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
I could be wrong, but I think the file size issues might be a problem with HW encoding on Handbrake, not necessarily on other SW.
I vaguely recall on the Handbrake forums, the person that developed the HW Video Toolbox Encoder saying that the file size problem might be a limitation of Handbrake, not necessarily other applications.
Although, I could be wrong about that.
Anyone know if the file size problem is common on other SW using HW encoding?
IF ya all want to do some comparison tests. Someone link to a video file and people can run it through HB then post part of the log file. It won't be scienticific but should be good enough to get a generalize idea how well it works, might be more real anyway. Just an idea.
ˆ These are the videos that the handbrake devs want people to use for benchmarking and comparison.
Recently, I've been doing lots of comparisons because I'm really in the market for a Studio Max.
My findings: VideoToolBox H.265 isn't very good, though it is very fast.
By the time you bump the quality to get good results, you might as well just stick with software H.264 - you'll get about the same file sizes. Though, on a Studio Max, software H.264 is 3x slower than VTB H.265. Software H.265 produces excellent results at much smaller file sizes, but takes 6x as long to encode as VTB H.265.
Recently, I've been doing lots of comparisons because I'm really in the market for a Studio Max.
My findings: VideoToolBox H.265 isn't very good, though it is very fast.
By the time you bump the quality to get good results, you might as well just stick with software H.264 - you'll get about the same file sizes. Though, on a Studio Max, software H.264 is 3x slower than VTB H.265. Software H.265 produces excellent results at much smaller file sizes, but takes 6x as long to encode as VTB H.265.
After looking at the file size differences between software H256 and hardware VTB 10-bit for two same movies, it looks like I was misremembering when I said VTB movies were 1.5-2x larger. Actually, they are 3-3.5x larger with VTB for similar qualities 
Actually, this is nothing new. Hardware encoders have always been worse than software encoders when it comes to quality and file size. That's why I've never used hardware encoders for my family movies, but always x264 and more recently x265.
Unfortunately, many are not aware of this. And also on sites like MacRumors the authors don't seem to know that, because you never read corresponding hints. Of course, this also applies to many well-known YouTubers who only look at the speed but not at the file size and quality.
If you use the software encoders from the big companies like Apple (FinalCut Pro, Compressor, etc.) and Adobe, you will probably still get worse results than with x264 or x265.
Unfortunately, many are not aware of this. And also on sites like MacRumors the authors don't seem to know that, because you never read corresponding hints. Of course, this also applies to many well-known YouTubers who only look at the speed but not at the file size and quality.
If you use the software encoders from the big companies like Apple (FinalCut Pro, Compressor, etc.) and Adobe, you will probably still get worse results than with x264 or x265.
I didn't know this was the case across the board, just thought it was a limitation of Handbrake.
From various tests on Handbrake that I have done, with quality as similar as I could get, the file size of HW encoding would be significantly larger than using SW encoding.
HW encoding would be faster, but the slower SW encoding was worth it for the much smaller file size.
This is for a Plex media server so lowering encode times isn't a big factor for me, but space savings at a decent quality is.
I am not aware of anything else. NVidia NVENC might be an exception in certain scenarios. But I can't say that for sure.
I once read about an M1 patch for ffmpeg, which Handbrake probably doesn't use. With the patch, the video toolbox with ffmepg should be better. But the file sizes should still be a lot larger than with a good software encoder (although possibly better than without patch).
Unfortunately, since the usual tests only compare speed without considering file size and quality, many have completely wrong ideas about encoders. Many are not even aware of the problem. Many don't even seem to know that there can be quality differences and file size differences.
There are even people who are not even aware that there is a hardware encoder and believe that the CPU would be so incredibly fast. If you then compare a Mac with a Windows computer and only use the hardware encoder on the Mac, it is clear that it is much faster.
Last edited:
Comparative tests are made using SAME settings (and of course obtaining SAME results, except time), where encoder type is intended as a setting. Otherwise i can say that fiat 500's lap time its better than that of a F1 while fiat 500 was tested on a short track and F1 was on nurburgring (full)...
Anyway, also the statement that hardware encoder on a mac its faster than windows is wrong. what if that windows pc its using an rtx 3090?
Last edited:
"and of course obtaining SAME results" => What do you mean by that? Quality? Filesize? Encoding time?
Different H.264 (or H.265) encoders lead to different results with the same settings.
Nowhere do I claim that. No idea where you read that?
You seem to have read my post out of context and therefore write such a post under my quote.
Last edited:
All the misunderstandings aside, the fact of the matter is that quality encoding (i.e. software) is much faster on Intel/AMD than with M1 Macs. Unfortunately, that's been true for awhile and continues to be so.
Despite the technical prowess of the M1 series, there's just no replacement for megahertz and cores.
Now, if there were a special prize for the most efficiency while software encoding, Apple deserves the award.
Despite the technical prowess of the M1 series, there's just no replacement for megahertz and cores.
Now, if there were a special prize for the most efficiency while software encoding, Apple deserves the award.
No.
Off the top of my head (dependent on actual settings) I know that the Ultra is ≈40% faster than the Max for 1080p H.265 software encodes. I've never tested the "regular" M1.
The M1 its 3,2ghz with 8 cores that’s more than average intels on macs lineup...
He's right.
To be clear, the M1 system on a chip is NOT a faster chip than intel. It is the combination of ASICs that doe "Apple stuff" REALLY well that matters. But if you aim slighly off to the side you could hit a codec that is not hardware optimized or software that does not necessarily exploit the architecture properly.
I really like the mac pro 2019 because it really does Everything well or well enough. And there are things like afterburner and powerful GPUs (even in MPX format) that put together offer a better "general purpose" computer for someone who wants to do a lot of different things.
Just an update: x265 is now 50% faster on arm64 than it was two years ago. So every benchmark on this thread will have to be redone.
Why are you quoting me when it was storkinsj, I just replied to avoid continuing this pointless discussione based on incorrect assumptions