Compared 4770k/5960x/3950x and Laptop
Posted: Mon Apr 13, 2020 4:49 pm
FWIW - I'm converting the bulk of my movies from ISO to MKV to save space on my 18TB media server (MKV typically takes half the space). I have several high performance (at the time I built them) desktops that I've used for photo processing. Getting familiar with Handbrake I did quite a bit of testing of various settings as well as trying my different systems. I threw in a new 10th gen Intel laptop I recently bought just for fun.....
For comparison, I purposely picked two files that were about the same size but resulted in drastically different MKV output file sizes and processing times.
Inputs are Blu Ray ISO files on hard drive:
Draft Day - 28,330,752 KB; run time 1:50:00; 1980x800 dimension
Little Miss Sunshine – 28,220,800 KB; run time 1:42:34; 1980x798 dimension
Platforms:
1) Ryzen 3950x; 16-core/32-thread; Overclock – all cores at 4.1 GHz (build just completed)
2) Intel I7-5960x; 8-core/16-thread; Overclock – all cores at 4.0 GHz (built 2015)
3) Intel I7-4770k; 4-core/8-thread; Overclock – all cores at 4.2 GHz (built 2014)
4) Dell XPS 13 laptop - Intel I5-10210u; 4-core/8-thread; 1.6 GHz up to 4.2 GHz turbo single core
Source Files:
Handbrake settings:
• Output = MKV
• Filters – all OFF
• Video – H.264; Framerate-Source; Constant Framrate; 2-pass encoding; Very Slow; Tune-None; Profile-High; Level-4.0; Constant Quality; RF-18
• Audio – DTS-HD MA – Auto-Passthru
• Subtitles – English; Forced Only; Burn In
• Chapters – Create chapter markers checked.
Results - hours:minutes:seconds; output filesize; average FPS:
Draft Day
3950x – 0:59:09; 6,899,711 KB; 56.30 FPS
5960x – 1:50:44; 6,901,982 KB; 30.07 FPS
4770k – 3:29:12; 6,920,530 KB; 15.92 FPS
10210u – 5:04:03; 6,920,432 KB; 10.92 FPS
Little Miss Sunshine
3950x – 1:33:30; 18,567,943 KB; 32.91 FPS
5960x – 3:18:51; 18,592,062 KB; 15.47 FPS
4770k – 6:19:34; 18,726,987 KB; 8.11 FPS
10210u – 9:25:49; 18,728,044 KB; 5.44 FPS
Commentary:
For all except the 3950x, the encode consumed 98-100% of CPU resources. On the 3950x, Little Miss Sunshine ran 70%-75% utilization and Draft Day ran 55% utilization.
I experimented with different Handbrake thread parameters and landed on 18 (threads=18) as a good balance.
3950x with threads=18 results:
Draft Day – 1:09:27; 6,917,406 KB; 47.95 FPS
Little Miss Sunshine – 1:53:15; 18,695,279 KB; 27.17 FPS
Anecdotal Observations:
Obviously the 3950x which is my new primary machine, will be most efficient running concurrent encodes. I used the same input file simultaneously running two and three Handbrake sessions.
Running multiple sessions without the thread parameter seemed to produce variable results – the process appears to leverage more/fewer threads at various points in a run. In some cases, it seemed one instance would grab most of the threads and finish much quicker leaving a ‘straggler’ that took much longer to finish. I settled on using the ‘threads=18’ as it gave me good results running 2 or 3 concurrent encodes while not being too restrictive with only a single session running.
For comparison, I purposely picked two files that were about the same size but resulted in drastically different MKV output file sizes and processing times.
Inputs are Blu Ray ISO files on hard drive:
Draft Day - 28,330,752 KB; run time 1:50:00; 1980x800 dimension
Little Miss Sunshine – 28,220,800 KB; run time 1:42:34; 1980x798 dimension
Platforms:
1) Ryzen 3950x; 16-core/32-thread; Overclock – all cores at 4.1 GHz (build just completed)
2) Intel I7-5960x; 8-core/16-thread; Overclock – all cores at 4.0 GHz (built 2015)
3) Intel I7-4770k; 4-core/8-thread; Overclock – all cores at 4.2 GHz (built 2014)
4) Dell XPS 13 laptop - Intel I5-10210u; 4-core/8-thread; 1.6 GHz up to 4.2 GHz turbo single core
Source Files:
Handbrake settings:
• Output = MKV
• Filters – all OFF
• Video – H.264; Framerate-Source; Constant Framrate; 2-pass encoding; Very Slow; Tune-None; Profile-High; Level-4.0; Constant Quality; RF-18
• Audio – DTS-HD MA – Auto-Passthru
• Subtitles – English; Forced Only; Burn In
• Chapters – Create chapter markers checked.
Results - hours:minutes:seconds; output filesize; average FPS:
Draft Day
3950x – 0:59:09; 6,899,711 KB; 56.30 FPS
5960x – 1:50:44; 6,901,982 KB; 30.07 FPS
4770k – 3:29:12; 6,920,530 KB; 15.92 FPS
10210u – 5:04:03; 6,920,432 KB; 10.92 FPS
Little Miss Sunshine
3950x – 1:33:30; 18,567,943 KB; 32.91 FPS
5960x – 3:18:51; 18,592,062 KB; 15.47 FPS
4770k – 6:19:34; 18,726,987 KB; 8.11 FPS
10210u – 9:25:49; 18,728,044 KB; 5.44 FPS
Commentary:
For all except the 3950x, the encode consumed 98-100% of CPU resources. On the 3950x, Little Miss Sunshine ran 70%-75% utilization and Draft Day ran 55% utilization.
I experimented with different Handbrake thread parameters and landed on 18 (threads=18) as a good balance.
3950x with threads=18 results:
Draft Day – 1:09:27; 6,917,406 KB; 47.95 FPS
Little Miss Sunshine – 1:53:15; 18,695,279 KB; 27.17 FPS
Anecdotal Observations:
Obviously the 3950x which is my new primary machine, will be most efficient running concurrent encodes. I used the same input file simultaneously running two and three Handbrake sessions.
Running multiple sessions without the thread parameter seemed to produce variable results – the process appears to leverage more/fewer threads at various points in a run. In some cases, it seemed one instance would grab most of the threads and finish much quicker leaving a ‘straggler’ that took much longer to finish. I settled on using the ‘threads=18’ as it gave me good results running 2 or 3 concurrent encodes while not being too restrictive with only a single session running.