M1 Pro v. M1 Max v. M1 Ultra

HandBrake for Mac support
Forum rules
An Activity Log is required for support requests. Please read How-to get an activity log? for details on how and why this should be provided.
Post Reply
wills
Posts: 5
Joined: Wed Mar 09, 2022 8:16 pm

M1 Pro v. M1 Max v. M1 Ultra

Post by wills »

Description of problem or question:
The M1 Pro has a single video decode engine, the Max two & Ultra four. Does Handbrake's speed scale with the additional encoders if using the VideoToolbox hardware encoder? Like is the M1 Max roughly 2x as fast as M1 Pro and M1 Ultra 2x as fast as M1 Max?



Steps to reproduce the problem (If Applicable):
N/A



HandBrake version (e.g., 1.0.0):
1.5.1



Operating system and version (e.g., Ubuntu 16.04 LTS, macOS 10.13 High Sierra, Windows 10 Creators Update):
macOS 12.3 Monterey



HandBrake Activity Log ***required*** (see How-to get an activity log)

Code: Select all


Deleted User 11865

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by Deleted User 11865 »

Not sure, but I guess probably not, unless you're doing several parallel encodes.
wills
Posts: 5
Joined: Wed Mar 09, 2022 8:16 pm

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by wills »

Yeah, I'm having a hard time finding any real comparisons or benchmarks in this regard. There is this article, where the reviewer says the M1 Max shows a 53% better Handbrake VTB encoding speed over the M1 Pro, but he really doesn't understand why and makes some guesses that the GPU/cooling is better.

https://www.macworld.com/article/622515 ... eview.html
rollin_eng
Veteran User
Posts: 4840
Joined: Wed May 04, 2011 11:06 pm

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by rollin_eng »

There is this discussion here:

https://github.com/HandBrake/HandBrake/issues/4028

Some benchmarks here and hopefully more to come:

viewforum.php?f=9
Deleted User 11865

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by Deleted User 11865 »

I mean, video decoding/filtering/scaling and all audio/subtitle processing is still done on the CPU…
wills
Posts: 5
Joined: Wed Mar 09, 2022 8:16 pm

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by wills »

Is there any way to see in macOS what is doing the actual work when running Handbrake? I looked at the Activity Monitor, but it doesn't show the video encoder specifically, and CPU usage is fairly low while doing HB encodes with VTB.
p_alex_r
Posts: 10
Joined: Mon Aug 16, 2021 11:28 am

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by p_alex_r »

I've been running my own personal tests over the past couple of weeks, here's a synopsis ...

M1, 1080p H.264 -> 1080p H.265 using VTB, approx 190fps
M1 Ultra, 1080p H.264 -> 1080p H.265 using VTB, approx 310fps

My reading of this is that the Ultra is only using 1 encode engine and that wguke the M1 cannot swamp its encode engine, the CPU is running 90+%, the Ultra sure can (sub 40% ish CPU, I wasn't watching it I got all the numbers from the log files) and it must be the encode engine that's the limiting factor.

Increasing the number of parallel encodes (I tried 2,3 & 4) and the Ultra managed around 520fps at best, with 3 parallel encodes being the quickest.

There are 8 files of various types in my test dataset, 1080p, 4k, h.264 & H.265. It's playback runtime is 1:57:58. The M1 completed the whole lot with a single thread in 43min 32 secs. The Ultra with a single thread, 18min 15 sec. The Ultra with three parallel encodes, 12min 32 secs (4 encodes was 16secs slower)

I'll let someone else work out the % increases :p
User avatar
s55
HandBrake Team
Posts: 10350
Joined: Sun Dec 24, 2006 1:05 pm

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by s55 »

M1 can be CPU bound, so if you want to see how fast VTB can go, disable all filters, audio encoding etc. See how it goes.
rollin_eng
Veteran User
Posts: 4840
Joined: Wed May 04, 2011 11:06 pm

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by rollin_eng »

Might be good to post your results with logs in the benchmark forum.
Falcon81
Posts: 6
Joined: Thu Mar 10, 2022 9:57 am

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by Falcon81 »

Has anyone tested x264/x265 on the Binned M1 Pro 8 Core vs M1 Pro 10 Core? Wondering if those 2 extra cores make any difference on HB?

I picked up a base 14" MB Pro with the Binned 8 Core CPU. The video toolbox is super fast however I do see some noticeable blocking that I don't seem to get with software encodes. It's present at QC46, 50 and 52. which produce bigger files and higher bitrates than using x265 on a CPU encode at RF22 with no blocking artifacts. for this reason I'm thinking of scrapping Video Toolbox for my 1080p H264 to 1080p H265 conversions.
mduell
Veteran User
Posts: 8187
Joined: Sat Apr 21, 2007 8:54 pm

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by mduell »

Yes, an extra 2 performance cores matter for most reasonable HB software encoddes.

VTB is fast garbage; but it's fast.
Falcon81
Posts: 6
Joined: Thu Mar 10, 2022 9:57 am

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by Falcon81 »

mduell wrote: Mon Sep 05, 2022 11:38 pm Yes, an extra 2 performance cores matter for most reasonable HB software encoddes.

VTB is fast garbage; but it's fast.

Hey thanks for the response, do you have any idea how much performance % an extra two cores would make? going from 8 to 10? Handbrake documentation suggests diminishing returns after 6 cores so I'm thinking no more than 10%?
Falcon81
Posts: 6
Joined: Thu Mar 10, 2022 9:57 am

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by Falcon81 »

Falcon81 wrote: Tue Sep 06, 2022 1:47 am
mduell wrote: Mon Sep 05, 2022 11:38 pm Yes, an extra 2 performance cores matter for most reasonable HB software encoddes.

VTB is fast garbage; but it's fast.

Hey thanks for the response, do you have any idea how much performance % an extra two cores would make? going from 8 to 10? Handbrake documentation suggests diminishing returns after 6 cores so I'm thinking no more than 10%?

Also the point about "fast garbage" - what is it about Hardware encoding that is stopping it from achieving the same or better results than software encoding?
mduell
Veteran User
Posts: 8187
Joined: Sat Apr 21, 2007 8:54 pm

Re: M1 Pro v. M1 Max v. M1 Ultra

Post by mduell »

2 extra performance cores, coming from 6, gives you at best +33% performance. Scaling isn't perfect, and the efficiency cores are worth something, so figure maybe +20% in real life. Diminishing doesn't mean 0%, it means anything less than a pro-rata core.

The hardware is fixed function, so it doesn't implement the same algorithms the software encoders do, particularly at their slower settings.
Post Reply