It's not a hardware problem. The title of the post is incorrect and you should have posted in the software section.
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 Max 32gb. It starts to make me sick.
- Thread starter elbateria
- Start date
- Sort by reaction score
I'm not the one who made the thread or video, don't get mad at me. The hardware is simply not working as advertised
But how can you be sure it's the hardware? The FCP and encoder are not the same - they might be the same version and all, but one is written for Intel and the other for Apple Silicon.
To go back to OP's original complaints, I've found out you can disable hardware encoding for Davinci, and ran the same export again: the result is that the encode takes almost 2,5 times longer (1:38 cpu vs 43 hardware), and during the software encoding a Videotoolbox Encode process shoots up to 500% CPU.
This tells me that Davinci on OP's machine is definitely using dedicated hardware and looking at the cpu usage is not a good way of telling whether it's working properly.
What I could not find anywhere is a benchmark of how fast should a fast h264 encode of a comparable Davinci project be. I've read on macrumors that 90fps speed for a 4k/60 project is good, but OP's thinks is horrible...
These machines have been tested and reviewed for three months now. The hardware works fine. Software is another thing. Apps and files can come in so many different configs and have so many different factors than sometimes bugs and performance issues can arise. I can get a video file from one camera and it flies through edit and render to whatever format. I do the same thing with a file from another camera with the SAME codec and then performance isn't the same.
People need to understand these complexities before just randomly blaming hardware with no educated reason. Creating decoding and encoding engines is easy. The software part and supporting hundreds of cameras is the difficult part.
Even software itself between platforms is sketchy sometimes. I once encoded some videos for an installation with ffmpeg on a pc and later the same day with ffmpeg on Mac. Same settings, possibly even the exact same ffmpeg version. One set of files worked properly, the other had some weird playback issue. I know, ffmpeg.
I was just browsing the blackmagic forums to see if anyone has insights encoding speeds there, and someone wrote something relevant to my original point about “pros”, which basically was something along the likes of: “encoding speed is irrelevant, it’s the last step before delivery and it’s going to be clogged by whatever effects you have on your project before hitting the hardware ecncoder limits. What you want is decode performance in order to have a responsive timeline which really is where the creative work happens”.
I could not agree more. The new beta native version of Ae is a world apart from the Rosetta version and makes actual productive work 100 times better. Render times on the same machine are almost halved.
Honestly when it’s time to encode for delivery I am ready for that handbrake cocktail and I don’t spend my time staring at the progress bar.
This is not to say that these encoding times cannot be better, they probably should beat a 5 years old intel machine, but it seems apple really listened to the pros, and made a machine that makes editing a breeze. I doubt anyone went there complaining about encoding times. If you’re an editing pro you spend your days on the timeline, not in the encoder. If you’re a pro encoder, you probably spend your days in a render farm.
I could not agree more. The new beta native version of Ae is a world apart from the Rosetta version and makes actual productive work 100 times better. Render times on the same machine are almost halved.
Honestly when it’s time to encode for delivery I am ready for that handbrake cocktail and I don’t spend my time staring at the progress bar.
This is not to say that these encoding times cannot be better, they probably should beat a 5 years old intel machine, but it seems apple really listened to the pros, and made a machine that makes editing a breeze. I doubt anyone went there complaining about encoding times. If you’re an editing pro you spend your days on the timeline, not in the encoder. If you’re a pro encoder, you probably spend your days in a render farm.
Last edited:
I ‘m currently doing different test, I will post them. The problem is not constant low performance, the problem is inconsistant performance and annoying bugs. Sometimes cpu uses 90% sometimes 30%. USB4 SSD performance is also random. I read about Monterey is having problems with CPU resources assignment, maybe this can explain everything.
About render times doesn’t matter for pros. What is a pro? A guy who is editing movies in L.A? This guy most probably is not using a laptop for tons of reasons. If you are working in advertising or content creation industry, render performance matters a lot. You’re constantly changing text legals, claims, CTAs, shot durations, framing, making versions… and every change should be repeated for 15 different languages, 4 different aspect ratios for mobile, social, channel and tv and uncountable resolutions. You should see some Compressor or AME queues before say “render times doesn’t matter”.
