Then you should just get a i7 desktop PC if you plan to do a lot of software encoding. M1 chips are more about power savings.
Last edited:
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
M1 handbrake performance very disappointing...
- Thread starter Damian83
- Start date
- Sort by reaction score
in the preset file, audio settings arent setup properly (i always select passthrough manually everytime). While on the activity log they're are still present, even if the resulting file have audio untouched (passthroughted). just to be sure, at end of every rip i demux audio from original file and final file, to compare them, and they're always identical. If i rip atmos 7.1, i have atmos 7.1 on the final file. Trust me, audio its the thing i care more than video
ehm, ive opened this thread not because i'm complaining about overall m1 speed. I'm fine with bot i5 and m1. I'm complaining about the fact that m1 its slower than i5 even if benchmarks says opposite. And it seems also some people here says the same (m1 faster than i5 on HB), so my point its not to find a way to rip faster. It's to make it faster than my i5, using SAME settings i'm already using
I'm getting 10.09 fps avg on my M1 Max with your settings, at 16854.15 kb/s, Avg QP:18.69 quality.
It is more than double of what you archived from your log and converting a source of much higher complexity (as you can see from the bitrate). I suspect that you M1 is either under serious throttling during the encoding(mostly the case), or your VP9 video source file decodes slowly using the M1 CPU as mine is H264.
I'll run tests to see if I get similar fps rates on my Mini (and let you know, it'll probably be a few days when I've time). But, while I can't comment on the speed of an i5 iMac vs M1 Air, I did run comparisons between my Mac Pro 5,1 12c 3.4GHz and my M1 Mini when I first got the Mini and the two were roughly the same speed for Handbrake (software) encodes and GeekBench5 (around 6700-6900). With some encodes, the MP would be quicker, with others the M1 Mini but the GB5 scores were close enough the same and the encode speeds were the same. But then, neither the M1 Mini or the MP will thermally throttle and the M1 Air does (though I've no idea how long it takes to do so)
thank u in advance if u make some tests and let me know. anyway, as i just said above the speed difference was noticeable almost instantly with no throttling involved. later ill try another movie fragment, as the one ive tested was very dark. most of it was just black. maybe its the reason...
I don't convert 4k files personally, my reasoning being they're already (generally) H.265 10bit so any size savings will be more down to data/quality loss than better compression (I know in your case you're also reducing resolution)
However, I have a set of files I use for speed/size testing across new versions of MacOS, Handbrake and hardware and just realised 2 of them are 2160p sources which I used to reduce to 1080p and I still have the notes of my test results from way back when
one converted on the M1 Mini at 12.5fps, the other at 17.2fps but both sources are way lower bitrate than yours so that probably accounts for the differences. I'll test with something that's around 80GB size and 80Mb/s bitrate
If your video is 1080 it's not worth it to do 265 encoding, stick with 264. Unless you want to drop bitrates to save space. but then you have to deal with longer encodes. 265 is more meant for 4k and above.
I have been meaning to compare my i7 PC to my M1 Max. I have found my PC to always outperform my Mac's by a significant amount. Not fair comparisons as the specs are different, but they are not that far off. Curious to see how the M1 Max stands side by side. Some things just seem to be quicker on Windows, USB is an easy example.
Oh my bad, seemly like that is hevc already. I've tried with one sample of HEVC 4K120 source and is having similar results:
encoded 11026 frames in 931.37s (11.84 fps), 6043.63 kb/s, Avg QP:17.68
The complexity of this file is not as high as my previous one, but still significantly higher than yours, and I'm getting even better fps. Perhaps passively cooled M1 just cannot do this kind of task well due to the thermal throttling.
Isn't that the point of H.265, the fact that you can get the same visual quality with half(ish) the bit rate ?
Just being nosey, what's your reasoning behind not worth encoding 1080p as H.265 ?
No problem, that's why we are here to discuss.
From what I understand 265 was made for larger formats and the benefits work more for them. Sure you can use it on lower formats to potienally save file sizes but since it is literally half the dimenson then less area to work with to save image quality. I'm not sure if it is a fact you can half the bit rate compared to 264 with 265 and get similar results. I've tested it before and I couldn't get similar results by doing it that way. Plus as mentioned longer encode times was not worth it for similar results and not much file size savings. I usually revisit this workflow once a year or when I major update to handbrake comes around. But so far for me 264 is just fine.
I think I used CQ 46 or 47.
"1.5-2x bigger for VT."
Still much bigger (though better than my experience)
@orionquest:
I think it also makes a big difference what kind of H.264 encoder you use. If you use x264, the difference between H.264 and H.265 might be much smaller than if you use another H.264 encoder. The H.264 encoders of the big companies are as far as I know worse than x264 (at least in the past => the Apple software H.264 encoder, which is offered e.g. in FinalCut or Compressor, was in the past even particularly bad - the Apple encoder could for years e.g. no Cabac => nevertheless many users used it => like now many use the VideoToolbox, although it cannot keep up in terms of file size and quality with x264 and x265).
Last edited:
That was mostly my understanding too, but just because something has greater benefit on 2160p doesn't mean it doesn't have significant benefits on 1080p too, just maybe not quite so great.
Fair enough. I've not used Handbrake with a bitrate setting for over a decade (and never with H.265), just simply with the "Constant Quality" slider. As such, the bitrate is left in the hands of people who know more than me and it seems to need little adjustment whether encoding with H.264 as H.265, both producing similar results, just one being much smaller than the other, and yes, taking longer to do it but in my view worth the wait.
