Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It's not completely insane.

1. N3B TSMC node used by M3 is reportedly low yielding. Apple likely wants to move away from it asap and move to N3E, which is the higher yielding node.

2. Rumors are that M3 Ultra is not planned due to low yields. M3 Max die shot does not have UltraFusion silicon. This supports the first theory.

3. Apple base the M chips on the A series cores. A18 Pro has already finished in design because Apple always starts making A series chips in Spring before a Fall debut. Therefore, A18 Pro is already in production using N3E node. Given enough early planning, it isn't surprising that M4 can be released early.

4. Reportedly, Apple plans to focus on GenAI at WWDC. They're also behind in GenAI. In order to catch up, they need to be aggressive. M4 is likely to have a big improvement in inference speed via the NPU and/or GPU.



I think 4 is underrated. From what I've seen Apple is making a play for on device inference. Gains on better screen, thinner phone, better camera selling points are pretty incremental now. But with on-device inference they now have access to: doubled the number, no cloud we care about privacy, on device translation capability, etc.


> They're also behind in GenAI

Where can I read up on that and what's state of the art in consumer GenAI chips?

Also thanks, all great points!


Think that whole major, unpatchable security flaw thing could also play a part? I could see them wanting to put distance between themselves on that fast too.


I don't think so. I think M4 was planned years in advance to debut around this time. You can't change this stuff in a few months.


M3s have a flag to disable the caching vulnerable to GoFetch on p-cores. M1s and M2s are the ones that have no mitigation besides running workloads on slower e-cores.


That‘s not true DMP can be disabled on M2 chips and probably also on the M1 chips. This was found by Hector Martin (@marcan) and will probably supported in Asahi Linux.


That's great news. Direct link for anyone else not up to date: https://social.treehouse.systems/@marcan/112238385679496096


I don’t recall any that affected silicon?


Key extraction from the HSM. It's unclear if this was ultimately patched in the kernel, but it appears patchable.

[1] https://gofetch.fail/

[2] https://social.treehouse.systems/@marcan/112238385679496096


Definitely not the HSM, this only affects the AP




Consider applying for YC's Fall 2026 batch! Applications are open till July 27.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: