My 3060 Laptop renders the BMW scene in 12.85sec. So that 4090 is mistaken or crippled somehow.
Edit: i saw the article and he certainly inverted the score of Classroom vs BMW and probably just put it in CUDA and not Optix
BTW
did Linus Media Group embargo’ed Apple ?
In the past They always had at least some Tech quickies or so.
No Apple hints since weeks.
And Mac Address seems to become iOS only.
i checked the daily build (jan 31) with metal gpu backend enabled, but i’m still getting that its not supported. how’d you get it working?
Apple doesn’t like LTT that much. They don’t get review samples if I remember correctly like other tubers (MKBHD for example). The last time the M1’s were released LTT had to go out and buy it themselves. If that still holds true they are probably still testing.
If you use the Metal backend, it actually works, even if it is greyed out. The code that greys it out on Mac wasn’t changed yet 
I find this just such a massive improvement from what the “meh, it’s still 5nm”, “meh, it’s only a couple more cores” sentiment that it has received!
As has been alluded before, Mac Studio M2 Max has become an even better potential proposition (if it comes out), M2 Ultra better offer something useful because M1 Ultra has been left in the dust and these laptops have become quite even better portable stations.
Seems not too bad to me. Still the M2 Max 38c is way overpriced.
Found my windows laptop in the list, scores 3650 points with its 3080 RTX . For 2499,- (ink. tax)
Compare to this the M2 Max 38c scores 1506 points. For 4149,- (ink tax)
Here is a link to my laptop:
- There is also a TI version, but its quite high in price, and scores not much more. The upgrade to 1 TB SSD is just 49,-, but it has 64 GRAM, so thats double then the Mac has. I think the compare is fair.
Thanks guys, definitely useful gains since Apple’s involvement in the Blender project was announced but RT cores are desperately needed if Apple are serious about 3D. However, the stop-gap GPU cache has gone a long way to fix the tile memory bottle neck. M3 should tell us more about their intentions IMHO. Cheers and thanks for all the clarification.
The display is quite dim at 380 nits compared to M2 Mac Pro 1,000 nits sustained, 1,600 nits peak (HDR).
Yes,
I think Maxtech has also to buy it themselves.
Even more for LTT that should be no financial problem,
I think they will get it easily more than back with a single video,
with their viewer numbers.
But somehow Linus seems to be tired of Apple.
But if you look at most benchmarks, you can clearly see that the
advancement is only (hardly) reached by the number of extra cores
in CPU and GPU.
Plus the same amount of extra power and heat and or fan rev needed.
OK, beside efficiency issues, we should better not care much where
it comes from and just buy.
So far I see only 2 options for the few more Apple positive Benchmarks
like for Blender.
- M1 really has a hardware bottleneck avoiding proper scaling and is
bad for certain higher demanding tasks with GPUs. - Apple artificially optimizes/accelerates Blender to prefer M2 over M1
or vice versa, optimized M2 for Blender as for popular Geekbench
In most other tests GPU, Video accelerating, … does not show any
architectural differences too.
M1 CPU usually always scaled quite well in the past, and in Blender
also GPU scaling Pro/Max/Ultra was on the better side.
As it arrived early, here you go as promised.
Tests done one the 16" M2 Max 30 core (on battery).
I did run every render twice.
Monster under the bed (the only one I checked on the release version).
3.41 = 1 min 38 sec
3.5 = 1 min 19 sec
Did not rerun all on the 14" 14 core but Monster under the bed takes 4 min 21 sec even with the 3.5 alpha.
I guess when this goes live it will be perhaps 1 min
https://developer.blender.org/D17153
Tree Creature 15 fps
Patry Tug from compiled shaders (material preview).
1st run 11 sec
2nd run 5 sec
Party Tug from solid view 28 sec
Amy animation playback
Material Preview 24 fps (left the frame limit on).
Eevee render 20 fps
Classroom 49 sec
BMW (just for fun) 22 sec
Lone Monk 6 min 40 sec
As for the M2 Pro. It looks like the CPU cores needs to be more powerful. Perhaps the M2 Ultra will eventually reach 25 fps.
I think that one will benefit of some optimization which is still up for review.
Cool. I didn’t know Affinity had a Benchmark setting. Just for S&G I ran my M1 Mini. Those M2 GPU scores are almost quadrupled in in the raster column compared to my M1 and almost tripled in the combined GPU. Nice!
Then I ran it on my Windows Laptop… embarrassing. 
@Tiwaz your M2 GPU raster score is almost 30X faster than my Windows laptop. lol
It’s cool to see bench test that are optimized for Metal starting to appear. Usually when Apple machines are benched against competing Windows products it with bench test that are built with Windows APIs in mind making the test almost like Apple starts with one hand tied behind its back.
Saying that, I’m sure if anyone runs the Affinity Bench with a 4090 the scores will be through the roof. But hey, those M2 scores are solid when comparing to my M1.
Yes easier than benching Photoshop where you have to get pudget bench which is not even Apple silicon native I think (or some of their benchmarks are not)
So far it is looking good and within expectations
I don’t have access to a 4090 so I can’t say.
Perhaps we can beat the 4090 in something.
Raster Single GPU 27669
Combined Single GPU 10150
Is one I could find