Perfomance degradation of H265 encoder (omxh265enc) in version 32.3.1 relatively to 28.2.1

Hi. I have updated one of my jetsons to 32.3.1 version and moved there my software with rtsp server, which sends full hd 60 fps stream.

I’m using this pipeline:

( appsrc name=vsrc ! videoconvert ! video/x-raw,format=BGRx,width=1920,height=1080,framerate=60/1 
! nvvidconv ! video/x-raw(memory:NVMM),format=NV12 
! omxh265enc  MeasureEncoderLatency=true bitrate=20000000 control-rate=2 ! rtph265pay name=pay0 pt=96 )

It worked perfectly on 28.2.1 version, but now it chokes and can’t provide real time perfomance.

I compared latencies, here is part of them from 32.3.1:

KPI: omx: frameNumber= 38 encoder= 13 ms pts= 633334200

KPI: omx: frameNumber= 39 encoder= 15 ms pts= 650000800

KPI: omx: frameNumber= 40 encoder= 13 ms pts= 666667500

KPI: omx: frameNumber= 41 encoder= 17 ms pts= 683334200

KPI: omx: frameNumber= 42 encoder= 13 ms pts= 700000900

KPI: omx: frameNumber= 43 encoder= 13 ms pts= 716667600

KPI: omx: frameNumber= 44 encoder= 13 ms pts= 733334300

KPI: omx: frameNumber= 45 encoder= 15 ms pts= 750001000

KPI: omx: frameNumber= 46 encoder= 19 ms pts= 766667700

KPI: omx: frameNumber= 47 encoder= 14 ms pts= 783334400

KPI: omx: frameNumber= 48 encoder= 14 ms pts= 800001000

KPI: omx: frameNumber= 49 encoder= 16 ms pts= 816667700

KPI: omx: frameNumber= 50 encoder= 14 ms pts= 833334400

KPI: omx: frameNumber= 51 encoder= 18 ms pts= 850001100

KPI: omx: frameNumber= 52 encoder= 13 ms pts= 866667800

This from an old version:

KPI: omx: frameNumber= 9 encoder= 3 ms pts= 150000200

KPI: omx: frameNumber= 10 encoder= 4 ms pts= 166666900

KPI: omx: frameNumber= 11 encoder= 5 ms pts= 183333500

KPI: omx: frameNumber= 12 encoder= 4 ms pts= 200000200

KPI: omx: frameNumber= 13 encoder= 4 ms pts= 216666900

KPI: omx: frameNumber= 14 encoder= 3 ms pts= 233333600

KPI: omx: frameNumber= 15 encoder= 4 ms pts= 250000300

KPI: omx: frameNumber= 16 encoder= 11 ms pts= 266667000

KPI: omx: frameNumber= 17 encoder= 5 ms pts= 283333700

KPI: omx: frameNumber= 18 encoder= 4 ms pts= 300000400

KPI: omx: frameNumber= 19 encoder= 4 ms pts= 316667100

KPI: omx: frameNumber= 20 encoder= 4 ms pts= 333333700

KPI: omx: frameNumber= 21 encoder= 4 ms pts= 350000400

KPI: omx: frameNumber= 22 encoder= 14 ms pts= 366667100

KPI: omx: frameNumber= 23 encoder= 4 ms pts= 383333800

KPI: omx: frameNumber= 24 encoder= 4 ms pts= 400000500

I also find out that there is a new encoder nvv4l2h265enc and tried to use it with this pipeline:

( appsrc name=vsrc ! videoconvert ! video/x-raw,format=BGRx, width=1920,height=1080,framerate=60/1 
! nvvidconv ! video/x-raw(memory:NVMM),format=NV12 
! nvv4l2h265enc MeasureEncoderLatency=true control-rate=1 bitrate=20000000 maxperf-enable=true 
! rtph265pay name=pay0 pt=96 )

According to latency measures it has much better perfomance, but it still can’t provide real time perfomance and chokes.(i can’t understand why)

