Hi! Many thanks for your analysis. Would you mind to update it for M2 (Pro) and Darktable (allegedly) having gone native on M* chips? As I'm considering buying a new machine, I would really appreciate any advice. I've been using Linux for past 20+ years and love it, but I miss some apps (Logos Bible Study, Affinity Photo, etc.). I don't like mac (after playing with it for a while) as it seems clumsy and command-line utilities I love so much seems to be much worse but I'm willing to give it a try---if the Darktable works ok with it including the expensive operations like Diffuse-and-sharpen module. (I mean for my home stuff. For my real work, Mac is no way and I will stick to Ubuntu.). Many thanks for any hint or help you could provide. Best wishes, Michal
Thanks for watching and your feedback! I'm not sure if I'll have an M2 Pro handy in the near-term, but I'm also curious about how the recent native builds perform against the Intel builds. I've been continuing to run Darktable on an Intel-based Mac because of the eGPUs and (annoyingly) software RAID drivers I installed once upon a time.
It's been four months, sorry it took so long. I did run a comparison of darktable on an M1 Max (64GiB RAM; not an M2) both compiled for ARM and ran through Rosetta. This wasn't as thorough as the original tests; my theory was that after an initial hit to precompile/translate the X86 instructions to native code, the Rosetta performance wouldn't be that different over the long run (under the theory that Darktable and its libraries generally wouldn't be optimized for Apple Silicon). Given that hypothesis, I was a little surprised to see a 37% speed-up (specifically, running darktable 4.6.1 on MacOS 14.1 (23B74)) the Intel code ran at 137% the time of the Apple Si binary; or the latter ran at 72.87% of the time of the Intel code). I'm curious if you've seen other numbers, and if they're in-line with that, since I didn't go deeper (e.g. look at the difference in individual module performance).
Thanks for watching! Apple hasn't updated their Mac Pro (desktop/workstation) line, which was what I was originally expecting to get-but they did release the Macbook Pros with the M1 Pro and M1 Max variants of the M1-I ended-up getting one of the latter rather than waiting for an updated Mac Pro (and principally for video workflow, where it made more sense than a GNU/Linux based system for a variety of reasons). I'm finishing-up analysis of the benchmarks-I rushed the first M1 comparison video, and am aiming to do a more comprehensive job on the next round-so I'm hoping to get that published at the start of December (school holidays are going to effectively knock me out for the next week). The short analysis: despite all of the "it's amazing" articles that have come-out over the last couple weeks-I'm still amazed by it, but it (M1 Max) also seems to fit right where I expected. In most of the workloads I've tested, it comes in at about 2x the performance of the 2019 MBP and half the performance of a maxed-out Mac Pro (based on other's benchmarks) for tasks like Davinci Resolve. I'll go into more detail about where it seems to excel compared to the 2019 16" Macbook Pro, and where it's more nuanced in that video.
@@NottaPro I expected that you bought one of new Mac pro with m1. I plan to buy 14" pro with 32gb (my the most demanding target is photo post production) . Eager to watch your detailed comparison (based on darktable experience) when it will be ready. PS: I like to speak with people who pay attention to the details - you one of them m. I glad to became your subscriber 🙂
You won't be disappointed with the 14" MBP's performance in darktable, and 32GiB ought to be more than enough if that's your main use. It stays quiet which, is a bonus, although I did get it up to 109ºF at the chasis-but that wasn't exactly standard usage. Thanks for subscribing!
@@NottaPro Would you test DT with release 3.8 on the M1 PRO / M1 MAX? I think lightroom recently made some impressive improvements running native on these machines. Not sure how much penalty DT gets by running Rosetta. But I'm also about to buy a 14" M1 PRO with 16BG and I love the fact DT now supports iOS. Thx already for your M1 analysis so far. Really much appreciated!
Already done! My most recent video on this channel is about the M1 Max (th-cam.com/video/nZKwVj57FJA/w-d-xo.html), and while my focus is more on Davinci/video-it does reprise the DT tests, and goes even deeper. Rosetta does have a penalty, but it seems to me most apparent on start-up (or as it sees code for the first time), which means the darktable-cli benchmark performs a *lot* worse that you'd expect during interactive editing. In the end, I decided to keep my 2019 Intel Core i9 MacBook for darktable, and use the 14" M1 Max for Davinci Resolve-but unless you're adding eGPUs to the Intel MBP, or doing a lot of scripted (darktable-cli) sessions, the M1 Max will be faster (or at least not slower) while being quieter and lasting longer. I also posted a related/follow-up video addressing what seems like silly reasons to put-off buying a MBP as well as why I'm not finding that newer eGPUs make sense-for me anyway-as the cost to performance increase isn't there for my workload: I'm not seeing dt able to push enough to the graphics card on either laptop to really make use of a significantly better card. Re: DT on iOS-I hadn't seen that? I've seen a competing product "darkroom" which is confusingly generic-I'd be very surprised to see darktable on iOS given its Mac port is barely supported (single dev AFAIK) and its core isn't very Mac-ish which would make a port a fairly significant rewrite. I'd love it if I were wrong. Thanks for watching!
