Somebody on phoronix commented that on his Xeon ffmpeg was severely slowed down by the cumulative fixes. I earlier reported the results from a quick and pointless test, which showed that on my machine the base position was *slower* than with new firmware and PTI fixes.
Then I remembered that I've got scripts, and notes of 'cut' details, that I used when preparing some videos. So here is a short test that uses a little of that to edit a video - and I ran it 10 times to try to judge how variable the numbers are. When ffmpeg is running it tends to use all cores for the main parts, other script code and sox probably only use one core. And yes, this is my Haswell, so presumably that is the "provokes reboot" firmware - so far, it seems ok to me (up for more than 24 hours twice, no problems in shorter uptimes - I've had to reboot to run these tests). Before each run, drop the caches. 0. Convert a video clip from mov to mkv Run ffmpeg to convert to mp4 using x264 and aac with possible video fades in and out, create a wav file from the mov file, use sox to fade the audio in/out and/or reduce the volume, run ffmpeg to create an mkv with the video from the mp4 and the audio from the second wav file. 1. Repeat for a different clip. 2. Use a png to create an 'end' label - use ffmpeg to convert it to mp4 with video fade, use sox to create a silent wav, use ffmpeg to merge them. 3. Use ffmpeg to merge all 3 files (including re-encoding to a lower quality). After each run I remove all the output. All times are in seconds. The standard deviation is from http://www.calculator.net/standard-deviation-calculator.html variant 0 : nopti, older microcode 0.000 57.561 0.001 57.683 0.002 57.877 0.003 57.446 0.004 57.544 0.005 59.494 0.006 59.907 0.007 58.315 0.008 59.390 0.009 57.947 mean 58.3164 variation +2.7% -1.3% stddev 0.92645803406786 variant 1 : PTI, current microcode 1.000 58.577 1.001 57.512 1.002 57.676 1.003 60.838 1.004 58.174 1.005 58.062 1.006 57.920 1.007 57.892 1.008 57.525 1.009 59.266 mean 58.3442 variation +4.3% -1.4% change over base +0.047% stddev 1.0233996504027 variant 2 : initial retpoline, PTI, current microcode 2.000 57.990 2.001 58.015 2.002 62.104 2.003 57.888 2.004 57.994 2.005 57.929 2.006 58.062 2.007 58.227 2.008 59.387 2.009 58.234 mean 58.583 variation +6.0% -1.2% change over base +0.45% change over PTI +0.409% stddev 1.3121555632707 For me, these are figures I can live with. ĸen -- Truth, in front of her huge walk-in wardrobe, selected black leather boots with stiletto heels for such a barefaced truth. - Unseen Academicals -- http://lists.linuxfromscratch.org/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page Do not top post on this list. A: Because it messes up the order in which people normally read text. Q: Why is top-posting such a bad thing? A: Top-posting. Q: What is the most annoying thing in e-mail? http://en.wikipedia.org/wiki/Posting_style