KPI: v4l2: frameNumber= 7 encoder= 4 ms pts= 116666800

KPI: v4l2: frameNumber= 8 encoder= 4 ms pts= 133333500

KPI: v4l2: frameNumber= 9 encoder= 3 ms pts= 150000200

KPI: v4l2: frameNumber= 10 encoder= 3 ms pts= 166666900

KPI: v4l2: frameNumber= 11 encoder= 4 ms pts= 183333600

KPI: v4l2: frameNumber= 12 encoder= 3 ms pts= 200000300

KPI: v4l2: frameNumber= 13 encoder= 4 ms pts= 216667000

KPI: v4l2: frameNumber= 14 encoder= 4 ms pts= 233333600

KPI: v4l2: frameNumber= 15 encoder= 3 ms pts= 250000300

KPI: v4l2: frameNumber= 16 encoder= 5 ms pts= 266667000

KPI: v4l2: frameNumber= 17 encoder= 4 ms pts= 283333700

KPI: v4l2: frameNumber= 18 encoder= 4 ms pts= 300000400

KPI: v4l2: frameNumber= 19 encoder= 3 ms pts= 316667100

KPI: v4l2: frameNumber= 20 encoder= 3 ms pts= 333333800

KPI: v4l2: frameNumber= 21 encoder= 4 ms pts= 350000500

So, what i’m doing wrong and how i can get real time perfomance on 32.3.1 version?

my jetson_clocks --show and tegrastats

SOC family:tegra186  Machine:quill
Online CPUs: 0-5
CPU Cluster Switching: Disabled
cpu0: Online=1 Governor=schedutil MinFreq=2035200 MaxFreq=2035200 CurrentFreq=2035200 IdleStates: C1=0 c7=0 
cpu1: Online=1 Governor=schedutil MinFreq=2035200 MaxFreq=2035200 CurrentFreq=2035200 IdleStates: C1=0 c6=0 c7=0 
cpu2: Online=1 Governor=schedutil MinFreq=2035200 MaxFreq=2035200 CurrentFreq=2035200 IdleStates: C1=0 c6=0 c7=0 
cpu3: Online=1 Governor=schedutil MinFreq=2035200 MaxFreq=2035200 CurrentFreq=2035200 IdleStates: C1=0 c7=0 
cpu4: Online=1 Governor=schedutil MinFreq=2035200 MaxFreq=2035200 CurrentFreq=2035200 IdleStates: C1=0 c7=0 
cpu5: Online=1 Governor=schedutil MinFreq=2035200 MaxFreq=2035200 CurrentFreq=2035200 IdleStates: C1=0 c7=0 
GPU MinFreq=1300500000 MaxFreq=1300500000 CurrentFreq=1300500000
EMC MinFreq=40800000 MaxFreq=1866000000 CurrentFreq=1866000000 FreqOverride=1
Fan: speed=255
NV Power Mode: MAXN
RAM 3874/7859MB (lfb 303x4MB) SWAP 68/3929MB (cached 3MB) CPU [2%@2035,8%@2035,100%@2035,2%@2035,1%@2035,1%@2035] EMC_FREQ 0% GR3D_FREQ 0% PLL@32C MCPU@32C PMIC@100C Tboard@28C GPU@30C BCPU@32C thermal@31.2C Tdiode@28.5C VDD_SYS_GPU 229/229 VDD_SYS_SOC 839/839 VDD_4V0_WIFI 0/0 VDD_IN 4084/4084 VDD_SYS_CPU 992/992 VDD_SYS_DDR 1313/1313
RAM 3875/7859MB (lfb 303x4MB) SWAP 68/3929MB (cached 3MB) CPU [2%@2035,31%@2035,100%@2035,4%@2035,3%@2035,2%@2035] EMC_FREQ 0% GR3D_FREQ 0% PLL@32C MCPU@32C PMIC@100C Tboard@28C GPU@30C BCPU@32C thermal@31.2C Tdiode@28.25C VDD_SYS_GPU 229/229 VDD_SYS_SOC 839/839 VDD_4V0_WIFI 0/0 VDD_IN 4275/4179 VDD_SYS_CPU 1144/1068 VDD_SYS_DDR 1332/1322
RAM 3875/7859MB (lfb 303x4MB) SWAP 68/3929MB (cached 3MB) CPU [13%@2035,22%@2035,100%@2035,9%@2035,17%@2035,12%@2035] EMC_FREQ 0% GR3D_FREQ 0% PLL@32C MCPU@32C PMIC@100C Tboard@28C GPU@29.5C BCPU@32C thermal@31.2C Tdiode@28.5C VDD_SYS_GPU 229/229 VDD_SYS_SOC 839/839 VDD_4V0_WIFI 0/0 VDD_IN 4428/4262 VDD_SYS_CPU 1297/1144 VDD_SYS_DDR 1351/1332
RAM 3875/7859MB (lfb 303x4MB) SWAP 68/3929MB (cached 3MB) CPU [4%@2035,15%@2035,100%@2035,2%@2035,2%@2035,3%@2035] EMC_FREQ 0% GR3D_FREQ 0% PLL@32C MCPU@32C PMIC@100C Tboard@28C GPU@29.5C BCPU@32C thermal@31C Tdiode@28.5C VDD_SYS_GPU 229/229 VDD_SYS_SOC 839/839 VDD_4V0_WIFI 0/0 VDD_IN 4199/4246 VDD_SYS_CPU 1068/1125 VDD_SYS_DDR 1332/1332

