Environment: * ICL GT2 B4 , graphics id 0x8a52 * 5.1.0-rc2-drm-tip : 79df6ae8e7d6fc94965a14e1fa04898954d40316 * https://github.com/intel/media-driver/ : 8773ef5 * https://github.com/intel/gmmlib : e65dff3 (includes GMM MOCS table fix) * https://github.com/Intel-Media-SDK/MediaSDK : 2c780ea We're using the following script to fix CPU/GPU frequency on ICL: #!/bin/bash set -ex cpupower frequency-set --governor performance cpupower frequency-set --min 1000000 cpupower frequency-set --max 1000000 echo 400 > /sys/class/drm/card0/gt_min_freq_mhz echo 400 > /sys/class/drm/card0/gt_max_freq_mhz echo 400 > /sys/class/drm/card0/gt_boost_freq_mhz However media workloads produce very different performance from run to run. perf stat -a -e i915/bcs0-busy/,i915/rcs0-busy/,i915/vcs0-busy/,i915/vecs0-busy/,i915/actual-frequency/,i915/requested-frequency/ ./sample_encode h264 -i input.yuv -qpi 26 -qpp 26 -qpb 26 -async 1 -hw -w 720 -h 480 -n 300 -r 1 -vaapi -g 100 File output is disabled as -o option isn't specified libva info: VA-API version 1.5.0 libva info: va_getDriverName() returns 0 libva info: User requested driver 'iHD' libva info: Trying to open /opt/intel/mediasdk/lib64/iHD_drv_video.so libva info: Found init function __vaDriverInit_1_4 libva info: va_openDriver() returns 0 Encoding Sample Version 8.3.26. Input file format YUV420 Output video AVC Source picture: Resolution 720x480 Crop X,Y,W,H 0,0,720,480 Destination picture: Resolution 720x480 Crop X,Y,W,H 0,0,720,480 Frame rate 30.00 Bit rate(Kbps) 2503 Gop size 100 Ref dist 1 Ref number 0 Idr Interval 0 Target usage balanced Memory type vaapi Media SDK impl hw Media SDK version 1.29 Processing started Processing finished Performance counter stats for 'system wide': 0 ns i915/bcs0-busy/ 4,078,817,486 ns i915/rcs0-busy/ 748,226,339 ns i915/vcs0-busy/ 0 ns i915/vecs0-busy/ 791 MHz i915/actual-frequency/ 3,137 MHz i915/requested-frequency/ 7.912914519 seconds time elapsed It looks like frequency is not actually fixed while the same approach works fine on SKL.
Most likely has the same root cause as Bug 108059
This was a problem with the HW, was proven to work in other machines.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.