CU: %08.83 time spent in slicetypeDecide (avg 99.764ms) and prelookahead (avg 2.322ms)
x265 [info]: CU: %04.74 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %21.45 %18.66 %59.90
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.67 %07.62 %89.72
x265 [info]: CU: Inter RDO time  per depth %21.70 %31.20 %21.77 %25.33
x265 [info]: CU: Inter RDO calls per depth %03.45 %13.32 %25.83 %57.40
x265 [info]: CU: 180030 64X64 CTUs compressed in 97.721 seconds, 1842.291 CTUs per worker-second
x265 [info]: CU: 23257 CTUs blocked by ThreadedME, %12.92 of total CTUs
x265 [info]: CU: 1.190 average worker utilization, %59.48 of theoretical maximum utilization

Pools=4, without threaded-me

encoded 353 frames in 9.62s (36.69 fps), 3987.81 kb/s, Avg QP:26.73
x265 [info]: CU: %14.29 time spent in motion estimation, averaging 9.003 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %50.48 %35.98 %43.44 %0.00
x265 [info]: CU: %03.82 time spent in intra analysis, averaging 8.711 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %170.76 %150.68
x265 [info]: CU: %18.71 time spent in inter RDO, measuring 25.763 inter/merge predictions per CTU
x265 [info]: CU: %12.29 time spent in intra RDO, measuring 22.346 intra predictions per CTU
x265 [info]: CU: %08.97 time spent in loop filters, average 0.509 ms per call
x265 [info]: CU: %00.20 time spent in weight analysis, average 0.890 ms per call
x265 [info]: CU: %27.50 time spent in slicetypeDecide (avg 93.634ms) and prelookahead (avg 2.808ms)
x265 [info]: CU: %14.21 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %20.70 %18.34 %60.96
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.54 %07.43 %90.04
x265 [info]: CU: Inter RDO time  per depth %22.18 %31.93 %21.49 %24.40
x265 [info]: CU: Inter RDO calls per depth %03.78 %14.32 %26.12 %55.78
x265 [info]: CU: 180030 64X64 CTUs compressed in 34.044 seconds, 5288.217 CTUs per worker-second
x265 [info]: CU: 3.539 average worker utilization, %88.47 of theoretical maximum utilization

Pools=4, with threaded-me (-76.5%)

encoded 353 frames in 40.99s (8.61 fps), 4107.94 kb/s, Avg QP:26.74
x265 [info]: CU: %73.04 time spent in motion estimation, averaging 98.823 CU inter modes per CTU
x265 [info]: CU: %01.39 time spent in intra analysis, averaging 9.206 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %166.00 %147.98
x265 [info]: CU: %06.26 time spent in inter RDO, measuring 25.443 inter/merge predictions per CTU
x265 [info]: CU: %04.48 time spent in intra RDO, measuring 22.889 intra predictions per CTU
x265 [info]: CU: %01.37 time spent in loop filters, average 0.226 ms per call
x265 [info]: CU: %00.07 time spent in weight analysis, average 0.918 ms per call
x265 [info]: CU: %08.75 time spent in slicetypeDecide (avg 95.434ms) and prelookahead (avg 2.280ms)
x265 [info]: CU: %04.64 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %21.42 %18.52 %60.06
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.67 %07.62 %89.72
x265 [info]: CU: Inter RDO time  per depth %21.71 %31.20 %21.72 %25.37
x265 [info]: CU: Inter RDO calls per depth %03.45 %13.32 %25.83 %57.40
x265 [info]: CU: 180030 64X64 CTUs compressed in 99.092 seconds, 1816.794 CTUs per worker-second
x265 [info]: CU: 105581 CTUs blocked by ThreadedME, %58.65 of total CTUs
x265 [info]: CU: 2.418 average worker utilization, %60.44 of theoretical maximum utilization

Pools=8, without threaded-me

encoded 353 frames in 5.59s (63.14 fps), 3987.81 kb/s, Avg QP:26.73
x265 [info]: CU: %13.75 time spent in motion estimation, averaging 9.003 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %50.48 %35.98 %43.44 %0.00
x265 [info]: CU: %03.66 time spent in intra analysis, averaging 8.711 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %170.76 %150.68
x265 [info]: CU: %18.11 time spent in inter RDO, measuring 25.763 inter/merge predictions per CTU
x265 [info]: CU: %11.79 time spent in intra RDO, measuring 22.346 intra predictions per CTU
x265 [info]: CU: %09.68 time spent in loop filters, average 0.595 ms per call
x265 [info]: CU: %00.22 time spent in weight analysis, average 1.069 ms per call
x265 [info]: CU: %28.85 time spent in slicetypeDecide (avg 65.631ms) and prelookahead (avg 3.820ms)
x265 [info]: CU: %13.95 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %20.76 %18.33 %60.91
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.54 %07.43 %90.04
x265 [info]: CU: Inter RDO time  per depth %22.35 %31.87 %21.40 %24.39
x265 [info]: CU: Inter RDO calls per depth %03.78 %14.32 %26.12 %55.78
x265 [info]: CU: 180030 64X64 CTUs compressed in 36.909 seconds, 4877.662 CTUs per worker-second
x265 [info]: CU: 6.602 average worker utilization, %82.53 of theoretical maximum utilization

Pools=8, with threaded-me (-72.8%)

encoded 353 frames in 20.56s (17.17 fps), 4107.94 kb/s, Avg QP:26.74
x265 [info]: CU: %71.58 time spent in motion estimation, averaging 98.823 CU inter modes per CTU
x265 [info]: CU: %01.35 time spent in intra analysis, averaging 9.206 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %166.00 %147.98
x265 [info]: CU: %06.15 time spent in inter RDO, measuring 25.443 inter/merge predictions per CTU
x265 [info]: CU: %04.38 time spent in intra RDO, measuring 22.889 intra predictions per CTU
x265 [info]: CU: %03.24 time spent in loop filters, average 0.572 ms per call
x265 [info]: CU: %00.07 time spent in weight analysis, average 0.927 ms per call
x265 [info]: CU: %08.67 time spent in slicetypeDecide (avg 91.504ms) and prelookahead (avg 2.584ms)
x265 [info]: CU: %04.55 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %21.42 %18.49 %60.10
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.67 %07.62 %89.72
x265 [info]: CU: Inter RDO time  per depth %21.88 %31.18 %21.72 %25.21
x265 [info]: CU: Inter RDO calls per depth %03.45 %13.32 %25.83 %57.40
x265 [info]: CU: 180030 64X64 CTUs compressed in 105.955 seconds, 1699.124 CTUs per worker-second
x265 [info]: CU: 75800 CTUs blocked by ThreadedME, %42.10 of total CTUs
x265 [info]: CU: 5.152 average worker utilization, %64.40 of theoretical maximum utilization

Pools=16, without threaded-me

encoded 353 frames in 4.22s (83.66 fps), 3987.81 kb/s, Avg QP:26.73
x265 [info]: CU: %12.48 time spent in motion estimation, averaging 9.003 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %50.48 %35.98 %43.44 %0.00
x265 [info]: CU: %03.34 time spent in intra analysis, averaging 8.711 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %170.76 %150.68
x265 [info]: CU: %16.58 time spent in inter RDO, measuring 25.763 inter/merge predictions per CTU
x265 [info]: CU: %10.75 time spent in intra RDO, measuring 22.346 intra predictions per CTU
x265 [info]: CU: %08.17 time spent in loop filters, average 0.626 ms per call
x265 [info]: CU: %00.18 time spent in weight analysis, average 1.104 ms per call
x265 [info]: CU: %35.57 time spent in slicetypeDecide (avg 39.176ms) and prelookahead (avg 3.940ms)
x265 [info]: CU: %12.93 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %20.62 %18.25 %61.13
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.54 %07.43 %90.04
x265 [info]: CU: Inter RDO time  per depth %22.58 %31.95 %21.27 %24.21
x265 [info]: CU: Inter RDO calls per depth %03.78 %14.32 %26.12 %55.78
x265 [info]: CU: 180030 64X64 CTUs compressed in 45.947 seconds, 3918.228 CTUs per worker-second
x265 [info]: CU: 10.889 average worker utilization, %68.06 of theoretical maximum utilization

