Discussion Apple Silicon SoC thread

[Eug]

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

 

[Hitman928]

The memory bandwidth Apple made available to a CPU in a laptop is insane. That is one area where it really is workstation class performance in a laptop. Now, I don't know how applicable it is for most users, but for those that can use it, there is absolutely 0 competition out there today in this format. The hardware accelerated tasks are obviously also very strong and are a big benefit of Apple being vertically integrated. From a CPU and GPU compute perspective, it is still quite impressive, but more in line with the node lead they enjoy compared to the other situations where the x86 competitors don't/can't offer competitive solutions in this format.

 

[StinkyPinky]

These "reviews" are terrible. Most of them are just running the same crap synthetic benchmarks we already had leaked to us. Is someone actually going to run those machines through their paces at some point? Anandtech is the only one that even made an effort and even that is somewhat lacking.

 

[igor_kavinski]

Anandtech is the only one that even made an effort and even that is somewhat lacking.

Well, AT admitted that it was tough trying to find something that makes use of the insane bandwidth and real world gaming benchmarks for the M1 are almost non-existent due to unavailability of a native AAA game. I guess the only other thing left to test is Windows ARM edition benchmarks running in Parallels.

 

[Panino Manino]

The expected numbers for these tasks.

 

[Eug]

The expected numbers for these tasks.

No such Max machine actually exists.

 

[igor_kavinski]

Some more benchmarks here: 14-inch MacBook Pro (M1 Pro) review: Life just keeps getting better for Mac users | Macworld

Seems the M1 Max's multicore prowess isn't particularly useful for the average user who should find the M1 Pro more than enough for their needs. Maybe data scientists may be better able to unlock the M1 Max's potential for their workloads.

 

[Eug]

Some more benchmarks here: 14-inch MacBook Pro (M1 Pro) review: Life just keeps getting better for Mac users | Macworld

Seems the M1 Max's multicore prowess isn't particularly useful for the average user who should find the M1 Pro more than enough for their needs. Maybe data scientists may be better able to unlock the M1 Max's potential for their workloads.

Even the M1 Pro 8-core (6+2) with 14-core GPU would be fine for most of those users, but it's interesting to see how many won't even consider that for their light workloads, just because it's not a "real" M1 Pro.

M1 Pro 8-core is probably 1/3 faster than M1 for CPU, and much, much faster than M1 for GPU.

BTW, someone posed an interesting question. Since those cores are clustered, how does Apple allow binning in the 8-core model? Does Apple allow deactivation of only one core per cluster?

 

[Hitman928]

Even the M1 Pro 8-core (6+2) with 14-core GPU would be fine for most of those users, but it's interesting to see how many won't even consider that for their light workloads, just because it's not a "real" M1 Pro.

M1 Pro 8-core is probably 1/3 faster than M1 for CPU, and much, much faster than M1 for GPU.

BTW, someone posed an interesting question. Since those cores are clustered, how does Apple allow binning in the 8-core model? Does Apple only deactivation of only one core per cluster?

1 core per cluster would be my guess.

 

[Roland00Address]

Even the M1 Pro 8-core (6+2) with 14-core GPU would be fine for most of those users, but it's interesting to see how many won't even consider that for their light workloads, just because it's not a "real" M1 Pro.

M1 Pro 8-core is probably 1/3 faster than M1 for CPU, and much, much faster than M1 for GPU.

BTW, someone posed an interesting question. Since those cores are clustered, how does Apple allow binning in the 8-core model? Does Apple allow deactivation of only one core per cluster?

Does the M1 Pro 8 core have any other limitation such as number of displays or anything else?

 

[Eug]

Does the M1 Pro 8 core have any other limitation such as number of displays or anything else?

Same number of displays for the M1 Pro 8-core and 10-core. The 8-core only has 14 GPU cores though, vs 16.

The 8-core comes with a lower wattage power brick.
You have to pay a bit extra to get the bigger power brick.

