Discussion Apple Silicon SoC thread

Eug · Nov 10, 2020

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:

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

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

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

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

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

poke01 · Nov 6, 2024

Early blender scores. I don't know what Apple did with M4 CPU but those are really good Blender scores. The Base M4 is on par with M3 Pro and M4 Pro is nearly on par with M3 Max.

mvprod123 · Nov 6, 2024

poke01 said:
View attachment 111055
Early blender scores. I don't know what Apple did with M4 CPU but those are really good Blender scores. The Base M4 is on par with M3 Pro and M4 Pro is nearly on par with M3 Max.

M4 GPU is only 13% faster in Blender than M3
M4 Pro GPU is 43% faster than M3 Pro.

Apple emphasized a 2x improvement in RT. I expected better results. Perhaps the reason is the limited memory bandwidth, which is not enough for the GPU. I hope for a better result for the M4 Max.

poke01 · Nov 6, 2024

mvprod123 said:
M4 GPU is only 13% faster in Blender than M3
M4 Pro GPU is 43% faster than M3 Pro.

Apple emphasized a 2x improvement in RT. I expected better results. Perhaps the reason is the limited memory bandwidth, which is not enough for the GPU. I hope for a better result for the M4 Max.

The GPU improvements are disappointing. Nvidia will have a big lead against the Mac Pro/Mac Studio with tasks that are not memory limited.

The Max chip has the same core count as last generation, in fact I expect less gains than M4 Pro.

mvprod123 · Nov 6, 2024

poke01 said:
The GPU improvements are disappointing. Nvidia will have a big lead against the Mac Pro/Mac Studio with tasks that are not memory limited.

The Max chip has the same core count as last generation, in fact I expect less gains than M4 Pro.

Well, we have to admit that Apple has better support for 3D rendering software than AMD and Intel GPUs. I wouldn't be surprised if Apple's RT is also better.

If we roughly calculate the score for the potential M4 Ultra, it could be as high as 4090.

mvprod123 · Nov 6, 2024

https://twitter.com/x/status/1854245853533438008

mvprod123 · Nov 6, 2024

It's even faster than new 9800X3D in Photoshop lol

gdansk · Nov 6, 2024

mvprod123 said:
It's even faster than new 9800X3D in Photoshop

Wow, 25% higher IPC in this test. Phenomenal.

digitaldreamer · Nov 6, 2024

mvprod123 said:
It's even faster than new 9800X3D in Photoshop lol

I wonder how much of a role the 128GB RAM played.

poke01 · Nov 6, 2024

that’s not a bad score at all for the 9800X3D considering it’s only $479.

I don’t think anyone is coming close to M4 Max on laptop in Photoshop. That’s an excellent score but it is very expensive to get an M4 Max

Meteor Late · Nov 6, 2024

poke01 said:
that’s not a bad score at all for the 9800X3D considering it’s only $479.

I don’t think anyone is coming close to M4 Max on laptop in Photoshop. That’s an excellent score but it is very expensive to get an M4 Max

There is not much core scaling in that benchmark, though, you can see how a 9700X is pretty much similar to a 9950X, so this means an M2 Pro will be similar to an M2 Max in CPU performance in that benchmark, M4 may not be that much behind too.

Meteor Late · Nov 6, 2024

I have checked I think, 12600 points for M4 Pro, 10353 for M4. But the benchmark has parts that use the GPU, hence if HWU tested with the same 4090 for all the CPUs, then there will be no difference in those GPU parts, unlike with Apple parts, M4 will have much worse GPU than M4 Pro and so on.

mvprod123 · Nov 6, 2024

Meteor Late said:
There is not much core scaling in that benchmark, though, you can see how a 9700X is pretty much similar to a 9950X, so this means an M2 Pro will be similar to an M2 Max in CPU performance in that benchmark, M4 may not be that much behind too.

Because Photoshop is very much an (almost) single-threaded benchmark.

The Hardcard · Nov 6, 2024

mvprod123 said:
https://twitter.com/x/status/1854245853533438008

The 4090 is severely bottlenecked on that system. With the 9950x the GPU score is 143, about double the M4 Max.

DZero · Nov 6, 2024

I want to see how ARL fares there.

mvprod123 · Nov 6, 2024

The Hardcard said:
The 4090 is severely bottlenecked on that system. With the 9950x the GPU score is 143, about double the M4 Max.

