Broadcom BCM2711 (Raspberry Pi 4)

Post your testing results with HandBrake.
Post Reply
Pluto
Posts: 3
Joined: Thu Jul 29, 2021 3:21 pm

Broadcom BCM2711 (Raspberry Pi 4)

Post by Pluto »

Machine: Raspberry Pi 4
CPU: Broadcom BCM2711 SoC / ARM Cortex-A72 CPU
Memory: 4GB 1600 MHz / LPDDR4-3200 SDRAM
Speed: 1.5 GHz / Not Overclocked
Number of CPUs: 4 core / 4 threads

HandBrake Version: 1.2.2 (Latest from rasbpian/buster repo)
Decode Format: MPEG-2
Decode Container: MKV and Decrypted DVD ISO
Encoder: H.264, slow, quality 20, film filter, high profile, level 3.1
Output Container: MKV
Video Size & Settings: 720x480, Decomb

Results:

The Hunger Games (2h22m) - 15.34fps - Log
Harry Potter and the Sorcerer's Stone (2h32m) - 15.4fps - Log
Harry Potter and the Chamber of Secrets (2h41m) - 13.09fps - Log
Harry Potter and the Prisoner of Azkaban (2h22m) - 15.73fps - Log
Harry Potter and the Goblet of Fire (2h37m) - 13.18fps - Log
The Journey of Natty Gann (1h41m) - 12.07fps - Log
National Lampoon's Christmas Vacation (1h37m) - 11.32fps - Log
Something Wicked This Way Comes (1h35m) - 12.97fps - Log
Sixteen Candles (1h33m) - 12.08fps - Log

Essentially I'm ripping DVDs I personally own and then here I'm converting them with HandBrake into an H.264 format. It seems after the first day of encoding though, it's slowed down 15-20% (I do have a fan installed to prevent throttling). So after today (the second day) if it does not post any faster speeds, I'll try debugging it.

I have over 100 discs to go through, so the Raspberry Pi is my best option (even if slow) because I can cheaply and efficiently run it 20 hours a day and monitor the results remotely. My alternative is a laptop that I take with me to work. The main benefit over a full-size PC is during a hot summer like we have here, this doesn't warm up the room noticeably.
User avatar
s55
HandBrake Team
Posts: 10350
Joined: Sun Dec 24, 2006 1:05 pm

Re: Broadcom BCM2711 (Raspberry Pi 4)

Post by s55 »

Note, source to source variance is normal. Typically the start/end of sources can spike the framerate a bit as there are usually simpler scenes, credits etc. More complex scenes will cause a drop.

Also be aware that decomb if it's doing stuff, can vary a lot from source to source as well.

Numbers seem reasonable to be honest.
Pluto
Posts: 3
Joined: Thu Jul 29, 2021 3:21 pm

Re: Broadcom BCM2711 (Raspberry Pi 4)

Post by Pluto »

s55 wrote: Thu Jul 29, 2021 6:00 pm Note, source to source variance is normal. Typically the start/end of sources can spike the framerate a bit as there are usually simpler scenes, credits etc. More complex scenes will cause a drop.

Also be aware that decomb if it's doing stuff, can vary a lot from source to source as well.

Numbers seem reasonable to be honest.
I put together the logs side by side of some of the most recent files I converted but in a more consumable way, and here are the results. They seem very consistent.

Click for full size
Image


First of all, I highlighted in yellow when it was obvious that Handbrake was impacted by external factors like SCP'ing additional videos into the system for conversion. So those results can mostly be ignored. In addition, as you said the credits of movies definitely causes spikes.

Notice the "Log Avg". This is the average rate the log said it took to encode the file. My calculated average FPS is of decoding the file (from the lines prefixed with "sync:". I did notice one more thing, which is when you divide my FPS average from the log's average, it is a consistent number of just over 1.44. Any time it drops below that you can assume it was because I paused encoding for some amount of time causing my calculated FPS to be lower than it actually was.

The one thing that's bugging me is why was I decoding 1.44x more frames than encoding? It was my understanding that the output would have a 25% higher number of frames (the log's FPS implies instead that the input has a 44% higher number of frames). I checked and found the input definitely has a framerate of almost exactly 23.976fps (the standard for DVDs). My output is configured to be 30fps constant, and the files definitely end up that way. Shouldn't the decoding rate (24fps video) be 0.8x the FPS of the encoding rate (30fps video)?

See here for the full log for "Super 8" as an example.
Pluto
Posts: 3
Joined: Thu Jul 29, 2021 3:21 pm

Re: Broadcom BCM2711 (Raspberry Pi 4)

Post by Pluto »

Never mind, ignore the last part. Excel was just being stupid with its math. It is the decoding rate that HandBrake calculates in its logs for average FPS. My numbers come out to exactly 1 when looking at the total average (except where I paused encoding).
user790
Posts: 8
Joined: Sat Aug 06, 2022 5:25 pm

Re: Broadcom BCM2711 (Raspberry Pi 4)

Post by user790 »

I looked around for a raspberry pi build of handbrake recently, but didn't find anything I could get to work. Where did you find yours?
mduell
Veteran User
Posts: 8187
Joined: Sat Apr 21, 2007 8:54 pm

Re: Broadcom BCM2711 (Raspberry Pi 4)

Post by mduell »

Build the flatpak?
Post Reply