Is anyone running Video AI on a Mac Mini M4Pro ? How’s it going ?
From the support side we have not seen too many users running M4’s at this time. Did have a user who tried it and found out they had a faulty hard drive in their system that had to be sent back but have not gotten an update from them after the replacement.
Very disappointing. It is cheaper, with better overall speed, to buy 2 x base spec Mac Minis instead. You would then have 32 cores of neural engine as well as 20 cores each of CPU and GPU. Compared to only 16 core neural engine, 12 core CPU and 16 core GPU for more than twice the price for the cheapest Mini M4 Pro.
(The licence for TVAI allows it to run on 2 machines at the same time.)
Thanks.
Andy
In general, TVAI seems less optimized (for performance) on Macs as it is for NVIDIA - maybe similar might be true for AMD.
Especially the Iris 2x model (and likely the other newer models as well) is lacking in speed.
That said, the advantages of M4 vs. M3 vs. M2 and even M1 are not THAT big - so it might be worthy looking for an older architecture but with higher core count than a similarly priced M4 system (but beware: TVAI doesn’t really scale very well with high GPU counts at AppleSilicon, either).
If you’re looking for the best performance you should go for a windows system with a RT at least x70 NVIDIA card - those have higher processing power and the software is better optimised.
I have been running on Mac mini M4 pro for some time now and I am pretty disappointed. Looking at the activity monitor the Mac is not doing much. I contacted support en they told me everything is ok. Looks like they would rather sell cloud conversion because that is a complete new source of revenue stream.
Only benefit of the M4 is the low power consumption compared to a Intel PC and a high performance Nvidia card.
Are you processing mainly standard definition sources? If so, TVAI will try to use the Neural Engine if possible (due to its energy efficiency) but that doesn’t show its usage in any of Apple’s own activity monitors. You can “force” the GPU cores into use by reducing the max memory to 10% in the TVAI settings.
If you’re processing HD and above, the GPU cores will usually be used anyway unless you’re in Low Power mode. You can monitor the GPU usage from the Activity Monitor and then use cmd-4 to show the GPU history.
Regards
Andy
I didn’t know you can do that. So set it to 10% GPU kicks in. What about the Neural Engine? Is it still utilized even at 10%. I’m gonna try it anyway and see if that actually does it. iMac M3 8/10 core.
Using free tools such as ASITOP and MacMon, I’ve found the Neural Engine usage depends on the model, power mode and source resolution. There are cases when the Neural Engine isn’t used at all - e.g. Rhea in standard power mode. However, bizarrely, Rhea is FASTER in Low Power mode on my M4 Pro as it uses the Neural Engine instead with minimal GPU.
Regards
Andy
I only do standard DVD Definition Upscale. No 4K or above. So DVD to HD 1080p is what I use it for. I did what you posted. I set it to 10% and the GPU history graph is FULL TILT and CPU graph is low. When I set it to 100% the GPU graph shows Med and CPU still shows Low. But the FPS is 16fps when 10% and 24fps when 100%. So on my particular machine set up. Seems that 100% is more efficient and faster for my particular workflow.
Update: Also the Fans are full blast when set to 10%. When set to 100% the fans are low and sometimes so low I can’t hear it and I think it stopped. Until I wake up the screen and it shows still processing.
Proteus Default Preset.
DVD Movie to FHD.
ProRes Standard Codec
iMac Hardware Overview:
Model Name: iMac
Model Identifier: Mac15,5
Model Number: Z19E000ZFLL/A
Chip: Apple M3
Total Number of Cores: 8 (4 performance and 4 efficiency)
Total Number of Cores GPU: 10
Total Number of Cores Neural Engine: 16
Memory: 24 GB
System Firmware Version: 11881.61.3
OS Loader Version: 11881.61.3
Mac OS: 15.2 (24C101)