 

[Eug]

delete - double post

 

[naukkis]

I don't think anyone was claiming the power efficiency wouldn't be really good, it's just that a particular poster was claiming 500% better than the most efficienct x86 competition which is bollocks and these new tests show as much. Also, everything looks super power efficient compared to Intel's chips .

His point was spot on, it was opposing claim that Apple's lead is mostly from node advantage - like 20% when it actually is more like a half a order of magnitude more efficient. If that's a 3 or 5 weren't his point.

Apple's cores just are from totally different league than x86 rivals, and denial against it still lives strongly.

 

[DrMrLordX]

These "reviews" are terrible. Most of them are just running the same crap synthetic benchmarks we already had leaked to us. Is someone actually going to run those machines through their paces at some point? Anandtech is the only one that even made an effort and even that is somewhat lacking.

It may not be possible to get any better reviews. How much "real world" software is there in common between MacOS and Windows or Linux? It's almost not worth complaining about it anymore, though. Some people just want their SPEC scores.

Meanwhile:

Apple's M1 Pro, M1 Max SoCs Investigated: New Performance and Efficiency Heights

www.anandtech.com

(check the M1 Max score for CBR23)

vs:

Cinebench R23 Scores [Updated Results]

Cinebench R23 is the newest instalment of the most popular CPU-Rendering Benchmark Cinebench. We have the Score Results for all modern Processors.

www.cgdirector.com

(check the CBR23 MT score for the 5980HS)

But:

So in 526.blender_r and 511.povray_r, the M1 Max massacres the 5980HS, but in Cinebench R23 it . . . doesn't? Wouldn't it be nice of AT to run a current build of Blender on their M1 Max review sample instead?

 

[naukkis]

So in 526.blender_r and 511.povray_r, the M1 Max massacres the 5980HS, but in Cinebench R23 it . . . doesn't? Wouldn't it be nice of AT to run a current build of Blender on their M1 Max review sample instead?

R23 test runt once is so short that turbo-boosted chips can maintain it with high boost - that doesn't mean anything. Run that R23 test longer and it will gave similar results to Anand's spec tests.

 

[TESKATLIPOKA]

This is for everyone who is claiming M1 Max is leagues ahead of It's x86 competition in energy efficiency.

Techspot review of Ryzen 7 5800U

M1 Max(Package power: 34W) vs Ryzen 9 5980HS(Package power: 35W)
R23 MT: 12375 vs 11024
Difference in performance is only 12% in favour of Apple at similar power draw.

What would happen If I compared It against 15W Ryzen 7 5800U?
M1 Max(Package power: 34W) vs Ryzen 9 5800U(Package power: 15W)
R23 MT: 12375 vs 7394
Apple is 67% faster, but consumes 127% more power than AMD, which makes this AMD CPU 36% more power efficient. BTW It was a 10min long run, just to be clear.

 

[DrMrLordX]

R23 test runt once is so short that turbo-boosted chips can maintain it with high boost - that doesn't mean anything. Run that R23 test longer and it will gave similar results to Anand's spec tests.

R23 specifically loops for 10 minutes to avoid that effect.

This is for everyone who is claiming M1 Max is leagues ahead of It's x86 competition in energy efficiency.

I just want to point out, that I have found sources showing the 5980HS to perform much better than that in R23. The link I pasted above showed it scoring 12844.

 

[TESKATLIPOKA]

I just want to point out, that I have found sources showing the 5980HS to perform much better than that in R23. The link I pasted above showed it scoring 12844.

I thought It was because of the 10 minutes long run, but you actually mentioned R23 loops for 10 minutes by default.

 

[naukkis]

Denial against Apple's cores seems to be strong - haven't you red Ian's review at all? Cinebench is absolutely low efficiency side of M1 as it is 10-thread 128 bit fpu against 16-thread 256 bit fpu cpus where SMT-scaling is great and increases efficiency a lot - and still M1 can be both better performing and more efficient than Zen3. That still is insane achievement - or putting it other way, x86 cpu manufacturers should be ashamed that they are beat so strongly by cpu from some lifestyle company which main target isn't selling cpus.....