View attachment 111118

This is a comparison with the laptop version of the 4090.

mvprod123 · Nov 7, 2024

poke01 said:
View attachment 111055
Early blender scores. I don't know what Apple did with M4 CPU but those are really good Blender scores. The Base M4 is on par with M3 Pro and M4 Pro is nearly on par with M3 Max.

M4 Max score:
+14% faster in CPU
+16% faster in GPU
vs M3 Max

FlameTail · Nov 7, 2024

Meteor Late said:
Eh 5.7GHz is on Desktop, on Laptop there is no 5.7GHz chip because it would consume a massive amount of power for single core. ARM limit is not 4.5GHz, I'm sure M4 can clock a bit higher, 200-300MHz more for sure, at what power cost though I don't know, probably very costly.

Apple doesn't bin based on frequency. If they did, I think there might be M4 chips that hit 4.7-4.8 GHz.

At any rate, I think there is a good possibility that M5 will breach the 5 GHz barrier next year, nor will it be the only ARM CPU to do so.

mvprod123 · Nov 7, 2024

The first reviews of M4 Macs have already appeared.

Cinebench 2024:

M4:
ST: 172
MT:1003

M4 Pro:
ST: 179
MT: 1744

M4 Max:
ST: 182
MT: 2043

FlameTail · Nov 7, 2024

mvprod123 said:
The first reviews of M4 Macs have already appeared.

Cinebench 2024:

M4:
ST: 172
MT:1003

M4 Pro:
ST: 179
MT: 1744

M4 Max:
ST: 182
MT: 2043

View attachment 111172

How is the single core score of M4 Pro/M4 Max higher than base M4?

Do they clock higher? Or perhaps they have a larger shared L2 cache !?

Eug · Nov 7, 2024

Reviews say the M4 Pro Mac mini can get a little bit noisy under full load.

For the record, the M4 Pro has a copper based cooling setup, whereas the M4 is all aluminum.

Meteor Late · Nov 7, 2024

FlameTail said:
Apple doesn't bin based on frequency. If they did, I think there might be M4 chips that hit 4.7-4.8 GHz.

At any rate, I think there is a good possibility that M5 will breach the 5 GHz barrier next year, nor will it be the only ARM CPU to do so.

But I'm not talking about binning here. I'm saying Apple is not pushing the single core to the max possible. I mean that in the current chips there is probably 200-300MHz headroom available if one is willing to consume like 20-25W single core, it's just that Apple doesn't want to. x86 camp is around 30W single core on Desktop, and that's without counting I/O die external power into account. Allow me to have doubts about Apple consuming what, around 10W single core and stopping there, not being able to clock 200-300MHz more using 20-25W.

Meteor Late · Nov 7, 2024

Eug said:
Reviews say the M4 Pro Mac mini can get a little bit noisy under full load.

For the record, the M4 Pro has a copper based cooling setup, whereas the M4 is all aluminum.

I mean, M4 P core can consume quite a bit more power according to Geekerwan testing, and M4 Pro this time has 10 P cores, not surprising at all.

DZero · Nov 7, 2024

Wait... how strong is the M4 Max GPU? Compared to dGPUs?

naukkis · Nov 7, 2024

Meteor Late said:
But I'm not talking about binning here. I'm saying Apple is not pushing the single core to the max possible. I mean that in the current chips there is probably 200-300MHz headroom available if one is willing to consume like 20-25W single core, it's just that Apple doesn't want to. x86 camp is around 30W single core on Desktop, and that's without counting I/O die external power into account. Allow me to have doubts about Apple consuming what, around 10W single core and stopping there, not being able to clock 200-300MHz more using 20-25W.

Apple chips aren't binned for clocks - so worst silicon limits clocks. With such binning x86 cpus wouldn't have 5.7GHz boost clocks but something closer 5GHz. Just binning silicon for clock would give Apple that 200-300 MHz more with same 10W single core consumption. Giving those best binned parts 20-25W single core power would increase clocks much more.

mvprod123 · Nov 7, 2024

DZero said:
Wait... how strong is the M4 Max GPU? Compared to dGPUs?

Blender

Discussion Apple Silicon SoC thread

Lifer

Platinum Member

Member

Platinum Member

Member

Member

Member

Diamond Member

Junior Member

Platinum Member

Member

Member

Member

Senior member

Senior member

Member

Member

Diamond Member

Member

Diamond Member

Lifer

Member

Member

Senior member

Senior member

Member