Hi, I don't speak very well English (I'm from Switzerland), that's why I didn't understand every thing you said in the video. You go really deep with the explanation. I just have simply question. I work (not as a professional) exclusively with Darktable. It's for me the most important software. I need to change my PC and I wanted to know if Darktable run well on m1. Is it usable? Or it's not enough optimise? I don't mean it's the best computer ever but for 900$ does it run enough good to be usable?? Thanks you for the time you take to read this message with so much mistakes.
Short answer: yes, it's definitely usable. If you're choosing between an Intel CPU with integrated graphics, and AMD CPU with integrated graphics, or the M1 (which only has integrated graphics); they're all going to be fairly close on performance. Longer answer: "It depends", as usual: darktable itself will run fine, but if you're working on high-resolution images you would do better with something that had a discrete graphics card (i.e. an AMD Radeon or nVidia Geforce with 8GB+ RAM, depending on your specific situation). Last I checked, the required libraries to build darktable for M1 haven't all been ported, which means you'll need to run darktable through Apple's Rosetta 2 layer; but I don't know how big of a difference native libraries would be, since it's unlikely darktable will ever be ported to the native Metal graphics API. The M1, running darktable through Rosetta, performs about on-par with the current 16" Macbook Pro (2019, 2.4GHz 8-core Intel Core i9) when the 16" is running on-CPU (i.e. without opencl/an AMD Radeon Pro graphics card); but the M1 will run cooler and quieter while doing so. The Core i9 will probably edge it out, for that matter. You don't mention what your current PC is: if you're currently running MacOS or Windows, and you don't have a discrete graphics card or eGPU; the M1 will probably be fine or better for you (the Apple silicon [M1] doesn't work with eGPUs). If you're running a GNU/Linux PC, I might be more hesitant. I'd probably go with a Ryzen-based machine with the option to add a discrete GPU as prices become more reasonable. In general... I'd try to avoid buying a computer in this market, component shortages are crazy, but if you need to, consider potential upgrade paths and what you might want to defer; an M1 is effectively non-upgrade-able except for external storage. There's also a risk that a MacOS upgrade will negatively affect darktable in some way, which only has a single volunteer maintaining the MacOS port (compared to several volunteers for Linux).
Im very interested in everything youre saying, but have a hard job hearing it because its dense, technical content but you’re speaking too quickly. A little slower next time, thanks!
Hi! Many thanks for your analysis. Would you mind to update it for M2 (Pro) and Darktable (allegedly) having gone native on M* chips? As I'm considering buying a new machine, I would really appreciate any advice. I've been using Linux for past 20+ years and love it, but I miss some apps (Logos Bible Study, Affinity Photo, etc.). I don't like mac (after playing with it for a while) as it seems clumsy and command-line utilities I love so much seems to be much worse but I'm willing to give it a try---if the Darktable works ok with it including the expensive operations like Diffuse-and-sharpen module. (I mean for my home stuff. For my real work, Mac is no way and I will stick to Ubuntu.). Many thanks for any hint or help you could provide. Best wishes, Michal
Thanks for watching and your feedback!
I'm not sure if I'll have an M2 Pro handy in the near-term, but I'm also curious about how the recent native builds perform against the Intel builds. I've been continuing to run Darktable on an Intel-based Mac because of the eGPUs and (annoyingly) software RAID drivers I installed once upon a time.
It's been four months, sorry it took so long. I did run a comparison of darktable on an M1 Max (64GiB RAM; not an M2) both compiled for ARM and ran through Rosetta. This wasn't as thorough as the original tests; my theory was that after an initial hit to precompile/translate the X86 instructions to native code, the Rosetta performance wouldn't be that different over the long run (under the theory that Darktable and its libraries generally wouldn't be optimized for Apple Silicon).
Given that hypothesis, I was a little surprised to see a 37% speed-up (specifically, running darktable 4.6.1 on MacOS 14.1 (23B74)) the Intel code ran at 137% the time of the Apple Si binary; or the latter ran at 72.87% of the time of the Intel code).
I'm curious if you've seen other numbers, and if they're in-line with that, since I didn't go deeper (e.g. look at the difference in individual module performance).
amazing detailed analysis. thank you for that!
did you finally buy m1 pro? :)
Thanks for watching!
Apple hasn't updated their Mac Pro (desktop/workstation) line, which was what I was originally expecting to get-but they did release the Macbook Pros with the M1 Pro and M1 Max variants of the M1-I ended-up getting one of the latter rather than waiting for an updated Mac Pro (and principally for video workflow, where it made more sense than a GNU/Linux based system for a variety of reasons). I'm finishing-up analysis of the benchmarks-I rushed the first M1 comparison video, and am aiming to do a more comprehensive job on the next round-so I'm hoping to get that published at the start of December (school holidays are going to effectively knock me out for the next week).
