My first forum post here - I hope I put it in the correct section, if not, feel free to move this thread and accept my humble apologies 🙂
I think I found a possible bug in the (as of right now) current Archicad 26 GER release for Apple silicon.
When comparing the ARM and Intel versions I only noticed Archicad 26 itself opening faster, while the rest seemed more or less the same (or just slightly faster - nothing like those gains described in the performance review).
However, when trying a quick test-rendering I noticed that it took just about twice as long to finish rendering - I know this was mentioned in the current issues but I wasn't aware that "marginally longer" meant almost twice as long - could just be a language barrier on my side though 😉 - anyway, I digress.
When checking my system usage while rendering (see the attached image) I noticed that Intel-AC26 uses every single available core to its full potential whereas ARM-AC26 uses only 5 cores to about 70% and a sixth for maybe 25%.
I strongly doubt this is intentional and think that there are still performance gains to be made if the current ARM release only uses half the available cores and of those used cores, none are even utilised fully.
ARM-AC26 also doesn't seem to use more than those 5 cores for any other task, though, due to limited testing (I got my hands on it just an hour ago), I cannot confidently say so.
Some additional information:
I'm using an M1 Max equipped Macbook Pro running MacOS 12.6.1.
Both test renderings were made using the same file with the same settings.
Thermal throttling wasn't an issue: temps under Intel didn't go above 83°C, and ARM didn't go above 70°C.
I should probably also mention that I use the education version (in case that makes a difference to my findings).
In the last couple of days, we received reports about the rendering slowness with Apple Silicon version of Archicad 26. With the help of the development, we are currently investigating the reason of it with high priority. I'll keep you updated about this matter when we know more about it.
As the rendering slowness doesn't happen in the Intel based version of Archicad 26 (not even on Apple Silicon computers), as a workaround, I suggest using that Archicad for rendering.
I just received the great news from the development, that the issue is located and can be fixed in the next update to Archicad 26 Apple Silicon version. We are planning to release it as a preview for the English and German speaking versions in February, then it becomes available for everyone a bit later.
Thank you for the patience!
[Edit: The issue has been registered as DEF-11704 and planned to be part of update 50xx.]
Does this mean that AC26 will use every core on M1 Pro/Max systems ?
If the core usage is fixed, there should be an *actual* improvement in speed on Pro/Max systems - since, as of now, the only actual improvement in my experience is still just the start-up time (which is back on AC25 levels after significantly worsening with AC26)
Right now, it seems AC just uses the 6 performance cores a base M1 has - which would be consistent with the findings I reported in my initial post.
The reason of this issue with the rendering was indeed the core usage: the Apple Silicon version of Archicad didn't use all of them, but the Intel did. This issue is going to be fixed in the next update, so thank you for reporting this with measurements, they helped us to localise the nature of the slowness.
On the other hand, the fix does not affect the other segments of Archicad. Core usage will not be changed anywhere else this time.
Thanks for understanding!