M4+ Chip Generation - Speculation Megathread [MERGED]

[JouniS]

If the M6 Max represents another 25% jump in GPU performance - are we hitting a point where 98% of power users feel they have enough performance for their software needs?

Power users are pretty good at finding uses for better hardware. For example, if you are into local LLMs, you could find immediate use for a 1000x faster GPU.

 

[innerproduct]

M3 ultra is way more than 35% faster in gpu for many loads. Just look at blender scores. 4000 -> 7000 or so. Btw, m5 shows up there now. 70% faster than m4. Wow. https://opendata.blender.org/benchm...PI&group_by=device_name&blender_version=4.5.0

 

[Tdude96]

Power users are pretty good at finding uses for better hardware. For example, if you are into local LLMs, you could find immediate use for a 1000x faster GPU.

Yep, any AI is going to benefit from whatever GPU and memory that can be thrown at it.

AI's pushing things in a somewhat different direction with its needs compared to others, but power users are always going to demand more performance no matter what, AI or not. Applications and needs evolve. Back in the day, it was "I need to edit HD video in real time." Then UHD, 4k, 8k. At 240 fps. Plus VR. And whatever else gets tacked on that's still to come. I don't think there's any danger of running out of needs for more powerful computers anytime soon. Then there are folks doing stuff like this: https://forums.macrumors.com/threads/new-m1-studio-crashes-with-heavy-load.2427579/

The need for more isn't going to be sated by M5 or M6 or M20.

 

[Pressure]

M3 ultra is way more than 35% faster in gpu for many loads. Just look at blender scores. 4000 -> 7000 or so. Btw, m5 shows up there now. 70% faster than m4. Wow. https://opendata.blender.org/benchm...PI&group_by=device_name&blender_version=4.5.0

That's quite a feat! Solid improvement. Can't wait to see what the bigger SoCs can offer.