People who want supporting evidence for their PoV.
While the Blender benchmark may not reflect the performance of the Mx GPU as well as Geekbench, it is easier to reason with because it distinguishes the Mx GPU by the number of GPU cores.
I think any consumer machines that come in now are going to be the 60 core. The 76 core counts as BTO I think and has a 1-2 week delay.
The complaint was that Geekbench gives M2 Ultra on average over 280k, but all results are below 225k.
It looks like the average score chart isn’t updated to dynamically, the 220000 scores look newer than the last update date on the main ranking chart.
It's more likely that the stock 60 core machines are arriving in customer's hands 2 days early.
The calculated average would only make sense if it uses results that are not published on the website.
They are always accurate.
Would this benefit GPU cores with Ray Tracing in a future iPhone (Sep 2023), M3 (Q1 2024) or M3 Extreme (Q1 2025)?Continuing my trawl through the WWDC sessions, came across this one:
Explore the USD ecosystem - WWDC23 - Videos - Apple Developer
Discover the latest updates to Universal Scene Description (USD) on Apple platforms and learn how you can deliver great 3D content for...developer.apple.com
Some pretty neat things in there. What stood out for me was:
- Model tag allows displaying USDZ in Safari (no more exporting to gltf)
- USD Files can be added Freeform(!)
- Contributed to the Maya USD implementation, hopefully making it suck less (had to use it recently and still bitter)
- Contributed to Blender USD support in 3.5, added USDZ support
- Support for materialX in Preview/MacOS along with a materialX editor in Reality Composer Pro
- Improved USD schema support in Preview (this plus metal support for MaterialX shaders could be awesome).
And then there were some nice performance bits and pieces:
- Can now display the high res version of the Alab scene (26GB GPU memory) on a laptop. Now super smooth (before it was...slow, so assuming this tied to updates to the Storm delegate)
- Blender 3.5 is full native metal app; metal viewport can be up to 4x faster than the old OpenGL one.
All of which puts paid to the idea that 'Apple doesn't care about 3D' and the old chestnut that has been repeated ad nauseam on the Mac Pro threads, 'Macs can't do 3D'. The integration of 3D via USD into the OS far exceeds that of Linux or Windows, and seems Apple are pretty keen to make 3D a first class citizen of their OS's from collaboration and review (Composer / Freeform) to supporting DCC apps used or authoring 3D content (Houdini, Maya and Blender all got a mention). It feels like most of these have been edging along in the background for a few releases now, but with the release of visionOS and the need to create 3D content for that all these various tools are really coming together.
tldr; after this wwdc feeling pretty sanguine about the future of 3D on AS
WWDC shows that Apple has put extra effort into its weakest points: 3D, gaming and machine learning.
At least as far as games and 3D, it seems that in both cases multi year projects finally reached tipping point. They'd been occasionally bits and pieces for the last year or two (some Metal improvements here, some RE:Villiage there), but feels like there's now a cohesive plans and direction in both areas.
Vray works fine with AS Machines. Mx Ultra is fast with Arnold and Vray based on CPU rendering. Redshift and Octane are different story on AS Macs. If M2 Ultra will be equal 3070 ti or 3080 you could use it for your compute demands by Rddshift and OCtane ( Both of them works well with C4D, Redshift works well on ZBrush ). You can check for 1st generation AS M1 Ultra rendering differences by render engine here.
opendata.blender.org
