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

 

[jeanlain]

The test is completely irrelevant UNLESS the laptop is plugged in.

Can you read? The post you quoted says that performance, as shown on the images above, was measured with all laptops plugged in.

Apple also showed results on battery, but these are not on the images above. You'd have to watch the video.

 

[eek2121]

Can you read? The post you quoted says that performance, as show on the screenshot above, was measured with all laptops plugged in.

Can YOU?

 

[StinkyPinky]

Apple is being Apple. The results will speak for themselves very soon. Someone should be able to get their hands on one very shortly.

 

[igor_kavinski]

If Apple's GPU can avoid the performance degradation on battery power and beat PC laptop GPU on battery, that's a big win.

 

[jeanlain]

Apple is being Apple. The results will speak for themselves very soon. Someone should be able to get their hands on one very shortly.

Assuming testers report precise CPU and GPU power consumption data, which is rather rare. I don't think I've ever seen detail data from power consumption during a single-core geekbench run for instance.

 

[uzzi38]

So if that is indeed a normal score, Apple’s minor perf advantage appears to come entirely from the node. If the chip didn’t have the two efficiency cores, if would have lower multicore performance than Ryzen H)



lol no. 3060 mobile.

EDIT: Apple was very misleading with the graphs. Ask @uzzi38 what he thinks about it.

The one graph that seems particularly misleading is the iGPU comparison. 32EU TGL is rather weak by todays standards (comparison chip is the 11800H) when 96EU TGL and Vega 8 on Renoir/Cezanne both exist and are 2-2.5x faster. The GPU performance is still outstanding given the power budget mind you, but Apple's graphs are highly exaggerated as per usual.

I think they're also comparing full system power or something and not pure iGPU power, because my 1135G7 never exceeds 13W in games. Most of the time even when maxed it caps out at 11.3W, the 13W was something I only ever saw when I set Yakuza 0 to some ridiculous internal resolution like 4x native or something.

Also, the CPU perf is actually kinda worse than I expected if I'm honest. 70% faster than an 11800H at 30W, but the 5900HX is already ~35% faster than the 11800H at that power. Little bit of a let-down there.

 

[eek2121]

If Apple's GPU can avoid the performance degradation on battery power and beat PC laptop GPU on battery, that's a big win.

The issue is that GPUs don't 'degrade' due to performance. My 3070 drops to 25W when unplugged. That is a far lower power budget than even my CPU gets. If Apple's charts are indeed accurate, they had 60W of headroom to play with on battery. I do agree with everyone that says 'wait for benchmarks', but I see all these ridiculous claims about performance, then I see Apple's marketing department being SUPER creative about the wording on every single slide. If they have really awesome chips, why not do it as AMD does and show some actual head to head benchmarks and performance numbers? Why not show Civ 6 or Subnautica, head to head, and give us the real numbers behind both framerate and power usage? Because the 3070 and 3080 will win out, that is why, and Apple's own chips will look far less impressive than they want.

 

[jeanlain]

Can YOU?

View attachment 51593

This is not from the images I am referring to, i.e. the "complete" grey curves on the previous page. Would you watch the video? Because you don't know what you're talking about.

The footnote you're showing refers only to the darker part of the curve. When plugged in, the 3080 mobile is like 5-10% faster than the M1Max, but it consumes like 3X the power.

 

[eek2121]

The one graph that seems particularly misleading is the iGPU comparison. 32EU TGL is rather weak by todays standards (comparison chip is the 11800H) when 96EU TGL and Vega 8 on Renoir/Cezanne both exist and are 2-2.5x faster. The GPU performance is still outstanding given the power budget mind you, but Apple's graphs are highly exaggerated as per usual.

I think they're also comparing full system power or something and not pure iGPU power, because my 1135G7 never exceeds 13W in games. Most of the time even when maxed it caps out at 11.3W, the 13W was something I only ever saw when I set Yakuza 0 to some ridiculous internal resolution like 4x native or something.

Also, the CPU perf is actually kinda worse than I expected if I'm honest. 70% faster than an 11800H at 30W, but the 5900HX is already ~35% faster than the 11800H at that power. Little bit of a let-down there.

The GB5 result is a bit baffling. I hope we get more results. It seems like it was underperforming in multicore.

As far as GPU performance, it's almost Apples and Oranges anyway because you have dedicated GDDR6 on the card along with DDR4 on the laptop vs. shared LPDDR5 on the Macbook. A fair comparison would have been to leave both plugged in and compare GPU core power between both systems.