Pools=16, with threaded-me (-61.9%)

encoded 353 frames in 11.07s (31.89 fps), 4160.14 kb/s, Avg QP:26.70
x265 [info]: CU: %71.09 time spent in motion estimation, averaging 54.005 CU inter modes per CTU
x265 [info]: CU: %01.31 time spent in intra analysis, averaging 9.240 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %165.24 %147.23
x265 [info]: CU: %06.15 time spent in inter RDO, measuring 25.557 inter/merge predictions per CTU
x265 [info]: CU: %04.27 time spent in intra RDO, measuring 22.876 intra predictions per CTU
x265 [info]: CU: %03.93 time spent in loop filters, average 0.765 ms per call
x265 [info]: CU: %00.08 time spent in weight analysis, average 1.213 ms per call
x265 [info]: CU: %08.66 time spent in slicetypeDecide (avg 105.079ms) and prelookahead (avg 2.724ms)
x265 [info]: CU: %04.51 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %21.67 %18.49 %59.84
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.69 %07.65 %89.65
x265 [info]: CU: Inter RDO time  per depth %22.63 %31.26 %21.34 %24.77
x265 [info]: CU: Inter RDO calls per depth %03.45 %13.35 %25.89 %57.31
x265 [info]: CU: 180030 64X64 CTUs compressed in 116.916 seconds, 1539.818 CTUs per worker-second
x265 [info]: CU: 19766 CTUs blocked by ThreadedME, %10.98 of total CTUs
x265 [info]: CU: 10.561 average worker utilization, %66.01 of theoretical maximum utilization

Pools=32, without threaded-me

encoded 353 frames in 3.63s (97.23 fps), 3987.81 kb/s, Avg QP:26.73
x265 [info]: CU: %12.02 time spent in motion estimation, averaging 9.003 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %50.48 %35.98 %43.44 %0.00
x265 [info]: CU: %03.19 time spent in intra analysis, averaging 8.711 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %170.76 %150.68
x265 [info]: CU: %15.99 time spent in inter RDO, measuring 25.763 inter/merge predictions per CTU
x265 [info]: CU: %10.21 time spent in intra RDO, measuring 22.346 intra predictions per CTU
x265 [info]: CU: %06.55 time spent in loop filters, average 0.546 ms per call
x265 [info]: CU: %00.18 time spent in weight analysis, average 1.172 ms per call
x265 [info]: CU: %39.30 time spent in slicetypeDecide (avg 31.057ms) and prelookahead (avg 4.292ms)
x265 [info]: CU: %12.56 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %20.59 %18.37 %61.04
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.54 %07.43 %90.04
x265 [info]: CU: Inter RDO time  per depth %22.44 %31.84 %21.34 %24.38
x265 [info]: CU: Inter RDO calls per depth %03.78 %14.32 %26.12 %55.78
x265 [info]: CU: 180030 64X64 CTUs compressed in 49.987 seconds, 3601.540 CTUs per worker-second
x265 [info]: CU: 13.768 average worker utilization, %43.02 of theoretical maximum utilization

Pools=32, with threaded-me (-55.7%)

encoded 353 frames in 8.19s (43.12 fps), 4156.02 kb/s, Avg QP:26.70
x265 [info]: CU: %74.08 time spent in motion estimation, averaging 53.968 CU inter modes per CTU
x265 [info]: CU: %01.09 time spent in intra analysis, averaging 9.222 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %164.91 %146.22
x265 [info]: CU: %05.40 time spent in inter RDO, measuring 25.451 inter/merge predictions per CTU
x265 [info]: CU: %03.57 time spent in intra RDO, measuring 22.857 intra predictions per CTU
x265 [info]: CU: %03.57 time spent in loop filters, average 0.957 ms per call
x265 [info]: CU: %00.05 time spent in weight analysis, average 1.125 ms per call
x265 [info]: CU: %08.29 time spent in slicetypeDecide (avg 93.884ms) and prelookahead (avg 4.091ms)
x265 [info]: CU: %03.95 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %21.80 %18.70 %59.50
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.69 %07.66 %89.65
x265 [info]: CU: Inter RDO time  per depth %23.42 %31.69 %20.84 %24.05
x265 [info]: CU: Inter RDO calls per depth %03.47 %13.37 %25.87 %57.29
x265 [info]: CU: 180030 64X64 CTUs compressed in 160.892 seconds, 1118.947 CTUs per worker-second
x265 [info]: CU: 10797 CTUs blocked by ThreadedME, %06.00 of total CTUs
x265 [info]: CU: 19.652 average worker utilization, %61.41 of theoretical maximum utilization

[Akilan-Sivakumar]

Original comment by divya_m (Bitbucket username :divya_m)
Posted on Bitbucket: 2026-03-12

I would recommend you to use this feature from 32 threads onwards. There will be degradation when we go below that limit.

Can you also let us know what all the clips you used for testing to reproduce this 32 thread degradation case?

[Akilan-Sivakumar]

Original comment by MTTF (Bitbucket username :MTTF)
Posted on Bitbucket: 2026-03-12

How many cores / threads did the test system on which the "giving speedups of upto 1.5x" statement was achieved on have?

[Akilan-Sivakumar]

Original comment by divya_m (Bitbucket username :divya_m)
Posted on Bitbucket: 2026-03-13

We tested on Intel Xeon-SP dual-socket Platinum 8180 with 112 logical Processor.

[Akilan-Sivakumar]

Original comment by MTTF (Bitbucket username :MTTF)
Posted on Bitbucket: 2026-03-18

Sadly, increasing the number of available resources does not solve the issue.

Tested using AMD Ryzen Threadripper 3990X (64C/128T).
The processors were split into two NUMA pools, since for whatever reason X265 is limited to 64 processors per pool.

1080p, without threaded-me:

./x265 --preset slow --crf 17.0 --input InputFHD.y4m --output-depth 10 --output OutputFHD.h265
y4m  [info]: 1920x1080 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: OutputFHD.h265
x265 [info]: HEVC encoder version 4.1+239-8be7dbf81
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
x265 [info]: Main 10 profile, Level-4 (Main tier)
x265 [info]: Thread pool 0 using 64 threads on numa nodes 0
x265 [info]: Thread pool 1 using 64 threads on numa nodes 1
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 5 / wpp(17 rows)
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : star / 57 / 3 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00
x265 [info]: Lookahead / bframes / badapt        : 25 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 4 / on / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-17.0 / 0.60
x265 [info]: tools: rect limit-modes rd=4 psy-rd=2.00 rdoq=2 psy-rdoq=1.00
x265 [info]: tools: rskip mode=1 signhide tmvp strong-intra-smoothing lslices=4
x265 [info]: tools: deblock sao
x265 [info]: frame I:      3, Avg QP:17.92  kb/s: 56230.96
x265 [info]: frame P:     73, Avg QP:18.07  kb/s: 37649.96
x265 [info]: frame B:    277, Avg QP:23.07  kb/s: 4877.46
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 22.62s (15.61 fps), 12091.21 kb/s, Avg QP:21.99
x265 [info]: CU: %48.34 time spent in motion estimation, averaging 85.191 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %66.65 %63.56 %62.94 %59.58
x265 [info]: CU: %00.74 time spent in intra analysis, averaging 10.794 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %458.38 %125.58
x265 [info]: CU: %24.36 time spent in inter RDO, measuring 67.900 inter/merge predictions per CTU
x265 [info]: CU: %06.25 time spent in intra RDO, measuring 26.458 intra predictions per CTU
x265 [info]: CU: %00.28 time spent in loop filters, average 0.179 ms per call
x265 [info]: CU: %00.03 time spent in weight analysis, average 1.711 ms per call
x265 [info]: CU: %13.42 time spent in slicetypeDecide (avg 59.176ms) and prelookahead (avg 6.624ms)
x265 [info]: CU: %06.57 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %11.80 %23.86 %64.34
x265 [info]: CU: Intra RDO calls per depth %00.00 %01.25 %08.06 %90.69
x265 [info]: CU: Inter RDO time  per depth %20.61 %34.47 %24.73 %20.19
x265 [info]: CU: Inter RDO calls per depth %02.34 %13.72 %28.98 %54.96
x265 [info]: CU: 180030 64X64 CTUs compressed in 379.435 seconds, 474.468 CTUs per worker-second
x265 [info]: CU: 16.776 average worker utilization, %13.11 of theoretical maximum utilization

1080p, with threaded-me (-24.3%)

./x265 --preset slow --crf 17.0 --input InputFHD.y4m --output-depth 10 --threaded-me --output OutputFHD.h265
y4m  [info]: 1920x1080 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: OutputFHD.h265
x265 [info]: HEVC encoder version 4.1+239-8be7dbf81
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
x265 [info]: Main 10 profile, Level-4 (Main tier)
x265 [info]: Thread pool 0 using 108 threads on numa nodes 0,1
x265 [info]: Thread pool 1 using 20 threads on numa nodes 1
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 4 / wpp(17 rows)+threaded-me
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: ThreadedME: task block / buf rows   : 1 / 5
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : star / 57 / 3 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00
x265 [info]: Lookahead / bframes / badapt        : 25 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 4 / on / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-17.0 / 0.60
x265 [info]: tools: rect limit-modes rd=4 psy-rd=2.00 rdoq=2 psy-rdoq=1.00
x265 [info]: tools: rskip mode=1 signhide tmvp strong-intra-smoothing lslices=4
x265 [info]: tools: deblock sao
x265 [info]: frame I:      3, Avg QP:17.92  kb/s: 56230.96
x265 [info]: frame P:     73, Avg QP:18.07  kb/s: 38181.08
x265 [info]: frame B:    277, Avg QP:23.07  kb/s: 5255.63
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 29.89s (11.81 fps), 12497.79 kb/s, Avg QP:21.99
x265 [info]: CU: %87.41 time spent in motion estimation, averaging 147.014 CU inter modes per CTU
x265 [info]: CU: %00.21 time spent in intra analysis, averaging 10.886 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %438.78 %118.32
x265 [info]: CU: %06.70 time spent in inter RDO, measuring 59.797 inter/merge predictions per CTU
x265 [info]: CU: %01.77 time spent in intra RDO, measuring 26.550 intra predictions per CTU
x265 [info]: CU: %00.18 time spent in loop filters, average 0.525 ms per call
x265 [info]: CU: %00.01 time spent in weight analysis, average 2.748 ms per call
x265 [info]: CU: %02.00 time spent in slicetypeDecide (avg 77.342ms) and prelookahead (avg 7.047ms)
x265 [info]: CU: %01.71 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %12.05 %24.78 %63.17
x265 [info]: CU: Intra RDO calls per depth %00.00 %01.26 %08.19 %90.56
x265 [info]: CU: Inter RDO time  per depth %19.52 %34.17 %24.65 %21.66
x265 [info]: CU: Inter RDO calls per depth %01.92 %11.79 %27.45 %58.85
x265 [info]: CU: 180030 64X64 CTUs compressed in 1750.221 seconds, 102.861 CTUs per worker-second
x265 [info]: CU: 11860 CTUs blocked by ThreadedME, %06.59 of total CTUs
x265 [info]: CU: 58.562 average worker utilization, %45.75 of theoretical maximum utilization

2160p, without threaded-me

./x265 --preset slow --crf 17.0 --input InputUHD.y4m --output-depth 10 --output OutputUHD.h265
y4m  [info]: 3840x2160 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: OutputUHD.h265
x265 [info]: HEVC encoder version 4.1+239-8be7dbf81
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
x265 [info]: Main 10 profile, Level-5 (Main tier)
x265 [info]: Thread pool 0 using 64 threads on numa nodes 0
x265 [info]: Thread pool 1 using 64 threads on numa nodes 1
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 6 / wpp(34 rows)
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : star / 57 / 3 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00
x265 [info]: Lookahead / bframes / badapt        : 25 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 4 / on / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-17.0 / 0.60
x265 [info]: tools: rect limit-modes rd=4 psy-rd=2.00 rdoq=2 psy-rdoq=1.00
x265 [info]: tools: rskip mode=1 signhide tmvp strong-intra-smoothing lslices=4
x265 [info]: tools: deblock sao
x265 [info]: frame I:      3, Avg QP:18.40  kb/s: 178124.72
x265 [info]: frame P:     71, Avg QP:19.04  kb/s: 183117.52
x265 [info]: frame B:    279, Avg QP:22.70  kb/s: 73789.75
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 71.68s (4.92 fps), 96665.90 kb/s, Avg QP:21.92
x265 [info]: CU: %52.23 time spent in motion estimation, averaging 175.801 CU inter modes per CTU
x265 [info]: CU: Skipped motion searches per depth %67.18 %66.32 %64.09 %59.70
x265 [info]: CU: %00.72 time spent in intra analysis, averaging 16.507 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %81.95 %85.38
x265 [info]: CU: %28.32 time spent in inter RDO, measuring 143.946 inter/merge predictions per CTU
x265 [info]: CU: %05.03 time spent in intra RDO, measuring 32.436 intra predictions per CTU
x265 [info]: CU: %00.10 time spent in loop filters, average 0.286 ms per call
x265 [info]: CU: %00.02 time spent in weight analysis, average 9.639 ms per call
x265 [info]: CU: %07.05 time spent in slicetypeDecide (avg 283.700ms) and prelookahead (avg 30.133ms)
x265 [info]: CU: %06.54 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %17.24 %21.45 %61.31
x265 [info]: CU: Intra RDO calls per depth %00.00 %01.87 %08.05 %90.09
x265 [info]: CU: Inter RDO time  per depth %18.60 %28.05 %27.56 %25.79
x265 [info]: CU: Inter RDO calls per depth %01.21 %07.30 %25.42 %66.07
x265 [info]: CU: 720120 64X64 CTUs compressed in 3360.211 seconds, 214.308 CTUs per worker-second
x265 [info]: CU: 46.876 average worker utilization, %36.62 of theoretical maximum utilization

2160p, with threaded-me (-44.1%)