I don't think monitoring the cpu is a good way of estimating performance, because I'm not sure whether the dedicated chip is part of the cpu readout. When I disabled hardware encoding in Davinci the encoding process occupied 500% cpu but the export time was three times slower than the hardware one, which occupied only 200% or so.
honestly if you have these crazy render queues you should probably not do this kind of work on a laptop.
But now that I have you here: so you're getting 95fps export from Davinci on a 4k/60fps timeline.
Do you have any other reference numbers of what is a great export fps performance? I could not find anything reliable.
So, I've found this video:
This is an export test in Davinci between a base 16 inch pro and an MSI GE66 Raider 11UH (gaming laptop with an i7 11800H and a RTX 3080), it costs something like 3800 euros. Fair comparison? You tell me, I know very little about pc specs. But it's a laptop and it costs a lot!
Well, the results are ridiculous: the M1 exports the same 1 hour project in just under 10 minutes, the pc takes 34 minutes.
Well, the results are ridiculous: the M1 exports the same 1 hour project in just under 10 minutes, the pc takes 34 minutes.
Alright now I'm officially in the rabbit hole. I've found this other video where a guy compares h264 export on Davinci 17 on an RTX 3090.
Project is a 4k timeline 1:04 minute long. Key detail: it's 24fps.
The Nvidia based encoding takes 19 seconds, the CPU based one takes 1:01.
So let's make a little math. 1:04 at 24 fps is 1536 frames. 1536 frames in 19 seconds is roughly 80fps speed during rendering with the Nvidia hardware encoder.
OP's project is 4k (at 60fps) and he's getting over 90fps during h264 exporting.
Does this mean that the M1 export is faster than a RTX3090?
Project is a 4k timeline 1:04 minute long. Key detail: it's 24fps.
The Nvidia based encoding takes 19 seconds, the CPU based one takes 1:01.
So let's make a little math. 1:04 at 24 fps is 1536 frames. 1536 frames in 19 seconds is roughly 80fps speed during rendering with the Nvidia hardware encoder.
OP's project is 4k (at 60fps) and he's getting over 90fps during h264 exporting.
Does this mean that the M1 export is faster than a RTX3090?
I don’t tested the RTX 3090, but export 4K footage to 4K H264 in a Ryzen 3900X + 5700XT was around 100fps. This is a 2020 mid price combo. Top equipment like 2080ti was way faster, around 150fps. Suposedly 3080 is around 6-8% faster encoding than 2080ti. So 90fps is not that bad but far from awesome. I’m pretty sure M1 Max is not faster exporting H264 than RXT3090 with a top CPU desktop like 5950x, in fact, I received feedback from some guys saying it cuts M1 Max H264 render times barely in a half. Ok, it’s not fair, desktop computer vs laptop, but I don’t understand this videos saying M1 Max is faster than 3090 exporting H264/H265.
PS: 90fps is the best scenario, it depends of a lot of factors depending how CPU resources are distributed when you export, so you get mixed results.
www.pugetsystems.com
PS: 90fps is the best scenario, it depends of a lot of factors depending how CPU resources are distributed when you export, so you get mixed results.
Premiere Pro 14.2 H.264 and H.265 Hardware Encoding Performance
In the 14.2 version of Premiere Pro, Adobe has added support for GPU-based H264/H.265 (HEVC) hardware encoding with both NVIDIA and AMD GPUs. How well does this feature work, and how much faster is it than the previous hardware encoding that utilized Intel Quicksync?
www.pugetsystems.com
Probably I can live having almost the same performance in my new M1 Max (3800€) than in my old PC (1800€) because this is a portable computer. But the problem is about consistency. Using thunderbolt 3 nvme (1400/2800) doubles exporting times. Why? There is no reading or writting bottleneck here. Use Zoom or recording screen sends performance near to zero. Why?
Ah, but render times are completely different from export times. As an editor for TV, I will take timeline performance over export times every day of the week and twice on Sundays. One saves hours of time, the other, a few seconds.