The short analysis: despite all of the "it's amazing" articles that have come-out over the last couple weeks-I'm still amazed by it, but it (M1 Max) also seems to fit right where I expected. In most of the workloads I've tested, it comes in at about 2x the performance of the 2019 MBP and half the performance of a maxed-out Mac Pro (based on other's benchmarks) for tasks like Davinci Resolve. I'll go into more detail about where it seems to excel compared to the 2019 16" Macbook Pro, and where it's more nuanced in that video.
@@NottaPro
I expected that you bought one of new Mac pro with m1.
I plan to buy 14" pro with 32gb (my the most demanding target is photo post production) .
Eager to watch your detailed comparison (based on darktable experience) when it will be ready.
PS: I like to speak with people who pay attention to the details - you one of them m. I glad to became your subscriber 🙂
You won't be disappointed with the 14" MBP's performance in darktable, and 32GiB ought to be more than enough if that's your main use. It stays quiet which, is a bonus, although I did get it up to 109ºF at the chasis-but that wasn't exactly standard usage. Thanks for subscribing!
@@NottaPro Would you test DT with release 3.8 on the M1 PRO / M1 MAX? I think lightroom recently made some impressive improvements running native on these machines. Not sure how much penalty DT gets by running Rosetta. But I'm also about to buy a 14" M1 PRO with 16BG and I love the fact DT now supports iOS.
Thx already for your M1 analysis so far. Really much appreciated!
Already done! My most recent video on this channel is about the M1 Max (th-cam.com/video/nZKwVj57FJA/w-d-xo.html), and while my focus is more on Davinci/video-it does reprise the DT tests, and goes even deeper. Rosetta does have a penalty, but it seems to me most apparent on start-up (or as it sees code for the first time), which means the darktable-cli benchmark performs a *lot* worse that you'd expect during interactive editing.
In the end, I decided to keep my 2019 Intel Core i9 MacBook for darktable, and use the 14" M1 Max for Davinci Resolve-but unless you're adding eGPUs to the Intel MBP, or doing a lot of scripted (darktable-cli) sessions, the M1 Max will be faster (or at least not slower) while being quieter and lasting longer. I also posted a related/follow-up video addressing what seems like silly reasons to put-off buying a MBP as well as why I'm not finding that newer eGPUs make sense-for me anyway-as the cost to performance increase isn't there for my workload: I'm not seeing dt able to push enough to the graphics card on either laptop to really make use of a significantly better card.
Re: DT on iOS-I hadn't seen that? I've seen a competing product "darkroom" which is confusingly generic-I'd be very surprised to see darktable on iOS given its Mac port is barely supported (single dev AFAIK) and its core isn't very Mac-ish which would make a port a fairly significant rewrite. I'd love it if I were wrong.
Thanks for watching!
nice video really good
Thanks for watching!
Hi,
I don't speak very well English (I'm from Switzerland), that's why I didn't understand every thing you said in the video. You go really deep with the explanation. I just have simply question. I work (not as a professional) exclusively with Darktable. It's for me the most important software. I need to change my PC and I wanted to know if Darktable run well on m1. Is it usable? Or it's not enough optimise? I don't mean it's the best computer ever but for 900$ does it run enough good to be usable??
Thanks you for the time you take to read this message with so much mistakes.
Short answer: yes, it's definitely usable. If you're choosing between an Intel CPU with integrated graphics, and AMD CPU with integrated graphics, or the M1 (which only has integrated graphics); they're all going to be fairly close on performance.
Longer answer: "It depends", as usual: darktable itself will run fine, but if you're working on high-resolution images you would do better with something that had a discrete graphics card (i.e. an AMD Radeon or nVidia Geforce with 8GB+ RAM, depending on your specific situation).
Last I checked, the required libraries to build darktable for M1 haven't all been ported, which means you'll need to run darktable through Apple's Rosetta 2 layer; but I don't know how big of a difference native libraries would be, since it's unlikely darktable will ever be ported to the native Metal graphics API.
The M1, running darktable through Rosetta, performs about on-par with the current 16" Macbook Pro (2019, 2.4GHz 8-core Intel Core i9) when the 16" is running on-CPU (i.e. without opencl/an AMD Radeon Pro graphics card); but the M1 will run cooler and quieter while doing so. The Core i9 will probably edge it out, for that matter.
You don't mention what your current PC is: if you're currently running MacOS or Windows, and you don't have a discrete graphics card or eGPU; the M1 will probably be fine or better for you (the Apple silicon [M1] doesn't work with eGPUs). If you're running a GNU/Linux PC, I might be more hesitant. I'd probably go with a Ryzen-based machine with the option to add a discrete GPU as prices become more reasonable.
In general... I'd try to avoid buying a computer in this market, component shortages are crazy, but if you need to, consider potential upgrade paths and what you might want to defer; an M1 is effectively non-upgrade-able except for external storage. There's also a risk that a MacOS upgrade will negatively affect darktable in some way, which only has a single volunteer maintaining the MacOS port (compared to several volunteers for Linux).
Im very interested in everything youre saying, but have a hard job hearing it because its dense, technical content but you’re speaking too quickly. A little slower next time, thanks!
Thanks for watching, and thanks for the feedback.