Thanks
True I don't doubt some encoders are better then others, and I never use VTB for that reason.
I tend to use Handbrake a lot to transcode from random formats into good mp4's
Also I should add, I generally always use Average Bitrate instead of Constant Quality, so I might be already at the lower end of the quality scale and this might be why 265 isn't such impact on the stuff I do.
Just sharing my experience.
The Benchmarks are accurate. The Benchmarks don't tell you how the hardware will perform with any software other than the benchmark software. In you example, a test from an oval track to nurburgring won't tell you how the car will perform off-road in the in Colorado Rockies. Different use cases require different designs.
If you have a specific use case in mind, always look for tests on the specific software you actually want to use.
ok ive made another test, this time ive choosed a colourful fragment with lot of action scenes. This time m1 started faster, and slowed after some minutes. The winner, again, was i5 by a 5% gap. This gap of course will be higher on a full movie, meaning that the fanless system seriously lowers performance. But now the interesting thing: ive also measured real time power consumption with the ups... both system restarted and with no other apps running, monitors off:
i5: 120w
m1: 20w at the beginning, then 10w during throttling!!!!
its really impressive! i was going to rip movies overnight in the garage, so i dont waste money on energy, but with so low consumption i can do this at home!
i5: 120w
m1: 20w at the beginning, then 10w during throttling!!!!
its really impressive! i was going to rip movies overnight in the garage, so i dont waste money on energy, but with so low consumption i can do this at home!
Are you certain that you are running the Apple Silicon-optimized version of Handbrake? It really should be faster than the Core i5 machine.
I got a similar result to the OP with different Intel Macs versus the M1 Mac Mini, and yes, I was using the AS optimized version of Handbrake.
While I have tested multiple Intel Macs, the example that I used earlier in the thread was the Late 2012 iMac with the 3rd Gen i7.
Geekbench has a multi-score of 3098 versus 7434 for the M1, meaning the M1 should be almost two and a half times faster than the 3rd Gen i7 in multi-core performance.
But, real world Handbrake encodes are at best only about twice as fast on the M1 with the tests I ran. Worse is that the M1 performance dramatically drops for more intense encodes, while the Intels stay steady. The hardest test I did, the M1 was only 50% faster than the 3rd Gen i7.
I suspect that if I would push the machines even harder, the 8 year older iMac might have caught up the the M1 in average FPS.
I am not surprised by this, as you were using a 9th generation i5 versus what I posted on here, a 3rd generation i7.
All the Intels I tested has the initial "burst" when first hitting start, but after about 10-30 seconds, would average FPS would drop.
The difference between the machines I posted about (Late 2012 iMac with i7) and the your iMac (2019 iMac with i5) is that the M1 is faster with Handbrake than the 3rd Gen i7 in my tests, but not faster than your 9th Gen i5.
Your tests probably has the M1 in an even worse position when compared to your 2019 i5, considering you were using the fan-less MBA.
But, does that mean the M1 is slower than your 2019 iMac? I don't think so, it just means that the M1 MBA is slower than your iMac when it comes to Handbrake.
ok but surfing web, reading mails, running messaging apps etc, doesnt need a fast pc... and most of us usually use a pc just for these standard things... but no matter, as i wrote above, m1 uses 10w during ripping so i can rip my entire collection without regreting ive wasted energy for shrink movies (most of them ill not even watch again) instead of, lets say, mining
One difference is might be that x86 has all sorts of special instructions meant for different purposes, while Apple breaks that logic out into coprocessors. So the Geekbench results are general number crunching, but there’s probably some x86 instructions designed for critical video compression operations— so if the M1 and the x86 are doing software only encodes, x86 has tools that M1 doesn’t that also don’t show up in the benchmarks.
Now, the M1 also has dedicated hardware encoders which, by all accounts, smoke the x86 on encode time, but it’s just not optimized for file size in the way people are configuring their software encode.
Last edited:
Yeah, but at the same time, I would bet that the M1 MBA would be noticeably faster than your 2019 iMac at these task that really benefit from fast single-core performance.
Another thing to keep in mind when comparing your 2019 iMac to the M1 MBA, is that the M1 is an entry level chip, while the 9th Gen i5 would be considered mid-tier, imo. That, and the lack of the fan in the M1, it is not surprising that some things, the mid-tier 2019 iMac will out perform a 2020 entry-level model Mac.
Honestly, the M1 has amazing performance for the price point, especially the Mac Mini. Keep in mind that the M1 Mac Mini, which was the cheapest Mac in the line up, had the fastest single-core performance out of any Mac ever, and kept that title for a year.
My biggest complaint about the M1 had more to do with the industry as a whole, and not specific with AS. Up until about the 2012, there was (relatively) big increases in performance about every new incremental release. This was true for the PPC years, and into the Intel years, but around 2013, and every year after, the performance increase was not as dramatic. This was around the "Haswell" or 4th gen for Intel.
I recall reading somewhere that Intel was going to focus on power consumption over performance beginning with Haswell, and I think that is what happened. The gains just disappeared.
This is why in this thread and others, I use the Late 2012 iMac with the 3rd Gen i7 in my examples when talking about comparisons, as apposed to the other Macs (and one PC) I tested. There just wasn't a big year to year difference after 2012. The performance difference between a Mac from 2012 and a Mac from 2020 is not that big when compared to a Mac from 2004 and 2012, or even a Mac from 2009 and 2012.