I currently have a 2018 MacBook Pro (i9, 32GB RAM) running Monterey. My wife recently suggested that I might want to upgrade to a new M1 MacBook Pro since they have significantly higher benchmark figures, but I am not sure how much the multi-core benchmark figures really apply to image processing.
When I bought the MBP I purposefully bought a model with high specs, thinking that it would dissuade me from wanting to upgrade quickly, and that has pretty much worked out. But I am curious about how much the multi-core benchmark figures actually translate into higher processing speeds with software like PL5.
When I process raw images with PL5 it generally takes a little less than 4 seconds to generate jpgs, with my latest test processing 27 raw images in 100 seconds. My MBP multi-core benchmark figure is about 4700 (and is currently set to process 5 images simultaneously) with the new M1 MBP multi-core benchmark figure being about 12,200, or about 2.5 times higher. Should I really expect images to process in about 40% of the time on an M1 machine? Or do the benchmark figures not really directly relate to images processing times? And do the main chip multi-core figures relate to the processing speeds? Or is that more the speed of the graphics card chips?
@MikeFromMesa, times don’t mean much if we don’t know what files were processed and what the settings were…but some figures were posted.
I recently developed 60 test images with PureRaw and got them developed, with DeepPrime, in about 6-8 seconds per image.
Comments: While I tested all this on a 2020 M1 MacBook Air (which is not the most powerful Apple Silicon Mac around), I’d not see an immediate need for you to buy a new MacBook. Developing images is something that I don’t watch either. I simply start (high number of) conversions and then do something else… One hint though: Set your Mac to not sleep, because it interrupts development on M1 Macs (open the terminal and enter “caffeinate” to keep tha Mac going. Afterwards, enter - to stop the caffeine…)
Thank you for the reply. And I should note that I don’t watch the development process either. Like you, I start it and go and do something else.
My current MBP responds immediately to editing changes with no noticeable visual lag, and 4 seconds seems like reasonably fast processing. Image processing is almost certainly the most intensive computational work that I do on my computer and I am not unhappy with the system I have, and hence my posting the question as compared to just going to the Apple Store and buying a new MBP. But I do wonder about how well the benchmark figures correspond to the actual image processing time.
Can’t comment against a machine that should be as powerful as yours however I’m sorely tempted to get an M1 machine shortly after recently picking up an M1 IPad Pro.
For my comparison it’s a 2019 MacBook Pro i5 8GB Ram (was for work not image processing just ended up being used for such) vs a 2021 iPad Pro 11 which is M1 and also 8GB.
Now I can only compare Lightroom CC but the iPad it’s fair to say destroys the MacBook.
Taking day the automask feature to select sky or subject. The M1 iPad has finished before the MacBook is halfway through the progress bar.
Obviously that’s a specific example but was one that caught me by how quicker the iPad was.
Whether you’d see such an improvement with PL5 I don’t know but in my case/machine I certainly think I will.
My several year old Dell with an i9 processor and 32gb ram and a mediocre gpu takes about 1 minute to process a 45 mp file from the Canon R5 (with many dxo adjustments) and Deep Prime. My $850 Macbook Air with the M1 chip processes 7 such files in 1 minute. So, seven times faster.
Benchmarks follow a predefined sequence of operations and attacking different subsystems in order to spit out a figure. Change one parameter or weighting and the result can be quite different. If one test results in a 7:1 speed difference, another test might produce 1:1.
The biggest benefit of Apple Silicon processors is the high performance / power consumption ratio, which should further increase with next generation chips.
We need to pay attention to the Apple Neural Engine in this discussion, which is reportedly the same in a 2020 M1 MacBook Air as it is in a 2021 MacBook Pro with either M1 Max or M1 Pro - and the neural engine is what PL uses for Deep Prime exporting. The M1 Max and M1 Pro will be faster than the M1 at many things, but not necessarily exporting from PL.
That is very interesting because my several year old MacBook Pro, also with an i9 chip, 32GB of RAM and a mediocre gpu also takes almost the exact same length of time to process an image with Deep Prime. And while I don’t feel the need to replace my MBP I have been thinking about replacing my 6 year old MBA, which would allow me to take it on trips rather the MBP.
Thank you for the information. When I look at the available MBA versions I don’t see one for $850 so can you tell me the specs of your MBA? And what year it is?
I looked on the Costco website and the MBA is not currently on sale, and the price break at Costco is only $50, which the same as what I would get on Amazon with one of those Amazon Prime cards (5% credit for future use). Still, a MBA is something that sounds like a good idea since my current one is about 6 years old and was never very fast. In fact it is so slow that when I ran the Dxo Pure Raw demo on some raw images it took something like 12 hours to complete the processing that my MBP did in about about 90 minutes.
That makes upgrading to an M1 MBA an appealing idea, but can you tell me how well it runs PL? I travel quite a bit in an RV but I am generally only taking photos in decent light, so I almost never need any of the Deep Prime processing. My MBP takes about 4 or 5 seconds to process an image without Deep Prime, and about 60-75 seconds to process one with Deep Prime, so it is reasonably fast for what it is - an older intel machine.
So how long does it take your MBA to process a raw image without Deep Prime? And with it? And how many images is the MBA set to process simultaneously? I am just wondering if it is fast enough to replace my current MBP.
