M1 Max and M1 Ultra Speed Improvements? Using the Mac Studio

Can anyone provide rendering speeds with the M1 Max and M1 Ultra chips found in the new Mac Studio computer?

I have a good idea on the speed for the M1 chip – but curious if the new Mac Studio computer out-performs the Mac Mini M1? Enough to validate the cost of $1995 vs $750 for the computer?

2 Likes

Mac Studio Ultra owner here. For me, using Proteus model & upscaling old video to 1080 HD 0.07 seconds per frame. I’m happy with that :slight_smile:

Hi - thanks for sharing. Can you tell me what was the resolution and specs of your original file and what settings for the Proteus model did you use?

I’d like to do a comparison to a similar job on a Mac Mini M1 to see if there is a dramatic performance difference.

Proteus Fine Tune - 720x576 upscaled to 1920x1080

Well - I did a test on my Mac Mini M1 with 16GB – and it renders at .11 seconds / per frame using your same settings.

So it appears that there’s not a dramatic performance increase when using a $4k Mac Studio with Ultra.

That’s a bummer, but it appears the best route for large projects is to chop the clip up and render on multiple M1 computers (@ $800 each). With 4 units running at the same time, you could basically render at .0275 seconds / per frame. Costing about $3200.

I just need a good automation process to do it.

1 Like

All of the M1 models have the same 16 core machine learning processor except the studio ultra which is 32 cores. I compared the M1 pro to M1 max and there was virtually no difference for me.

I agree getting multiple cheap mac minis is the cheapest way to process things at the moment. You can just divide a movie by frames or on a keyframe cut and recombine.

Iphone 13 processor had 30% more powerful machine learning than iphone 12 so if that is used as the foundation for M2 then M2 should give a decent boost.

The codebase as it stands today isn’t optimised yet to take advantage of Max & Ultra. If you absolutely positively need to do the fastest you can now, then yes, buy a few M1 mini’s and cut and recombine them like rapsfan has also said. But the beta of version 3 is coming on in leaps and bounds but that’s all I’m prepared/allowed to say on a non beta forum, I’m sure you’ll appreciate that!

This is where the new M2 chip will have the advantage over the current M1 SOC’s. M2 has a 16 core Neural engine capable of 15.8 Trillion operations per second vs 11 trillion per second on the M1 Chips.

Can anyone confirm that the next version will fully support a machine with the M2 chip?

For me, having a bunch of MacBooks, just because of the M2 chip is silly, unless it offers more than double the performance compared to a Mac mini M1 computer for $600 to $800.

I only need one screen for 12 Mac minis LOL.

Thanks for any details.

I did read that the M2 Neural Engine is 40% faster.

https://www.apple.com/newsroom/2022/06/apple-unveils-m2-with-breakthrough-performance-and-capabilities/

Do we know for certain that VEAI uses Apple M1/M2 hardware neural engine? This is a cross-platform (windows-Mac) product and could potentially be built on AI software not optimized for, or not even compatible with, Apple’s hardware neural engine.

As far as I know, yes.

Because it doesn’t get faster with the bigger chips as long as the neural engine doesn’t get bigger.

Some chips have the same neural engine although they actually have more cpu and gpu cores.

Please correct me if I’m wrong, but I’ve seen no one say that this software will run faster with the soon to come Apple M2 chip.

It is confirmed that a significant increase in speed was achieved with the M1 chip, yet only slight performance increases with the M1 Max and Ultra chips.

That make’s sense, thanks Thomas.

Digital foundry took a look at the M1 ultra, and VEAI was in its suite of tests.

Ultra was about 50% faster than the max according to them doing 11s 1080p to 4k60 (Artemis high)

M1 Max MBP (wasnt clear if 24 or 32 core model) @ 6min 1s

Mac Studio (64c) 4min 12s

Intel 9980/5500m MacBook Pro 13m 33s

Interesting discussion. Clearly, the M1 Neural engines do a lot of the heavy lifting with VEAI. The 50% performance increase on the Ultra Chip makes sense since it has 32 Neural Cores as opposed to 16 on all the other chips. I just bought the base 14" MBP with 8core CPU and 14-core GPU. Does this mean the performance of the 10 Core and 16-core M1Pro CPU will be almost the same?

On another note, M2 looks promising with its 40% Neural Engine performance boost. That could mean the Macbook Air with M2 could perform almost as good as a Mac Studio with M1 Ultra??

“The Neural Engine Cores, GPU cores, encoder cores in that order to optimize performance for Video Enhance AI. - I would recommend purchasing a machine with memory of 16GB or above.”

This was from an email response I received from Topaz labs recently where I asked what components of the Apple M1 SOC are of most importance when it came to using Video Enhance AI.