Something similarly groundbreaking would requite a similarly radical redesign. Maybe there will be one another couple decades from now, but I can't even begin to imagine what it would be. Maybe something related to ML if anything.
Sure but..already the second gen didn't show gains _that_ impressive over the first one.
If M1 successors will start showing Intel 2010s gains gen over gen, it's not gonna do great.
It seems natural that M series chips won’t forever be uniquely positioned ahead of the pack. I don’t think that anyone besides Apple actually wants that in the first place though. I have every expectation that they’ll keep up with the pack, or simply move back to third party chips, even if that means changing architectures, which Apple is not scared of doing.
I don’t think performance is the only metric. For all we know, there could have been some considerable cost reduction.
Just... staggered. What an upgrade.
Yes, that was my experience with well-established players like AMD, Nvidia and Qualcomm.
Then again, there just aren't that many people in the world with experience designing GPUs or writing systems software for them. What you would often see is a manager and their whole team switching to a competitor. People prefer working with other people they already know and trust.
In addition, GPUs are massively complex and it simply takes a good amount of time to understand a new architecture well enough to make substantial contributions. I spent five years at NVidia and felt that I had only scratched the surface.
When you're as vertically integrated and rich as Apple it makes sense.
Being rich isn’t really related, it’s that they can still profit while doing this (some people say “sheep” while others say “good products with great integration”), where nobody else can, which makes them rich. ;)
It isn't just that Apple can afford a particular part. Apple funds some of the R&D for that part, the building of the factory (sometimes, multiple factories) that create that part, and will commit to buying 100% of its capacity for years.
They aren't just outbidding others for capacity. They are making long-term strategic partnerships to _build_ that capacity.
> Putting memory on package runs counter to repairability, modularity, etc.
Everything is a trade-off. For traditional OEM integration for desktops and laptops, memory on package would be too limiting. Apple is one of the few companies that can pull it off.
Sure, but do we expect AMD/Intel to do this for their processors anytime soon?
Apple can decide that their custom processor only needs two memory configurations to meet their product need.
Intel and AMD can't easily make that decision for all the OEMs in the world, and the architectural ramifications pretty much demand one to decide one way or another for a given chip generation.
It's definitely possible with stacked chips we see a similar mind of process-and-packaging driven boost.
I'd expect more revolutionary changes to not come from digital logic improvements, but rather:
1. Adding specialized silicon for specific tasks
2. manufacturing changes, such as moving to chiplets, allowing for a wider variety of processors.
"...as well as the integrated circuits that were developed in Israel, and the jewel in the crown: the Israeli team played a central role in developing the premium version of the company's flagship M1 processor, including the M1Pro and M1Max chips designed to support premium Mac computers such as MacBook Pro and MacBook Studio. These chips were built here in Israel while working with other teams worldwide, including at the headquarters in Cupertino. The integration with the verification applications and processes was also carried out here."
https://en.globes.co.il/en/article-apple-to-open-jerusalem-d...