Discussion Apple Silicon SoC thread

Page 293 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Eug

Eug

Lifer
Mar 11, 2000
23,725
1,263
126
M1
5 nm
Unified memory architecture - LP-DDR4
16 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 12 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache
(Apple claims the 4 high-effiency cores alone perform like a dual-core Intel MacBook Air)

8-core iGPU (but there is a 7-core variant, likely with one inactive core)
128 execution units
Up to 24576 concurrent threads
2.6 Teraflops
82 Gigatexels/s
41 gigapixels/s

16-core neural engine
Secure Enclave
USB 4

Products:
$999 ($899 edu) 13" MacBook Air (fanless) - 18 hour video playback battery life
$699 Mac mini (with fan)
$1299 ($1199 edu) 13" MacBook Pro (with fan) - 20 hour video playback battery life

Memory options 8 GB and 16 GB. No 32 GB option (unless you go Intel).

It should be noted that the M1 chip in these three Macs is the same (aside from GPU core number). Basically, Apple is taking the same approach which these chips as they do the iPhones and iPads. Just one SKU (excluding the X variants), which is the same across all iDevices (aside from maybe slight clock speed differences occasionally).

EDIT:

Screen-Shot-2021-10-18-at-1.20.47-PM.jpg

M1 Pro 8-core CPU (6+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 16-core GPU
M1 Max 10-core CPU (8+2), 24-core GPU
M1 Max 10-core CPU (8+2), 32-core GPU

M1 Pro and M1 Max discussion here:

Page 78 - Discussion - Apple Silicon SoC thread

Page 78 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
forums.anandtech.com forums.anandtech.com

M1 Ultra discussion here:

Page 109 - Discussion - Apple Silicon SoC thread

Page 109 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
forums.anandtech.com forums.anandtech.com

M2 discussion here:

Page 127 - Discussion - Apple Silicon SoC thread

Page 127 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
forums.anandtech.com forums.anandtech.com

Second Generation 5 nm
Unified memory architecture - LPDDR5, up to 24 GB and 100 GB/s
20 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 16 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache

10-core iGPU (but there is an 8-core variant)
3.6 Teraflops

16-core neural engine
Secure Enclave
USB 4

Hardware acceleration for 8K h.264, h.264, ProRes

M3 Family discussion here:

Page 215 - Discussion - Apple Silicon SoC thread

Page 215 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
forums.anandtech.com forums.anandtech.com

M4 Family discussion here:

Page 263 - Discussion - Apple Silicon SoC thread

Page 263 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
forums.anandtech.com forums.anandtech.com
 
Last edited:
  • Like
Reactions: Nothingness, Nosware, Charolette-Yog and 7 others
Sort by date Sort by votes
richardskrad

richardskrad

Member
Jun 28, 2022
54
53
51
There’s no perceptible difference between using the M1 and M4 in day to day use. In the age of LLM’s, forget the CPU. It’s now about unified memory, GPU and the Neural Engine and Apple is already falling behind there. The 38 TOPS in the M4 is behind Qualcomm and Intel’s upcoming Windows chips.
 
  • Like
Reactions: Tlh97, Mopetar and dr1337
Upvote 0 Downvote
T

The Hardcard

Member
Oct 19, 2021
101
143
86
richardskrad said:
There’s no perceptible difference between using the M1 and M4 in day to day use. In the age of LLM’s, forget the CPU. It’s now about unified memory, GPU and the Neural Engine and Apple is already falling behind there. The 38 TOPS in the M4 is behind Qualcomm and Intel’s upcoming Windows chips.
Click to expand...
Apple remains in the driver’s seat there. unified memory has nothing to do with the physical placement of the memory. It is about the different IP blocks GPU, CPU, NPU, having coherent, memory management units, eliminating the need for data transfers. Without that, even blocks sharing the same memory have to move data back-and-forth.

Apple could maintain the top AI capabilities if they ever decide to adjust the extreme memory and SSD upgrade pricing. on top of unified memory, Apple’s Machine learning research group came up with an effective way to have some of the deep learning layers run off the SSD while mitigating the latency impact.

High memory capacity Macs Have proven to be very effective at running LLMs. Unfortunately, the memory and SSD cost is severely high.
 
  • Like
Reactions: Viknet
Upvote 0 Downvote
poke01

poke01

Golden Member
Mar 8, 2022
1,272
1,446
106
Let’s address Geekerwan’s inconsistency between their videos. I don’t think they should be cross compared at all.

IMG_0421.png
This is their iPhone 15 Pro video. Take a look at the floating point number it’s higher than M4 at similar clocks. Clear indication of different datasets between videos.

IMG_0423.png
This is from their M3 iMac video. Here the floating point is lower than A17 Pro. Keep an eye on the Integer number for the M3 and i9-14900K.

IMG_0426.png
Here the Integer numbers for M3 and i9 -14900K are lower when compared to their last M3 video. So clearly different datasets are being used for each video.

IMG_0419.png
Some using using this chart say the IPC improvement is 4.6%. But it’s not same clocks and there are way too many variables.
Especially the frequency.


IMG_0427.png
However, this screenshot should provide the most accurate IPC improvement over M3. The frequency was controlled and fixed for M4 using Xcode and was tested at room temperature. Here we see an Integer improvement of 7.3%.
 