./x265 --preset slow --crf 17.0 --input InputUHD.y4m --output-depth 10 --threaded-me --output OutputUHD.h265
y4m  [info]: 3840x2160 fps 30/1 i420p8 frames 0 - 352 of 353
raw  [info]: output file: OutputUHD.h265
x265 [info]: HEVC encoder version 4.1+239-8be7dbf81
x265 [info]: build info [Linux][GCC 15.2.1][64 bit] 10bit
x265 [info]: using cpu capabilities: MMX2 SSE2Fast LZCNT SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
x265 [info]: Main 10 profile, Level-5 (Main tier)
x265 [info]: Thread pool 0 using 96 threads on numa nodes 0,1
x265 [info]: Thread pool 1 using 32 threads on numa nodes 1
x265 [info]: Slices                              : 1
x265 [info]: frame threads / pool features       : 6 / wpp(34 rows)+threaded-me
x265 [info]: Coding QT: max CU size, min CU size : 64 / 8
x265 [info]: ThreadedME: task block / buf rows   : 1 / 3
x265 [info]: Residual QT: max TU size, max depth : 32 / 1 inter / 1 intra
x265 [info]: ME / range / subpel / merge         : star / 57 / 3 / 3
x265 [info]: Keyframe min / max / scenecut / bias  : 25 / 250 / 40 / 5.00
x265 [info]: Lookahead / bframes / badapt        : 25 / 4 / 2
x265 [info]: b-pyramid / weightp / weightb       : 1 / 1 / 0
x265 [info]: References / ref-limit  cu / depth  : 4 / on / on
x265 [info]: AQ: mode / str / qg-size / cu-tree  : 2 / 1.0 / 32 / 1
x265 [info]: Rate Control / qCompress            : CRF-17.0 / 0.60
x265 [info]: tools: rect limit-modes rd=4 psy-rd=2.00 rdoq=2 psy-rdoq=1.00
x265 [info]: tools: rskip mode=1 signhide tmvp strong-intra-smoothing lslices=4
x265 [info]: tools: deblock sao
x265 [info]: frame I:      3, Avg QP:18.40  kb/s: 178124.72
x265 [info]: frame P:     71, Avg QP:19.03  kb/s: 183536.08
x265 [info]: frame B:    279, Avg QP:22.70  kb/s: 74937.85
x265 [info]: Weighted P-Frames: Y:0.0% UV:0.0%

encoded 353 frames in 128.48s (2.75 fps), 97657.50 kb/s, Avg QP:21.92
x265 [info]: CU: %82.69 time spent in motion estimation, averaging 203.824 CU inter modes per CTU
x265 [info]: CU: %00.32 time spent in intra analysis, averaging 16.841 Intra PUs per CTU
x265 [info]: CU: Skipped intra CUs at depth %-nan %62.29 %61.59
x265 [info]: CU: %10.41 time spent in inter RDO, measuring 119.522 inter/merge predictions per CTU
x265 [info]: CU: %02.30 time spent in intra RDO, measuring 32.770 intra predictions per CTU
x265 [info]: CU: %00.08 time spent in loop filters, average 0.673 ms per call
x265 [info]: CU: %00.01 time spent in weight analysis, average 16.089 ms per call
x265 [info]: CU: %01.89 time spent in slicetypeDecide (avg 455.429ms) and prelookahead (avg 36.953ms)
x265 [info]: CU: %02.30 time spent in other tasks
x265 [info]: CU: Intra RDO time  per depth %00.00 %18.23 %23.06 %58.71
x265 [info]: CU: Intra RDO calls per depth %00.00 %02.00 %08.72 %89.28
x265 [info]: CU: Inter RDO time  per depth %18.15 %26.77 %26.70 %28.38
x265 [info]: CU: Inter RDO calls per depth %01.09 %06.40 %23.40 %69.12
x265 [info]: CU: 720120 64X64 CTUs compressed in 10177.518 seconds, 70.756 CTUs per worker-second
x265 [info]: CU: 34973 CTUs blocked by ThreadedME, %04.86 of total CTUs
x265 [info]: CU: 79.214 average worker utilization, %61.89 of theoretical maximum utilization

[Akilan-Sivakumar]

Original comment by gunasrij (Bitbucket username :gunasrij)
Posted on Bitbucket: 2026-03-20

@{5e9db3c685c25b0c31366259} Thanks for the information and test results.
We are observing that the Threaded-ME feature can exhibit performance degradation on higher-frequency systems compared to lower-frequency setups.
We are investigating the underlying cause and will get back once we have more insights.