Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster


Apple today announced the M6 chip, its first built using a 2-nanometer process and featuring a Dual 16-core Neural Engine.

m6-chip.png

Apple says the 2nm process packs greater transistor density into a smaller die, contributing to the chip's overall gains in performance and power efficiency.

The M6 chip features a larger, more powerful 12-core CPU, two more cores than the M5, made up of two super cores, four performance cores, and six efficiency cores. Apple says the chip delivers the world's fastest single-threaded performance and up to 1.2x faster multithreaded performance than the M5, and up to 2.4x faster than the M1.

The chip also features a 12-core GPU, with two more cores than the M5, with a Neural Accelerator built into each core. Apple says this delivers nearly a 30% increase in peak GPU compute for AI compared to the M5, and more than 8x compared to the M1, resulting in faster prompt processing for on-device LLMs.

The GPU also gains updates to the shader core architecture, Dynamic Caching, and hardware-accelerated ray tracing, along with a 50% increase in geometry rates for more complex graphics.

The M6 introduces a Dual 16-core Neural Engine, offering up to 2x the peak compute of previous generations. Apple says system frameworks can automatically use both engines at once, allowing applications to run AI models faster.

The chip supports up to 32GB of unified memory and up to 170GB/s of unified memory bandwidth, which is a 10% increase over M5 and 2.5x more than the M1.

The M6 debuts in the new Mac mini, which launches on Tuesday, September 22.

Article Link: Apple Reveals M6 as First-Ever 2nm Chip
 
Last edited:
  • Love
Reactions: blob.DK
Hmm… so around 10% architectural GPU compute improvements and 20% thanks to more cores?

Up to 24 GPU cores for M6 Pro chips whenever (if) they are released?
 
The M6 was rumored to have 200 GB/sec of unified memory bandwidth. 170 is an improvement, of course, but still a bit of a disappointment.
 
It’s a jump, but these Mac mini machines are probably bullet proof for the kinds of usage they typically get. So expect 10+ years of working order. €2 a week.
Bold of you to assume Apple will support this device for that long. Given their recent Mac history, that isn't exactly a given.
 
  • Like
Reactions: Morod
Excited to see the GPU benchmarks… those claims of a 2x improvement over M4 are pretty huge. Im probably skipping the next few generations due to price hikes but if they keep up these year on year GPU gains then its gonna be a hell of an upgrade from my M3 Max one day!
 
  • Like
Reactions: PBG4 Dude
if you use M1 this is huge upgrade...gpu and bandwidth alone
Cannot wait for the Max for my maya usage

M5 Ultra comes with PCIe Gen 6
 
It’s a jump, but these Mac mini machines are probably bullet proof for the kinds of usage they typically get. So expect 10+ years of working order. €2 a week.

Only if you don't want to make use of whatever new-fangled "AI" additions come down the pike in a couple of years.

It's quite possible, maybe even likely, that a few years from now 16GB RAM will just not be enough to host the bestest of thingabobs.
 
Only if you don't want to make use of whatever new-fangled "AI" additions come down the pike in a couple of years.

It's quite possible, maybe even likely, that a few years from now 16GB RAM will just not be enough to host the bestest of thingabobs.
Buy what you need now and in the near future and not what you hypothetically might need in a couple years.

If someone wants to potentially future-proof for an unknown to-be-released AI function, someone can do that, but that’s a want and not a need.
 
  • Like
Reactions: Happy_John
The M6 was rumored to have 200 GB/sec of unified memory bandwidth. 170 is an improvement, of course, but still a bit of a disappointment.
It's only a disappointment if you believed the 200GB/sec rumours for which there was no basis whatsoever.
considering the M1 was 80gb/s....more than twice in bandwidth alone
The M1 was just under 70 GB/sec, the rest of the M1 family had faster RAM.

The M6 weighing in with 170GB/sec on the same bus width is quite good.
 
Buy what you need now and in the near future and not what you hypothetically might need in a couple years.

If someone wants to potentially future-proof for an unknown to-be-released AI function, someone can do that, but that’s a want and not a need.
And by the time those potential “needs” become realised, there’ll be newer, faster, shinier SoCs bring released.

Future proofing might if made sense a decade or two ago, but it really doesn’t now. The industry has changed.

Much as I hate it, hardware you buy now is simply not futureproofable, it is designed to be disposable and replaceable.

Hardware released now is not being designed for the tasks of tomorrow - they’re only designed for the tasks of today.

The whole iPhone 16 range is a good example of this.
 
  • Like
Reactions: neuropsychguy
This is also the first time Apple put all three tiers of CPU cores on a single chip. The choice of 6 E-cores is interesting. Perhaps because they are so small that using just 2 or 4 would not have saved a lot of space. Or am I the only one thinking 2 + 6 + 2 would be a more practical configuration than the 2 + 4 + 6 we got? Of course 4 + 4 + 2 would be even better.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.