Attachments

  • IMG_0418.png
    IMG_0418.png
    3.5 MB · Views: 11
Last edited:
  • Like
Reactions: name99, Tlh97, Mopetar and 4 others
Upvote 0 Downvote
poke01

poke01

Golden Member
Mar 8, 2022
1,272
1,446
106
Hitman928 said:
Edit: @Nothingness to give a possible answer to the lingering question as to why he has 2 different IPC improvement values between the videos, the M4 has the same IPC in both videos (with <1% error margin) but the A17 has slightly higher IPC than the M3. My guess is that the A17 does have slightly higher IPC due to the 3X larger SLC on the die which reduces the IPC improvement from 7.3% to 4.6%.
Click to expand...
That would explain the higher IPC results for A17 Pro


I
 
Upvote 0 Downvote
poke01

poke01

Golden Member
Mar 8, 2022
1,272
1,446
106
still doesn’t address the fact why the M3 and i9-14900K had different results in the M3 video and the M4 video. So there still be inconsistency?
 
Upvote 0 Downvote
N

Nothingness

Platinum Member
Jul 3, 2013
2,688
1,273
136
poke01 said:
still doesn’t address the fact why the M3 and i9-14900K had different results in the M3 video and the M4 video. So there still be inconsistency?
Click to expand...
I didn't watch Geekerwan videos. He might explain the conditions of his runs. At first approximation, as you wrote, one should not compare results extracted from different videos as he obvously changes things (different SPEC results for M4 for instance). Note I don't dismiss his results, I just miss some information to feel more confident.
 
  • Like
Reactions: Tlh97, SarahKerrigan and poke01
Upvote 0 Downvote
H

Hitman928

Diamond Member
Apr 15, 2012
5,543
8,638
136
poke01 said:
@Hitman928 @Nothingness

Ahh this is so much clearer now. This explains so much.

Yeah, let’s stick to comparing M-series to M-series from now on.
Click to expand...

We can, but the previous comparison point for x86 and Arm was the A series CPUs, so if that is the high water mark for the industry, I think it's only fair to compare the M-series to it as well. In the end, though, the few percentage points difference doesn't change anything in the discussion about Apple slowing way down in IPC improvements compared to their competitors. That doesn't mean this will always be the case, but it's a very interesting trend.

As far as Geekerwan tests go, not having a clear and repeatable test procedure makes it harder to put confidence in the results in general, even in the same video when we don't know for sure how the tests are being conducted for each CPU. It seems to be the best we have to go with for now, though, since Andrei and Dr. Cutress left. The new staff has still ran SPEC for the last desktop releases, but that's it and who knows if even that will continue.
 
  • Like
Reactions: Tlh97, The Hardcard and Nothingness
Upvote 0 Downvote
N

Nothingness

Platinum Member
Jul 3, 2013
2,688
1,273
136
igor_kavinski said:
I think it will continue. They wouldn't want to waste their license cost and AT is one of the very few that is posting these SPEC results in reviews.
Click to expand...
The license cost is a one time pay and it's relatively inexpensive (currently $1200, but I think it used to be less a few years ago) so I'm alas not sure the cost is enough to mandate use of SPEC.

And SPECv8 should be coming. Work on it has started a few years ago. https://www.spec.org/cpuv8/
 
Upvote 0 Downvote
reggie_fils_aime

reggie_fils_aime

Member
Mar 8, 2024
61
165
66
richardskrad said:
There’s no perceptible difference between using the M1 and M4 in day to day use. In the age of LLM’s, forget the CPU. It’s now about unified memory, GPU and the Neural Engine and Apple is already falling behind there. The 38 TOPS in the M4 is behind Qualcomm and Intel’s upcoming Windows chips.
Click to expand...

Saying that there's "no perceptible difference" between M1 and M4 is very funny to bring up right before TOPS performance. If you could find one in a 1,000 person auditorium with an actual daily use-case for an NPU or TOPS, I'll give you a thousand dollars. As is, the average use for an NPU is to do stuff that Apple has already incorporated into MacOS, like blurring out backgrounds and facial tracking during a zoom meeting.
 
  • Like
Reactions: Tlh97, Viknet, Ghostsonplanets and 2 others
Upvote 0 Downvote
SarahKerrigan

SarahKerrigan

Senior member
Oct 12, 2014
553
1,233
136
Nothingness said:
The license cost is a one time pay and it's relatively inexpensive (currently $1200, but I think it used to be less a few years ago) so I'm alas not sure the cost is enough to mandate use of SPEC.

And SPECv8 should be coming. Work on it has started a few years ago. https://www.spec.org/cpuv8/
Click to expand...

When I bought my CPU2017 license at release, I believe it was US$1k.
 
  • Like
Reactions: Tlh97, igor_kavinski and Nothingness
Upvote 0 Downvote
poke01

poke01

Golden Member
Mar 8, 2022
1,272
1,446
106

Gokhan Avkarogullari (@gavkar@mastodon.gamedev.place)

Game Porting Toolkit 2 is out https://developer.apple.com/games/game-porting-toolkit/ #wwdc24
mastodon.gamedev.place mastodon.gamedev.place
Apple actually did this before Qualcomm. It’s out today as a beta. AVX2 support is going to a game changer for Mac gaming. The community can make ports easier now.

Nice to see head of Metal group promoting community apps.

Gokhan Avkarogullari (@gavkar@mastodon.gamedev.place)

@jimmyjamesuk@mastodon.social @schwa@mastodon.social if you are planning on porting your windows game to Apple platforms your tasks got easier. If you are someone who use crossover or whiskey to play windows games on the Mac now a lot more of those work and lots of them work faster/better
mastodon.gamedev.place mastodon.gamedev.place
 
Last edited:
  • Like
Reactions: Ghostsonplanets
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom