performance of darktable 4.2 on macbook air m1 base model

Hi, is anyone running darktable 4.2 on a macbook air m1 with only 8gb of RAM? The reason I’m asking is that I want a light laptop with good battery life and found a slightly used macbook air base model for a really good price, but I’m not sure if the 8gb version would be enough to give me a good performance.
Thanks!

8gb is the minimum recommended, but it really depends on what you are doing.

I usually don’t edit large batches of images, but I do intend to use resource-intensive modules such as diffuse or sharpen if it is usable on it.

For D&S, you want processing power, ideally dedicated GPU.

1 Like

If you’re getting a new computer to run darktable, don’t expect to have the best experience using macOS. I switched to Linux to get the most out of darktable. I have a well specced M1 Max machine at work (64 GB RAM) running macOS, but I can’t say darktable feels faster than on my soon 7 year old Acer laptop (with a GTX 970M GPU) running Manjaro.

Asahi Linux is making a lot of progress, but it still has some way to go.

I’m not getting it for darktable, because I do most of my edits on a linux desktop. I just need a small and light laptop that could run darktable acceptably well, but it does not need to be a “studio workhorse”. If the M1 air can run it well enough with 8gb, it would be a nice option for me (where I live the price hike for the 16gb version is outrageous); if not (like in “if you turn on diffuse or sharpen you will regret for life”), then I would look at something else.

1 Like

Using the images you have and in your current system, test how much memory the system is using. It should not take too much more memory than that.

From my experience, 8 GB w/o a dedicated GPU are hardly enough for an acceptable workflow. I was doing this for the past 10 years, and it was OK up to ~20 MP images with less demanding modules, but even for those, it was only acceptable. Running other software in parallel, such as a browser with a couple of open tabs, meant to be a huge issue, and I had to restart darktable every ~30 minutes to free the memory, as even in lighttable, the memory use accumulated over time.

I think it would be good if you could test this for a day or two, to see if it is OK for you.

1 Like

Thanks, I think I will try and find soneone with an m1 air that can help me to test it. Thanks for the advice.

As I understand it, the M1 Macs are borderline terrible with DT. No GPU support, and lots of visual bugs both during editing and exporting.

you need to enable opencl yourself - the calculations if cpu performance outperforms gpu aren’t very reliable for m1 - even well spec’d M1 models won’t enable GPU by default

@MStraeten , and do you have any first hand experience with performance on a M1 with only 8GB of RAM? If not, is there any particular macbook you would recommend that is also able to run dt decently?

I’ve also been on the lookout for a new laptop and weighing an M1 Air as an option. I benchmarked an 8GB M1 Air with 7 core GPU along with a X1 Carbon (with i5-6300u) and desktop with an (archaic) GTX 1050 2GB both running Windows 10 and darktable 4.2.0.

Here are my findings using the arecibo benchmark from here.

X1 Carbon CPU only: 36.7s
X1 Carbon OpenCL: 25.44s
M1 Air CPU only: 9.48s
M1 Air OpenCL: 2.94s
GTX 1050 2GB: 10.23s

Anecdotally darktable performance on the M1 Air was actually pretty good when it worked. As MStraeten mentioned support for OpenCL is flaky (I had to restart the benchmark 3 times to actually get the kernels to compile) and darktable has a few significant MacOS-specific bugs at the moment. OpenCL has also been deprecated by MacOS for a couple years now.

All that said the M1 Air is still frustratingly good. There doesn’t seem to be another laptop with comparably good display, battery life, weight, build quality, performance, and display at a similar price point on the market. If anyone has any ideas or suggestions though I’m all ears :slight_smile:

3 Likes

I think you should test with a raw file/xmp that mimics the edits you normally do. Sarunas xmp is a great way to see how dt / hardware progresses over time, but it is now a bit dated.

1 Like

Yeah definitely a good idea, I’m curious how the M1 Air would handle Diffuse & Sharpen.

1 Like

It works. I’m running darktable on mba m1 worh 8gb. Don’t upgrade to ventura. Is useless on ventura. Gtk issue. But is working.

Asahi linux project for M1 does not yet have GPU support let alone an opencl running …
prepare to be very patient to get the most out of Dt on an M1

nevermind that, sorry I’m outdated on this knowledge and need to catchup, sorry for the noise !

And again I’m not speaking of other x86 optimizations (SSExxx and stuff) that the M1 arch cast you out of (correct me if I’m wrong !)

Do the amount of ram used depedent on the raw size (res*bit depth) beign open in darkroom view ? or is it irrelevant ?

1 Like

More megapixels = more ram.

1 Like

Well, I had the opportunity to find a M1 air with 16GB of RAM for a good price, and decided to try it. I have been using it with darktable and photoshop for a month, and I’m happy to report that dt 4.2 ARM version works great. I have been editing 24 and 42MP images with it. I manually enabled open cl, and it works fine. It is quite fast overall, even when using resource intensive modules. Export times are also totally fine. The only strange thing that happened so far it that it crashed about 4 or 5 times during editing. Coincidence or not, it did it when I was using the color zones module. I thought about opening an issue on GitHub, but it doesn’t seem to be something easy to reproduce. I have used color zones many other times without any issues. Should I open an issue on GitHub anyway, so maybe someone can investigate it?