Did a quick test with one image and 4 virtual copies set to different presets (I used the the standard presets 1-5) and got all output files in a total of
7 seconds on a iMac 2019 (8-core, 40 GB RAM, PL set to process 3 images simultaneously)
5 seconds on M1 MBA 2020 (8-core, 16 GB RAM, PL set to process 3 images simultaneously)
Processing 60 test images with PureRAW (another of DxO’s apps, which is like a no-frills extract of PhotoLab) and DeepPrime, I got all output files in a total of
8.7 minutes on the iMac
6.5 minutes in the MBA
Does this help? I don’t think so, because processing times depend on many things. Running a batch of heavily customized images on the iMac, I get the fans up to speed, while the MBA quietly reduces its speed if gets too hot, which is of no concern to me in real life.
If you use the same image as me (see link above), you can compare your figures to mine, although with a grain (or ton) of salt.
Actually, yes, it does help, because it gives me some indication of how the MBA compares with the iMac and it (the iMac) should be considerably faster than my MBP. Thank you very much for taking the time.
It may be that I can use a new M1 MBA as a main machine in place of my MBP. Since I use it in clamshell mode the smaller screen size does not matter. I use an external 32 inch Monitor hooked up to the MBP.
I do have a couple of questions.
> on a iMac 2019
Is.this an i7 or an i9 chip?
> 5 seconds on M1 MBA 2020 (8-core, 16 GB RAM, PL set to process 3 images simultaneously)
Does this mean 5 seconds/image?
> 6.5 minutes in the MBA
Similarly, does this mean 6.5 minutes/image? Or 6.5 minutes for all? If each image, that seems like a long time for a new M1 machine so I suppose it is the total time. Also, were you processing to jpgs? Or dng? I am not sure it makes much difference, but I processed to jpgs so I suppose it is a valid data point.
I did a similar test on my MBP and I seem to remember that it took less time than that, if that is a time per image figure, but then I did not document the time and I can not redo the test because I did not buy PureRAW. Perhaps I should try downloading it to see if I can run the test again.
I’ve rephrased my post above. Times are total times as reported by PL’s indicator, which is only showing full seconds, and the timestamps of the 60 output files respectively.
If you take the example file, you can replicate the first test with the virtual copies. No need to test PureRAW because you don’t have the same set of images to test with. I had added the 60 image test results because I had run that test with DeepPrime while the first test was without DP.
It is your PureRAW figures that are the most impressive compared to my MBP. 6.5 minutes for 60 images using DeepPRIME means about 6.5 seconds/image and my testing using DeepPRIME (I was able to download a later version so I do have a working trial version) showed me that running it using my raw images (Olympus M5 Mk3) took about 25 seconds each, so your MBA is much faster.
Interestingly enough the processing speed is heavily dependent upon image size. My Olympus takes relatively small (20MP) images but has a High Def setting that produces 80MP raw images by sensor shifting. I don’t use it much because the photos require that the camera not move at all for the time it takes to finish the sensor shifting and even a decent tripod does not do a good enough job. The only success I have had is when I hold the camera down on a rigid surface to make sure it does not move. Anyway doing PureRAW processing of those 80MP images took about 75 seconds, or about 3 times as long as the regular raw images.
Your times with the MBA are so impressive I am going to look into upgrading to one of the smaller M1 machines, so thank you very much for your help.
Get as much RAM as possible before anything else. Nevertheless, saving a few bucks now doesn’t pay in the long run, unless all your stuff is on a NAS… RAM is shared between CPU and GPU and if you’re getting low in RAM, it kind of hits twice.
(Robert Silber -- Canon | Mac Studio M2 Max)
So true! I learned long ago to spend as much as I can buying into the tech curve, which is always a moving target. Even with a NAS, I tend to favor a 1TB drive, and all the RAM possible. I do miss the day that the user could easily upgrade RAM and drives on a Mac.
When I bought my current MBP I opted for 32GB of RAM and the fastest chip I could get. Both of those decisions have worked well for me and as a general rule I have been very pleased with how fast the machine is.
However the M1 chip and the associated 16 core neural network seems to have changed all that. The figures I am seeing for a simple 8GB RAM setup are far faster than those for my MBP, so I am torn. My current machine is fast enough for my needs, but a faster machine is always nice. I could justify spending enough for a MBA, even the higher priced one, but it is hard for me to justify spending the price for a full out spec-ed MBP since it is 3 times the price of the MBA and my current machine is more than adequate for my current needs.
I just wonder if the smaller MBA with the newer hardware would be enough to replace my current MBP. As always, upgrading hardware becomes a choice between desire and cost.
My current memory size is 512GB and that is more than I need because I have a 1TB usb SSD and I keep all of my photo stuff on there. Interestingly enough the PureRAW test I did using my external SSD took less time than an identical test using my onboard memory (which is also solid state). Don’t know why, but the internal SSD was about 5% slower than the external SSD. Go figure.
Well, @MikeFromMesa, it’s your cash. Buy small if it’s for « fun » only, or don’t buy and keep your current machine running for another 1-2 years. If you plan to replace your Book now, the story might be different.
I have just installed a minimal Mac mini M1 (8Gb/256Gb). Average export times for 20MB M4/3 files is 6 seconds. I hardly have time to get organized to do something else while exporting but don’t think there is any slowdown of concurrent activity.
It’s highly recommended and I am reorganizing my workflow around it.