Help with M2 MB Air Vs M4 MB Air performance comparison insights?

That is strange. Was the M4 given time to finish all its "new computer/new OS " housekeeping (can be a day in some cases)?

Hard to tell from the numbers. In general, the first thing most long-serving contributors suspect on hearing "slow" are add-on apps and utilities. They can mask other issues. The only tool we have to help on this setting is a system config report. Fortunately there is a safe way to accomplish that.

If you want a data-driven evaluation in this setting where we can neither see nor touch your computer, please post an EtreCheck report. We can quickly and within the limitations of these forums help you determine what issues are at play without our playing a protracted game of "20 Questions" with you that could go on for days. EtreCheck Pro is available here:

https://etrecheck.com/index

The free version will do nicely for this purpose, although the app is worthy of our financial support.

We can see hard data about drive performance, software issues and interferences, and RAM usage. Etrecheck is the development of a long-serving and trusted ASC contributor. It is a reporting app, not a "fix-it” app, expressly for displaying information in these forums to help us help you remotely. It will not reveal any personal or secure information.

Please see this excellent user tip on how to post long text reports like EtreCheck's into a forum response:

How to use the Add Text Feature When Post… - Apple Community

Please post the entire report unedited. What seems insignificant to a new Etrecheck user can speak volumes to those of us who have reviewed thousands of those reports.

I recommend opos the M4 report, thenwe;sse if posting the M2 might add anything.

The importing and exporting more likely due to the SSD speeds still pretty much the same between the two computers. I not long ago bought an M4 MacBook Air to replace an M1 MacBook Air. Both have a 512GB SSD and I did a speed test using different apps and found that the speeds are not really that much different. So, with importing and exporting, most of the time is spent pushing files around so you may not see that great of a time difference as the processor is not a major portion of the process.

As for the denoise performance, if it is truly almost an hour for a single image, you must have some Lightroom settings not right (or Lightroom is just a real dog when it comes to noise reduction). I have to apps (don't mess with Adobe anymore), DxO PhotoLab and ON1 PhotoRAW, and their noise reduction on a 6774x4492, 24 bit image takes maybe at most 20 seconds on my M4 MacBook Air w/24GB RAM. BTW, with superior noise reduction results than Lightroom.

However, if you are referring to batch processing those 50 images you mentioned for noise reduction then that is a different story and again, pushing the files around is likely consuming most of the time so only moderate improvement would be seen.

Also, there is also the matter of how Lightroom actually does the noise reduction and if they are using the most efficient methods for Apple Silicon.

With all that said, the truth of the matter is that benchmarks are just a synthetic method of of comparing processors and real world use depends on so many factors (like app coding efficiency, computer resource utilization, etc.) that it is not surprising that real world operations may not see the 50% or so improvement in speed in real life and can sometimes see less than one might expect.

woodmeister50 wrote:

Also, there is also the matter of how Lightroom actually does the noise reduction and if they are using the most efficient methods for Apple Silicon.

Within a given generation of Apple Silicon chips (M1 - M4),

• You get more GPU cores as you go from a plain chip, to a Pro chip, to a Max chip
• You get the same number of Neural Engine cores (16) in all of those chips

Thus, if an operation can be accelerated either using the GPU, or using the Neural Engine, the Neural Engine may provide the greatest acceleration on a plain chip (like the ones in MacBook Airs). You might need to go up all the way to a Max chip before the GPU can outrun the Neural Engine on certain tasks.

However, I remember that some developers of photo processing programs ran into issues with the Neural Engine approach creating unwanted results, and so were going back to GPU-based acceleration (as the default or only choice).

I found these threads with a little searching. I'm not certain, but it looks like Adobe disabled Neural Engine use for Denoise on purpose.

Neural Engine bug: Adobe, DXO - Apple Community

Reddit – Lightroom Classic no longer using the Apple Silicon neural engine (and I'm... happy about it?)

FredMiranda.com – Adobe disables Neural Engine Support again

Adobe Community – P: Status of Apple Neural Engine Use?

That last thread includes posts by someone whom the Adobe forums identified as an Adobe employee.

njwight wrote:

Finder folder (333GB )copy to SSD: M2 9:01:00 M4: 8:44:20 (-0:16:40) 

This is going to be dependent on interface speed, drive speed, and Finder overhead more than on CPU speed.

You were running the same version of macOS on both laptops, so the Finder overhead should have been equal (barring system overhead like Spotlight indexing on a new machine).

Both Macs had a 1 TB SSD (not "hard drive"). M2 MacBook Airs with 256 GB SSDs had somewhat slow SSDs, because all 256 GB of storage was on a single flash chip, and there was no opportunity to use two chips at the same time. But with 1 TB SSDs, both of your machines should have had multiple flash chips, and their internal drives should have been very fast.

So this particular result might not indicate anything wrong.