Yes I don’t thing so either, probably no this generation, but with the “more space” that 3nm gives them they could.
have there been many m1 ultra blender users in the wild to speak to how blender handles the 64 core gpu? from what i recall even the 32 core max starts to fall off in terms of performance scaling (although as a daily user its still vastly better than any laptop apple offered previously).
if i end up upgrading earlier than usual, it’ll be primarily for more ram, and any slight performance bumps would just be a bonus. all the speed in the world doesn’t help me if my system keeps crashing out of memory when trying to do lookdev on larger scenes!
1 Like
Don’t know, good question.
I agree on the more RAM definitely would get 64 GB next time. My M1 only has 32 GB.
Also recently I have been upgrading / changing my Macs way to often. Not like PCs which I used to keep for like 5 years (sometimes with a GPU change).
I had 3 Macs in 3 years 
4 Likes
normally i update my mac every 3 years. but apple was making garbage laptops for a while there, i ended up keeping my 2013 mac for 5 years, and then i guess i made it 3 years with my 2018, but that was mostly due to the component shortages. i was ready to throw the thing in the garbage after barely a year!
2 Likes
Mmh interesting…
From the 3.3 changes (and 3.4 for that matter).
https://developer.blender.org/rBe11c899e715b01f65b0a3b9b99cd69cf460209b1
I guess that means no more 5-6 min compile times.
Also a shame that this is windows only (I did not try yet if it runs under Paralells, there is a chance it might as it has only DX11 under requirements).
https://actorcore.reallusion.com/auto-rig
well that sucks. i was fine waiting for 5 mins once to save an hour of rendering on every overnight render. although i agree having it be an option so users can opt into the compile wait. i just hope that gets added promptly and not in some nebulous “future”.
have any of the metal optimizations survived? its hard to keep track because its felt like every time one gets added to a beta, it gets retracted a few weeks later due to one issue or another.
yes I did not mind the wait either, plus it is not like it had to recompile them that often.
Not sure might have a look when I get home, but then again who know how much it will change till 3.3 comes out.
2 Likes
I have to admit that I’m a bit disappointed by the optimization development for Metal. I thought that by now we’d have cut render times in half from when Cycles for Metal was initially introduced, but that doesn’t seem to be the case.
I had read on the dev boards that there was potential for up to a 60% decrease in render times, but either the Apple team hasn’t been able to figure it out, or the speed gains result in other slowdowns that make them inefficient.
1 Like
Yes I remember that too, like 2x improvements were talked about and that has been a while ago.
My guess its that either Apple underestimated the complexity of Blender or their own API is holding them back.
Worst case their hardware is not as good as they thought ?
At least their system is stable, still trying to figure out what is up with my windows machine.
3 Likes
I think it’s a combination of factors, including that the team from Apple actually working on this is probably severely under-staffed.
But fundamentally it doesn’t change the reality that Cycles for Metal will likely stay at the current performance level for quite some time.
2 Likes
True l, I am not going to complain at least there is GPU rendering on Macs again.
The thing that worries me more is there have been no news on metal viewport and Eevee next is non functional at the moment.
1 Like
What I mean is there has been no movement in months from what I remember, or at least is what it feels like.
2 Likes
we probably got spoiled by the surprise announcement of cycles metal support and then having a usable beta within a what felt like a few weeks. i assume they must have been working on it quietly for a while prior to that.
that said, i hope the viewport project hasn’t stalled or anything. i am getting a lot of application hangs and gpu errors in the regular viewport on my m1 max (apologies if i’ve shared these screens before)
Hmm maybe that’s what the view port compositor build always freezes when using the compositor for the viewport.
I tried under Parallels 18. Yes, it works fine. BTW Parallels 18 supports DirectX 12.
2 Likes
Hmmh,
I re-tested Blender M1 GPU Rendering with 3.4 Alpha
for Classroom and BMW.
Not only that I still don’t see any speed improvements.
When I just activate GPU Rendering, without “experimental”,
it will still Render CPU only.
If I activate “experimental”, it will Render by GPU
(90-92% load mostly only on my M1 Mini 16 GB)
but the Rendering looks pretty wrong.
Like no (soft) shadows in BMW, seems to float over ground
and strange (non-blurry) reflections of the Windows at the inner
Wall.
it is pretty amazing how rigging exploded in recent years
I am happy I only need to teach a car rig this semester
characters are so much more labor intensive
the motion tracing tools even with a phone app are still fantastic
This seems more in line with what I expect Apple to do. I don’t think they would confuse the M2 line with a 5nm and 3nm version of the CPU with the same fundamental name designation.
I suspect the 3nm tech is mid-2023, probably with a bump in the Mac Studio line first, and perhaps a new iMac 27" (or 32"). It also seems like a missed opportunity not to designate the 3nm chips with a matching M3 name.
I am predicting the M2 line will be one of Apple’s shortest lifespan CPU’s to date. It does feel more of a marketing-driven filler than an actual next gen CPU.
That’d be great. Right now my PC feels like a temporary band aid while waiting for Mac salvation. 
I really hope the next generation Mac Studio with new Apple Silicon chipset will indeed be announced somewhere next year, and will ideally feature hardware raytracing, to undo the GPU let-down of the current Mac Studio line-up.
1 Like