Headless Minis

I probably have to replace my MacBook Pro in the very near future and I was looking at the M1 MacMini as one possible replacement. It occurred to me I might be able to set up the Mini as a headless server that handled household chores such as TimeMachine, media server, and so on.

I also wondered if it would be possible to run PL4 off of the Mini by connecting to it from a (new) MacBook Air using Apple’s Remote Desktop or similar.

The advantages of this are that PL4 would be able to take advantage of the better processor speed on the headless Mini, and the new Air would not need 16GB of memory or 1 TB of disk space. The Air could indeed be much skinnier, as it was intended, and I wouldn’t have to buy a new MBP.

I’m pretty sure from research that this architecture is supported. PL4 would run off the Mini and I would see the screen from my remote desktop. I’ve read a bit about how remote desktop has a bit of lag, and this might become a factor while editing images in PL4.

Has anyone else, considered or successfully built such an arrangement?

Thanks in advance.

According to geekbench, the M1 Mini and MBA have very similar benchmark values. I don’t expect that your planned setup will do much more than add complexity.

I’ve run an Intel Mac Mini as a headless server a while ago. Setup required a keyboard and a monitor, once the setup was done, things worked as expected. I did not do any high load stuff though.

That’s a great resource, thank you. My 2015 MBP is in the middle of the performance list, but the new iron tests very good. Yea I am probably overthinking (and over-engineering) it by adding the server. I’ve always wanted a file server and a true NAS set up is probably cheaper than this solution. The possibility of moving apps off to an app server though is enticing.

At this point, I’ll probably just wait until the M1/M1X/M2 MacBook Pro’s appear in the Fall.

Waiting can save money and if your current hardware still works, there is no need to rush it.

I have a 2020 M1 MBA (16GB/1TB) and it runs smoothly, in many cases, it feels snappier than my 8-core 27in iMac (40GB/2TB). Exporting images takes longer though. I’ve posted some figures somewhere in here…

I’m a fan of the no-fan design of the MBA. And although the iMac is mostly silent, its fan emits an almost subliminal noise that creates a feeling of weightlessness when the iMac is powered down.

Could you give us an example of RAW file and time processing on M1 MBA, preferably with deep prime ?
I would compare with my old MBP mid 2012.
Time processing could be as much as 15 minutes with Deep Prime…

Yes ALL of the current Apple Silicon Macs use the same M1 chip with only very minor differences. However that does not negate your desire to “slim down” a MacBook Air.

I would, however, caution against operating something as interactive as PL4 over a screen share, no matter what technology you use. I would expect the latency to be really annoying.

The approach I have taken is to put all my RAW photos on a small external SSD. I have a 1TB Samsung T5 which is tiny and weighs basically nothing. Apart from the slight fuss of plugging it in (to my M1 MacBook Pro) and having it dangle there while in use, it’s a great solution and if I outgrow it before the MBP is due for replacement, I can simply buy a new SSD.

Smart way, no noise and much smaller power consumption than a server or NAS. I’ve also used a T5 as external boot drive for the M1 MBA as well as a bootable clone for the iMac. External hard disks should work too, if you can bare the speed drop after being spoilt by SSD speeds.

You could post a link to a raw file from the sample galleries of dpreview.com or any other source that permits downloads - check copyright notices. We could use such a file as reference for simple benchmark measurements. If DxO provided sample files, that would be perfect.

Example links:

Procedure (must be followed for reproducible measurements)

  • Process both images to DNG
  • Parallel processing set to 3
  • Provided presets or settings, no sliders moved

First column: iMac 27in, 2019, 3.6 GHz 8core i9, 40GB, Radeon Pro 580X 8GB and DPL4 on Mojave.
Second column: M1 MacBook Air, 8+8core, 16GB and DPL4 on Big Sur

  • 13 / 14 s @ No Correction
  • 33 / 81 s @ NoCorrection plus DeepPrime
  • 17 / 21 s @ DxO Standard
  • 35 / 81 s @ DxO Standard set to DeepPrime
    — / 84 s with parallel processing set to 2
    — / 90 s with parallel processing set to 1

Settings play an important role in how much time is used to process the images. Increasing the value for parallel processing does not boost performance that much. I find the sweet spot to be somewhere around 3. Going beyond 3 does not do much good on my systems and batch size.

If we want to play that game seriously, we’d need to define how to set DPL first and also how to run the tests. I’d propose to create a separate thread - including a table hosted in the internet, so that results can be compared more easily. The table should also contain columns for tech specs.

@CaptainPO, if DxO should be interested in such user system benchmarks, could DxO then provide a few resources like test images, the table mentioned above and some brain power on how to set up systems and procedures for comparability?

Let us make the bench ourselves !
Results with example files :
NEF : 9’
ARW : 8’
MBP 2012 2,3 GHz 16 Go
Just activated DeepPrime
Who else can make a test ?
Specially with a recent computer M1 or iMac.

My MacBook Pro 2012 is the last one that had a HD and a DVD drive and where you could change them for HD or SSD.
I have a 1To SSD and a 4 To SSD inside.
Very practical indeed, but in next one, I will be obliged to use external drive.
My pictures weight 3 To and I manage them grouping ra, jpeg originel, jpeg developed, .dop and .json files together and I keep file names synchronized.
So I would not like to put raw apart form jpeg.

Well, in my first test, I did not respect all that procedure but with deep prime, process time is multiplied by n, n>>1 (*)
The results with MBP 2012 2,3 GHz 16 Go and your procedure

  • 26s @ No Correction → DNG (10 s → JPEG)
  • 34s @ DxO Standard → DNG
  • 14’56s @ NoCorrection plus DEEPPRIME → DNG
  • 15’27s @ DxO Standard plus DeepPrime → JPEG
  • 16’51s @ DxO Standard plus DeepPrime → DNG
    Well, we don’t play in the same school yard ; you are in the upper school!
    Another interesting lesson : you iMac is not ridiculized by MacBook Air, reassuring but DxO is not optimized for ARM.
    That comfort me to wait for big iMac replacement or super Mini, perhaps end of the year.

(*) n~30 for me ; I wonder if my graphics card is used ; the setting is automatic or CPU only, no choice for graphic, so I don’t no the behavior.

Your figures for DeepPrime suggest that your GPU does not contribute much, if used at all.
Compared to your figures, the GPU in the M1 MBA seems to be fairly helpful. DxO has not published anything yet that is optimised for Apple Silicon SOCs, but the apps run well on the new chips and Rosetta seems to be up to the task too.

Meanwhile, I tested develop times depending on how parallel processing is set and on the number of pictures processed. Here are the figures for the total processing times for 1, 2, 3 etc. parallel images and the No Correction preset.

05 images: 13, 08, 07, 08, 09, 07, 07, 07, 08, 07, 07, 07
15 images: 41, 24, 21, 19, 19, 22, 22, 22, 22, 22, 19, 25

As we can see, times don’t change much above a setting of 3 and I’d not venture into saying which setting above 3 gives the best results in all possible cases. Default on my iMac was a setting of 6.
Maybe the sweet spot is at 4, taking into account this post.

I don’t know if GPU is used ; there is an option in Settings but it’s greyed.

Well now I can compare with MBP 16"
21"" for the two pictures and Deep Prime ; good !

If I am reading those figures correctly the M1 MBA is much, much slower in processing raw images with Deep Prime than your iMac.

I have a relatively well-spec-ed 2018 MacBook Pro (32GB RAM, 8 core i9) which seems comparable to your iMac (except for the graphics card) and that appears to mean that even an M1 MBA would be slower than my MBP. I guess I should keep my money in the bank instead of sending it to Apple.

These figures were from April, which is a while ago…
Maybe, I’ll retest later…

Using the two linked images. M1 MacBook Pro, 8+8 core +16 core neural engine, 16GB and PL5 on Monterey using the neural engine.

  • 10 s @ No Correction
  • 21 s @ No Correction plus DeepPRIME
  • 20 s @ DxO Standard (13 s to full sized JPEG)
  • 23 s @ DxO Standard set to DeepPRIME (20 s to full sized JPEG)

I never shied away from using DeepPRIME in PL4, simply leaving it to churn away on its own if I had a lot of photos. Now I have even less reason to avoid it.