Macs With M5 Chips Expected to Launch in This Order

[MacMan988]

I'm sure those are built for AI (Apple Intelligence)

 

[Rychiar]

The only thing that has challenged my 32GB M2 Max Mac Studio is AI. This round I am gagging to buy a 128+GB M4 Ultra and put the LLMs to work.

I only use AI in adobe and I don't know how much local power that uses

 

[ChrisA]

Under is more likely.

Apple had some early years with very high uplift.

View attachment 2471332
View attachment 2471333
View attachment 2471334


But in recent years, that has decreased a bit. A15, 16, 17 and M2 and M3 were closer to 10% than 20%.

A18 / M4 is once again a high uplift.

View attachment 2471335


But we haven't seen anything in a while like, say, the A10, with 50%, the A9, with 64%, and the A6, with a whopping 166% (the A6 was the one that changed from ARM Cortex to an Apple design, so it's arguably the first "Apple Silicon").

Therefore, my guess is A19 / M5 will be closer to 10%, and maybe we'll see a cadence of two years (A20 / M6), or even four years (A22 / M8) where we get a 20% uplift.

Gains in performance now are caused by making the feature size smaller and then either running the clock faster or putting more cores in the same space. They might continue doing this for a while but there is a hard limit at maybe around 0.4 nm (twice the size of a silicon atom) We are getting very close to what is physically possible and I doubr we will see ten more years of dramatic feature size reduction.

To go much faster after reaching the practical smallest feature size Apple will have to start moving software functions into specialized hardware, like the Neural processor.

This is about desktops not laptops.

The performance you get from a studio Ultra is quite absurd especially if they don’t maintain it and treat the pro users right. I can get the best Windows system out there and still have a lot of money left for the cost of these Macs. And the performance is so much better.

Better for what tasks? The vast majority of Apple's customers use their Mac as a media player, for web surfing or online shopping. Any Chromebook is up to that task and will perform well enough.

So it seems we must be talking about the 1% who are working with large media files, specialized kinds of engineering or experiments with AI research. Macs are strikingly good at some of these things. AI in perticular.

We complain about the cost of RAM on Mac but people who do that compare MAC RAM to commonity PC RAM. This is wrong. They should be comparing the cost of Mac RAM to the price Nvidia charges for video RAM on their GPUs. When I run an AI model on an Nvidia GPU, I need 8GB of ram to fit an 8GB model. But on a Mac, I only need 8GB of system RAM. So the Mac with 64GB installed is a total bargain compared to an Nvidia GPU with 64GB of VRAM. This is a big deal tot he 1% who need it. For the other 99%, that 8 minute cat video on YouTube will always take 8 minutes to watch no matter how fast your computer is.

SO pperformance deppends on what you are doing, some tasks are not sped up at all and others are very much.