Hi,
Please run sudo nvpmodel -m 0 and sudo jetson_clocks
And also sudo tegrastats to get NVENC clocks
We are deprecating omx plugins. Please run with nvv4l2h265enc.

Hi. Thanks for your answer. I have done everything you said, but still have poor perfomance. So my pipeline is:

appsrc name=vsrc ! videoconvert ! video/x-raw,format=BGRx, width=1920,height=1080,framerate=60/1 ! nvvidconv ! video/x-raw(memory:NVMM),format=NV12 ! nvv4l2h265enc MeasureEncoderLatency=true control-rate=1 bitrate=20000000 maxperf-enable=true ! rtph265pay name=pay0 pt=96
Part of measure latency:

KPI: v4l2: frameNumber= 0 encoder= 49 ms pts= 0

KPI: v4l2: frameNumber= 1 encoder= 204 ms pts= 16666600

KPI: v4l2: frameNumber= 2 encoder= 27 ms pts= 33333300

KPI: v4l2: frameNumber= 3 encoder= 6 ms pts= 50000000

KPI: v4l2: frameNumber= 4 encoder= 4 ms pts= 66666700

KPI: v4l2: frameNumber= 5 encoder= 5 ms pts= 83333400

KPI: v4l2: frameNumber= 6 encoder= 3 ms pts= 100000100

KPI: v4l2: frameNumber= 7 encoder= 4 ms pts= 116666800

KPI: v4l2: frameNumber= 8 encoder= 3 ms pts= 133333500

KPI: v4l2: frameNumber= 9 encoder= 3 ms pts= 150000200

KPI: v4l2: frameNumber= 10 encoder= 4 ms pts= 166666800

KPI: v4l2: frameNumber= 11 encoder= 3 ms pts= 183333500

KPI: v4l2: frameNumber= 12 encoder= 3 ms pts= 200000200

KPI: v4l2: frameNumber= 13 encoder= 3 ms pts= 216666900

KPI: v4l2: frameNumber= 14 encoder= 3 ms pts= 233333600

KPI: v4l2: frameNumber= 15 encoder= 3 ms pts= 250000300

KPI: v4l2: frameNumber= 16 encoder= 3 ms pts= 266667000

KPI: v4l2: frameNumber= 17 encoder= 3 ms pts= 283333700

KPI: v4l2: frameNumber= 18 encoder= 4 ms pts= 300000400

KPI: v4l2: frameNumber= 19 encoder= 3 ms pts= 316667100

KPI: v4l2: frameNumber= 20 encoder= 3 ms pts= 333333700

And part of tegrastats:

RAM 4949/7859MB (lfb 191x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2035,100%@2036,100%@2035,100%@2035,100%@2035,100%@2034] EMC_FREQ 7%@1866 GR3D_FREQ 6%@1300 NVENC 1164 APE 150 MTS fg 4% bg 0% PLL@38.5C MCPU@38.5C PMIC@100C Tboard@32C GPU@35C BCPU@38.5C thermal@37.2C Tdiode@34C VDD_SYS_GPU 456/347 VDD_SYS_SOC 1599/978 VDD_4V0_WIFI 0/0 VDD_IN 9520/8750 VDD_SYS_CPU 4870/5018 VDD_SYS_DDR 1694/1498
RAM 4928/7859MB (lfb 191x4MB) SWAP 0/3929MB (cached 0MB) CPU [99%@2010,100%@2035,100%@2035,100%@2012,100%@2014,100%@2015] EMC_FREQ 10%@1866 GR3D_FREQ 4%@1300 NVENC 1164 APE 150 MTS fg 4% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35C BCPU@39C thermal@37.4C Tdiode@34.25C VDD_SYS_GPU 532/359 VDD_SYS_SOC 1750/1030 VDD_4V0_WIFI 0/0 VDD_IN 9862/8824 VDD_SYS_CPU 4794/5003 VDD_SYS_DDR 1827/1520
RAM 4927/7859MB (lfb 191x4MB) SWAP 0/3929MB (cached 0MB) CPU [98%@2034,100%@2035,100%@2035,100%@2035,100%@2034,100%@2033] EMC_FREQ 12%@1866 GR3D_FREQ 9%@1300 NVENC 1164 APE 150 MTS fg 5% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35C BCPU@39C thermal@37.4C Tdiode@34.25C VDD_SYS_GPU 456/365 VDD_SYS_SOC 1827/1079 VDD_4V0_WIFI 0/0 VDD_IN 10086/8903 VDD_SYS_CPU 4870/4995 VDD_SYS_DDR 1923/1545
RAM 4927/7859MB (lfb 191x4MB) SWAP 0/3929MB (cached 0MB) CPU [99%@2035,100%@2035,100%@2034,100%@2037,100%@2035,100%@2035] EMC_FREQ 12%@1866 GR3D_FREQ 13%@1300 NVENC 1164 APE 150 MTS fg 1% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35C BCPU@39C thermal@37.4C Tdiode@34.25C VDD_SYS_GPU 532/375 VDD_SYS_SOC 1750/1119 VDD_4V0_WIFI 0/0 VDD_IN 10086/8972 VDD_SYS_CPU 5022/4996 VDD_SYS_DDR 1865/1564
RAM 4912/7859MB (lfb 192x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2035,100%@2035,100%@2033,100%@2034,100%@2034,100%@2034] EMC_FREQ 12%@1866 GR3D_FREQ 1%@1300 NVENC 1164 APE 150 MTS fg 2% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.6C Tdiode@34.25C VDD_SYS_GPU 609/388 VDD_SYS_SOC 1750/1154 VDD_4V0_WIFI 0/0 VDD_IN 10010/9030 VDD_SYS_CPU 4946/4993 VDD_SYS_DDR 1808/1577
RAM 4892/7859MB (lfb 196x4MB) SWAP 0/3929MB (cached 0MB) CPU [99%@2035,100%@2034,100%@2036,99%@2030,100%@2034,100%@2035] EMC_FREQ 12%@1866 GR3D_FREQ 5%@1300 NVENC 1164 APE 150 MTS fg 3% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35.5C BCPU@39C thermal@37.7C Tdiode@34.5C VDD_SYS_GPU 609/400 VDD_SYS_SOC 1827/1189 VDD_4V0_WIFI 0/0 VDD_IN 10429/9103 VDD_SYS_CPU 5020/4995 VDD_SYS_DDR 1904/1595
RAM 4892/7859MB (lfb 196x4MB) SWAP 0/3929MB (cached 0MB) CPU [98%@2034,100%@2034,100%@2034,100%@2034,100%@2035,100%@2034] EMC_FREQ 13%@1866 GR3D_FREQ 10%@1300 NVENC 1164 APE 150 MTS fg 3% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35C BCPU@39.5C thermal@37.4C Tdiode@34.5C VDD_SYS_GPU 456/402 VDD_SYS_SOC 1827/1221 VDD_4V0_WIFI 0/0 VDD_IN 10391/9168 VDD_SYS_CPU 5096/5000 VDD_SYS_DDR 1942/1612
RAM 4892/7859MB (lfb 196x4MB) SWAP 0/3929MB (cached 0MB) CPU [99%@2004,100%@2034,100%@2035,100%@2007,100%@2009,100%@2007] EMC_FREQ 13%@1866 GR3D_FREQ 4%@1300 NVENC 1164 APE 150 MTS fg 2% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35.5C BCPU@39C thermal@37.6C Tdiode@34.5C VDD_SYS_GPU 532/409 VDD_SYS_SOC 1750/1246 VDD_4V0_WIFI 0/0 VDD_IN 10201/9217 VDD_SYS_CPU 5022/5001 VDD_SYS_DDR 1865/1624
RAM 4844/7859MB (lfb 205x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2035,100%@2035,100%@2035,98%@2035,100%@2034,100%@2032] EMC_FREQ 13%@1866 GR3D_FREQ 7%@1300 NVENC 1164 APE 150 MTS fg 3% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.6C Tdiode@34.5C VDD_SYS_GPU 456/411 VDD_SYS_SOC 1827/1273 VDD_4V0_WIFI 0/0 VDD_IN 10125/9258 VDD_SYS_CPU 4870/4995 VDD_SYS_DDR 1904/1637
RAM 4841/7859MB (lfb 205x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2009,100%@2035,100%@2034,99%@2008,100%@2004,100%@2008] EMC_FREQ 13%@1866 GR3D_FREQ 6%@1300 NVENC 1164 APE 150 MTS fg 4% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.6C Tdiode@34.5C VDD_SYS_GPU 532/416 VDD_SYS_SOC 1827/1297 VDD_4V0_WIFI 0/0 VDD_IN 10086/9294 VDD_SYS_CPU 4870/4989 VDD_SYS_DDR 1904/1648