Oh, man, come on.
I've looked at the link you posted from Puget (a company building pc desktop workstations for video) and if you look at the results, the closest situation to your initial 'benchmark' is this: Prores to H264:
the 2080ti gets around 100fps.
Uhm you're telling me you launch an encoding process from Davinci and at the same time on the same machine you make a zoom call or screen record? two activities that encode video in real time? Yeah, no wonder something has to give up, and the system picks the non-realtime task.
Maybe this is all a terrible dream.
1) these are still desktop numbers, with very high power draws (and prices!). You can browse Puget's own benchmarks and you can see that their custom built desktops with 850w go head to head with the macbooks. In Premiere pro there's a score of 1236 for the m1 mx vs 1300 for an i9 12900k with an rtx 3080, so pretty much a draw. (see here: https://www.pugetsystems.com/benchmarks/?age=0&benchmark=premiere&application=&specs=#results-table )
2) not all compressions are the same! this one is an h264 with high bitrate going down to a still pretty high bitrate. the higher the target bitrate, the less the compression engine has to work. Plus the source material is relevant: not all codecs and contents are the same. It makes sense to compare these numbers only with the exact same source materials and target compression.
At this point I'm convinced you had unrealistic expectations. It's too bad, but by all accounts you got yourself a pretty decent machine.
2) not all compressions are the same! this one is an h264 with high bitrate going down to a still pretty high bitrate. the higher the target bitrate, the less the compression engine has to work. Plus the source material is relevant: not all codecs and contents are the same. It makes sense to compare these numbers only with the exact same source materials and target compression.
At this point I'm convinced you had unrealistic expectations. It's too bad, but by all accounts you got yourself a pretty decent machine.
which you can do by downloading the Davinci benchmark (https://www.pugetsystems.com/labs/articles/PugetBench-for-DaVinci-Resolve-1523/) that includes the source materials, but I fear you'll be disappointed if you don't get the same results as a mini-fridge sized machine that costs more and consumes more power. Oh, and that you have to buy an extra monitor for.
No-one seems to know if he enabled his GPU or not in render settings on the PC - that makes a huge difference.
From my own tests (and the work I actually do for my income which is the key thing for me) on rendering out HD files to h264 mp4's, I got this in Davinci Resolve:
Job one with a little NR, and simple colour correcting and a little sharpening:
PC (Ryzen Threadripper 3970X / 3090 / 64MB Ram): 3 mins, 16 secs
13" M1 Macbook Pro (8‑Core CPU and 8‑Core GPU / 16GB Ram): 24 mins, 06 secs
16" M1 Macbook Max (10-Core CPU 32-Core GPU 64GB Ram): 5 mins, 12 secs
Job Two with a max NR, and more colour correcting and a little sharpening:
PC (Ryzen Threadripper 3970X / 3090 / 64MB Ram): 10 mins, 50 secs
13" M1 Macbook Pro (8‑Core CPU and 8‑Core GPU / 16GB Ram): 1 hour, 20 mins
16" M1 Macbook Max (10-Core CPU 32-Core GPU 64GB Ram): 17 mins, 31 secs
The M1 did do well, prob better than I thought but on some projects, it just doesn't work as fast despite being the same file type which is weird or maybe some bug?
My main reason for returning though is because its a laptop and a lot of the price is in the screen, keyboard and trackpad and I'll almost never use them as I use it on a big external with a normal keybaord/mouse.
I will probably buy the new Mini or Pro providing the Mini is of equal power or the Pro isn't priced too high.
I gather english is not your first language, but I wrote that you can download the benchmark and in it you can find the media assets they use: you can import them in da vinci launch the compressions as they are in the benchmark image you put up and see the performance for yourself.
This is all very vague: is the performance underwhelming in the timeline? or the encoding times are disappointing when compared to a desktop with very different features? As someone else wrote here, if you base your evaluation on encoding times, and this is so important to your bottom line, picking a laptop in the first place is a questionable choice.
I see.
Also no one here is talking about the quality issue of hardware vs software encoding, which is probably too much of a pro issue!