 

[Hitman928]

His point was spot on, it was opposing claim that Apple's lead is mostly from node advantage - like 20% when it actually is more like a half a order of magnitude more efficient. If that's a 3 or 5 weren't his point.

Apple's cores just are from totally different league than x86 rivals, and denial against it still lives strongly.

Those claiming the M1's lead was only due to the node advantage were a lot more accurate than him claiming M1 had a 500% advantage.

 

[naukkis]

Those claiming the M1's lead was only due to the node advantage were a lot more accurate than him claiming M1 had a 500% advantage.

That's just denial again.

 

[StinkyPinky]

Anyway, Alder Lake is due anytime now so I feel that is probably a better real world comparison since that is a new architecture

 

[Hitman928]

It may not be possible to get any better reviews. How much "real world" software is there in common between MacOS and Windows or Linux? It's almost not worth complaining about it anymore, though. Some people just want their SPEC scores.

Meanwhile:

Apple's M1 Pro, M1 Max SoCs Investigated: New Performance and Efficiency Heights

www.anandtech.com

(check the M1 Max score for CBR23)

vs:

Cinebench R23 Scores [Updated Results]

Cinebench R23 is the newest instalment of the most popular CPU-Rendering Benchmark Cinebench. We have the Score Results for all modern Processors.

www.cgdirector.com

(check the CBR23 MT score for the 5980HS)

But:

So in 526.blender_r and 511.povray_r, the M1 Max massacres the 5980HS, but in Cinebench R23 it . . . doesn't? Wouldn't it be nice of AT to run a current build of Blender on their M1 Max review sample instead?

SPEC usually does custom compilations of each software it benchmarks and tends to compile more for the least common denominators. For instance, in the x264/x265 tests they compile without AVX or SSE flags on x86 which is not at all how the software is actually used. This could paint a very different picture of video encoding performance as x86 kind of relies on those more modern instructions. Unless Anandtech is custom compiling the sub tests and turning these flags back on? Either way, some of the tests make it hard to really compare performance for real world apps when they're not testing the apps in the way they are actually compiled/run in the real world.

 

[Hitman928]

That's just denial again.

Ok, then show me where M1max or pro has a anywhere near a 500% advantage over Zen 3 in Cinebench. That was his claim. Or really any compute bottlenecked application that isn't using the custom accelerators.

(Edited)

 

[moinmoin]

Regardless of what everybody states, hypes up or play down regarding the M1 family's performance, I think we can all agree that if Intel and AMD care about efficiency Apple clearly is the competition to beat right now.

As for reviews I'm still missing one that explains to me why according to Apple itself M1 Pro/Max Macbook Pro's efficiency while browsing is only half that of M1 Macbook Air. Has anybody seen one such yet?

 

[DrMrLordX]

Denial against Apple's cores seems to be strong - haven't you red Ian's review at all? Cinebench is absolutely low efficiency side of M1 as it is 10-thread 128 bit fpu against 16-thread 256 bit fpu cpus where SMT-scaling is great and increases efficiency a lot - and still M1 can be both better performing and more efficient than Zen3. That still is insane achievement - or putting it other way, x86 cpu manufacturers should be ashamed that they are beat so strongly by cpu from some lifestyle company which main target isn't selling cpus.....

Um.

Blender can absolutely use AVX, AVX2, and even AVX-512 (I think?) and M1 Max tears apart the competition in that sub-bench. See below, as well.

SPEC usually does custom compilations of each software it benchmarks and tends to compile more for the least common denominators. For instance, in the x264/x265 tests they compile without AVX or SSE flags on x86 which is not at all how the software is actually used. This could paint a very different picture of video encoding performance as x86 kind of relies on those more modern instructions.

Just another reason to be skeptical of SPEC results.