RAM 4833/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2035,100%@2034,100%@2034,98%@2034,100%@2035,100%@2035] EMC_FREQ 13%@1866 GR3D_FREQ 6%@1300 NVENC 1164 APE 150 MTS fg 2% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.6C Tdiode@34.75C VDD_SYS_GPU 685/427 VDD_SYS_SOC 1827/1319 VDD_4V0_WIFI 0/0 VDD_IN 10201/9332 VDD_SYS_CPU 4870/4984 VDD_SYS_DDR 1865/1657
RAM 4835/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [98%@2034,100%@2035,100%@2035,98%@2035,100%@2035,100%@2034] EMC_FREQ 13%@1866 GR3D_FREQ 8%@1300 NVENC 1164 APE 150 MTS fg 3% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35.5C BCPU@39C thermal@37.9C Tdiode@34.75C VDD_SYS_GPU 685/437 VDD_SYS_SOC 1750/1336 VDD_4V0_WIFI 0/0 VDD_IN 10125/9364 VDD_SYS_CPU 4717/4974 VDD_SYS_DDR 1904/1667
RAM 4832/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2035,100%@2034,100%@2034,98%@2035,100%@2035,100%@2033] EMC_FREQ 13%@1866 GR3D_FREQ 9%@1300 NVENC 1164 APE 150 MTS fg 2% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35.5C BCPU@39C thermal@37.9C Tdiode@34.75C VDD_SYS_GPU 609/444 VDD_SYS_SOC 1827/1355 VDD_4V0_WIFI 0/0 VDD_IN 10277/9399 VDD_SYS_CPU 4870/4970 VDD_SYS_DDR 1904/1676
RAM 4832/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2036,100%@2035,100%@2034,98%@2035,100%@2035,100%@2035] EMC_FREQ 13%@1866 GR3D_FREQ 6%@1300 NVENC 1164 APE 150 MTS fg 1% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.6C Tdiode@34.75C VDD_SYS_GPU 685/453 VDD_SYS_SOC 1750/1370 VDD_4V0_WIFI 0/0 VDD_IN 10053/9423 VDD_SYS_CPU 4794/4963 VDD_SYS_DDR 1827/1682
RAM 4833/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2035,99%@2035,100%@2033,100%@2035,100%@2035,100%@2035] EMC_FREQ 13%@1866 GR3D_FREQ 7%@1300 NVENC 1164 APE 150 MTS fg 4% bg 0% PLL@39C MCPU@39C PMIC@100C Tboard@32C GPU@35.5C BCPU@39C thermal@37.9C Tdiode@34.5C VDD_SYS_GPU 532/456 VDD_SYS_SOC 1750/1383 VDD_4V0_WIFI 0/0 VDD_IN 9900/9440 VDD_SYS_CPU 4870/4960 VDD_SYS_DDR 1770/1685
RAM 4836/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2034,100%@2035,100%@2035,100%@2035,100%@2035,100%@2034] EMC_FREQ 12%@1866 GR3D_FREQ 4%@1300 NVENC 1164 APE 150 MTS fg 5% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.9C Tdiode@34.75C VDD_SYS_GPU 609/461 VDD_SYS_SOC 1750/1396 VDD_4V0_WIFI 0/0 VDD_IN 9976/9458 VDD_SYS_CPU 4717/4951 VDD_SYS_DDR 1827/1690
RAM 4833/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2034,100%@2034,100%@2035,99%@2033,100%@2037,100%@2036] EMC_FREQ 12%@1866 GR3D_FREQ 3%@1300 NVENC 1164 APE 150 MTS fg 1% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@36C BCPU@39.5C thermal@37.9C Tdiode@34.75C VDD_SYS_GPU 685/468 VDD_SYS_SOC 1750/1408 VDD_4V0_WIFI 0/0 VDD_IN 10048/9478 VDD_SYS_CPU 4870/4949 VDD_SYS_DDR 1846/1695
RAM 4833/7859MB (lfb 208x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2031,100%@2035,100%@2034,100%@2035,100%@2034,100%@2035] EMC_FREQ 12%@1866 GR3D_FREQ 7%@1300 NVENC 1164 APE 150 MTS fg 2% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.9C Tdiode@34.75C VDD_SYS_GPU 532/470 VDD_SYS_SOC 1751/1419 VDD_4V0_WIFI 0/0 VDD_IN 9710/9486 VDD_SYS_CPU 4641/4939 VDD_SYS_DDR 1808/1699
RAM 4832/7859MB (lfb 209x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2034,100%@2034,100%@2035,98%@2034,100%@2034,100%@2034] EMC_FREQ 12%@1866 GR3D_FREQ 7%@1300 NVENC 1164 APE 150 MTS fg 1% bg 0% PLL@39.5C MCPU@39.5C PMIC@100C Tboard@32C GPU@35.5C BCPU@39.5C thermal@37.9C Tdiode@34.75C VDD_SYS_GPU 532/472 VDD_SYS_SOC 1827/1431 VDD_4V0_WIFI 0/0 VDD_IN 10015/9502 VDD_SYS_CPU 4794/4934 VDD_SYS_DDR 1884/1704
RAM 4835/7859MB (lfb 209x4MB) SWAP 0/3929MB (cached 0MB) CPU [100%@2036,100%@2035,100%@2034,99%@2033,100%@2036,100%@2035] EMC_FREQ 13%@1866 GR3D_FREQ 9%@1300 NVENC 1164 APE 150 MTS fg 1% bg 0% PLL@39.5C M

