Apple is an existential threat to the PC

Apple M1: 4+4 SoC with best [insert what you want] in the world. 8-core GPU. Fastest integrated GPU? 16-core Neural Engine. Thunderbolt and USB 4.

MacBook Air M1

Mac Mini with M1 and a fan.
 
Last edited:
And MacBook Pro 13" too. I guess they had to flip the 13" otherwise - based not the numbers they gave -the 13" Air would have murdered the 13" MBP in performance. Hardware shipping next week. It'll be interesting to see how fast they are once they're out there. I imagine compatible benchmarks will be rare but it'll have to be application run benchmarks for a while.
 
And MacBook Pro 13" too. I guess they had to flip the 13" otherwise - based not the numbers they gave -the 13" Air would have murdered the 13" MBP in performance. Hardware shipping next week. It'll be interesting to see how fast they are once they're out there. I imagine compatible benchmarks will be rare but it'll have to be application run benchmarks for a while.
They are still selling Intel MBP 13 alongside, and at higher price points. That’s not really a move that says “gonna murder my old products”.
 
They are still selling Intel MBP 13 alongside, and at higher price points. That’s not really a move that says “gonna murder my old products”.
Some people need 80x86, like anybody needing Bootcamp. I need Bootcamp, but may get away with VMWare in the future, depending on whether they leverage a Rosetta2-like translation layer in Windows. I can see Apple being happy to drop BootCamp but maybe is Microsoft push Windows 10 for ARM and that brand of ARM is compatible with M1, M2, M3... M10.
 
It is a bit underwhelming but expected. M1 is basically A12Z on a new process, maybe updated IP blocks, higher clocks, and some Mac perks.

4+4 CPU cores
8 GPU cores (now 2 TFlops, up from 1.x TFlops that A12Z has)
Up to 16GB soldered RAM

The baseline to leap from is so low that not “beating” these aging 14nm+++ dual-core and quad-core chips is a almost crime.

So they seem to be playing the “same performance, lower power” angle, instead of making larger chips that roll competitors in raw performance. Kinda make sense for these mobility first & cost sensitive ranges. The performance crowd is probably indifferent about this though.
 
Last edited:
So they seem to be playing the “same performance, lower power” angle, instead of making larger chips that roll competitors in raw performance.
There made so many points of higher CPU and GPU performance throughout. 3X here, 6X here, 2X here. Did you stop watching after the first 20 minutes?
 
There made so many points of higher CPU and GPU performance throughout. 3X here, 6X here, 2X here. Did you stop watching after the first 20 minutes?
I watched it, but I believe more in absolute numbers than selected comparisons, especially when they are comparing to who knows what.

Truth remains that it has 4+4 CPU cores and 2 TFlops 8-core GPU, which is same in scale as A12Z. Unless you make a strong bet on these cores being architecturally very different from A12Z and A14, extrapolation can give a sense of performance that isn’t too far fetched.
 
I watched it, but I believe more in absolute numbers than selected comparisons, especially when they are comparing to who knows what.

Truth remains that it has 4+4 CPU cores and 2 TFlops 8-core GPU, which is same in scale as A12Z. Unless you make a strong bet on these cores being architecturally very different from A14, extrapolation can give a sense of performance that isn’t far fetched.

A14 is something like ~40% faster than A12Z in single thread performance, so I don't know what you mean by "same in scale" as A12Z.
 
I watched it, but I believe more in absolute numbers than selected comparisons, especially when they are comparing to who knows what.
What absolute numbers are you waiting for? How are you proposing to compare 80x86 vs ARM? 3x quicker to compile in Xcode makes the point for me. That's a real world task, and there were plenty more.

This is 5nm, you'll see big improvements.
 
A14 is something like ~40% faster than A12Z in single thread performance, so I don't know what you mean by "same in scale" as A12Z.
Same in numbers of cores, and I did note that “A12Z [...] with updated IP blocks” which may include CPU cores.
 
Same in numbers of cores, and I did note that “A12Z [...] with updated IP blocks” which may include CPU cores.

The Apple M1 is a totally different beast. The 4 high-performance (ultra-wide execution) cores have 192KB instruction cache, 128KB data cache and 12MB of shared L2 cache. The 4 high-efficiency (wide execution) cores have 128KB instruction cache, 64KB data cache and 4MB of shared L2 cache.

Update: Oh, you were going with 8 cores is equal 8 cores logic. Ok then...
 
Last edited:
What is also intrestin, they say only 8 GPU Cores. No Shader count no clock speed and the fans are happy? Realy!?
 
The Apple M1 is a totally different beast. The 4 high-performance cores have 192KB instruction cache, 128KB data cache and 12MB of shared L2 cache. The 4 high-efficiency cores have 128KB instruction cache, 64KB data cache and 4MB of shared L2 cache.
These are not “totally different beast” when comparing to numbers from A13 (3+3). Say AnandTech’s educated guess for the big Lightning core was 128KB L1I, 128KB L1D and 8MB L2.

For L1I size, there wasn’t any A14 information that I know of, so it is hard to say whether the 192KB L1I is a feature of the new generation of cores on 5nm (A14 included), or that they are M1 exclusive.

As for L2, yes, it is true that it gets 4MB more than A12Z at the same core count. But A12Z is also one process node behind, and it also wasn’t quite clear if it is an architectural-wise bump/trend (so A14 gets more L2/core too) or it is M1 exclusive.

Clock speed is unknown. But with such large cache and known tight cache latency, it is unlikely that it will clock exceptionally higher than iPad Pro maximum boost. But for sure, Mac would have the thermal capacity to enable sustained performance at higher clocks.

https://www.anandtech.com/show/14892/the-apple-iphone-11-pro-and-max-review/3

I am sorry if having said it is “underwhelming” triggers people unintentionally. What I meant by “upgraded A12Z” is in terms of its positioning and design target, and the likeliness of A12Z’s 5nm successor having the same specification (= the bin for fanless MacBook Air). Of course it is going to get the newer IP blocks, and take advantage of the new process node.
 
Last edited:
So does it use HBM?
According to Apples Finnish pages, it does.
https://www.apple.com/fi/mac-mini/
Yksi yhteinen muisti. Vahvempaa joukkovoimaa. M1-siru mahdollistaa jopa 16 Gt:n supernopean yhteismuistin. Vähäisen viiveen HBM-muisti on integroitu yhteen komponenttiin, minkä ansiosta apit voivat jakaa dataa tehokkaasti prosessorin, näytönohjaimen ja Neural Enginen kesken. Näin kaikki tekemäsi asiat hoituvat nopeasti ja sulavasti.
 
Back
Top