Hi,
Could you also run the same on r28.2.1 and compare tegrastats?
sudo nvpmodel -m 0
sudo jetson_clocks
sudo tegrastats

Looks like the performance is capped by CPU on r32.3.1. Not sure how it runs on r28.2.1. Comparing tegrastas should bring some information.

After few experiments i have found that bottleneck was in videoconvert part of pipeline (videoconvert ! video/x-raw,format=BGRx, width=1920,height=1080,framerate=60/1 ) . I don’t know why, but in 32.3.1 it works much slower that in previous version. So before this i had bgr frames at input and then they converted to rgba using videoconvert. Now, instead of that i convert frames with opencv:

cvtColor(frame, frame, COLOR_BayerBG2RGBA); 
or 
cvtColor (frame, frame, COLOR_BGR2RGBA);

and use modified pipeline:

( appsrc name=vsrc ! nvvidconv ! video/x-raw(memory:NVMM),format=NV12 
! nvv4l2h265enc MeasureEncoderLatency=true control-rate=1 bitrate=20000000 maxperf-enable=true 
! rtph265pay name=pay0 pt=96 )

and everything works like expected

Hi,
r28.2.1 is with gstreamer v1.8.3 and r32.3.1 is with gstreamer v1.14.5. Probably it it reason for the deviation.