 

[eek2121]

This is not from the images I am referring to, i.e. the "whole" grey curves on the previous page. Would you watch the video, because you don't know what you're talking about.

I did watch the video. Yes it is on the image. You don't understand the graph.



EDIT: In case you don't understand, they aren't measuring real performance. They are "extrapolating" perf/watt.

 

[igor_kavinski]

Why not show Civ 6 or Subnautica, head to head, and give us the real numbers behind both framerate and power usage?

I don't think they regard gamers as their target audience for their "Pro" laptops, else they would have showed AAA games in their stream. But they do want game developers to use their laptops.

 

[jeanlain]

EDIT: In case you don't understand, they aren't measuring real performance. They are "extrapolating" perf/watt.

They aren't measuring real performance? So they're lying I suppose?
First you claimed they only used iGPUs, then that they only measured on battery, and now that you've been proven wrong, you assert that they haven't actually measured performance. What's next?

EDIT; you think that the dimmed part of the curve is extrapolation from Apple's part, based on measures on battery power? That Apple couldn't be bothered to plug the laptop in to measure actual performance in this situation?

 

[Ajay]

Eh boy. Guess I'll leave this thread till some reviews come out. Stuff always starts going to sh*t when we don't have enough info

 

[Thala]

I did watch the video. Yes it is on the image. You don't understand the graph.

It is you talking nonsense the whole time. The light grey graph represents the laptop being plugged in, while the dark grey graph shows the power/performance limit of a laptop on battery. So Apple is demonstrating both a) the power saving compared to a plugged-in laptop at roughly iso-performance and b) the performance gain compared to a laptop on battery at roughly iso-power.
There is _NOTHING_ misleading about this.

 

[Ajay]

I don't think they regard gamers as their target audience for their "Pro" laptops, else they would have showed AAA games in their stream. But they do want game developers to use their laptops.

Nope, they really haven't for years. Really, it's just the iMacs that need some gaming performance as they are likely to be in a family environment (for duh childrens, teens, non-hardcore gamers).

 

[nxre]

Stop trying to make sense of Apple's graphics, it's just marketing.
Considering this is shipping next week, I'm expecting reviews this thursday-friday, so we won't have to wait much longer for some concrete numbers to judge, instead of having to argue over highly simplified marketing graphics.

Onto that geekbench score, single core is expected as I don't think they will be increasing clock frequencies within the same gen(M1/Pro/Max), only core counts/cache/bandwidth. MC is lower than expected, they claimed a 70% improvement over M1 if I remember correctly, which would put it in the ~12500-13000 region. They don't usually inflate their numbers, so either this is a bad sample or the performance improvement doesn't transfer as well to geekbench as to other workloads.

 

[jeanlain]

The light grey graph represents the laptop being plugged in,

To be fair, I'm not sure how they establish the curve, since that would require measuring a range of GPU power modes (which they could achieve by underclocking I suppose). I take the the far end of the curve represent actual measured performance/wattage when plugged in and in ideal conditions. The end of the darker part of the curve represent perf/watt on battery power. As for everything in between, I don't know, but that doesn't really matter.

 

[Thala]

To be fair, I'm not sure how they establish the curve, since that would require measuring a range of GPU power modes (which they could achieve by underclocking I suppose). I take the the far end of the curve represent actual measured performance/wattage when plugged in and in ideal conditions. The end of the darker part of the curve represent perf/watt on battery power. As for everything in between, I don't know, but that doesn't really matter.

You can model such a curve very precisely with just a few measurement samples. Apparently you are getting 2 samples by just plugging the device in and let it run on battery. I am sure you can get more samples when playing with the power settings.
In any case both of Apples claims refer to both above mentioned trivial measurement points - independent if the whole graph is 100% correct at any position in between.

 

[jeanlain]

You can model such a curve very precisely with just a few measurement samples. Apparently you are getting 2 samples by just plugging the device in and let it run on battery. I am sure you can get more samples when playing with the power settings.
In any case both of Apples claims refer to both above mentioned trivial measurement points - independent if the whole graph is 100% correct at any position.

I'd be nice if anandtech could generate these curves when comparing CPUs and GPUs. So far, Nuvia are the only ones I know which produced these kinds of results (and I wouldn't be surprised if they inspired Apple).

 

[Schmide]

Napkin numbers

xbox x
CPU ~2 tflop
GPU ~12 tflop
watts ~150-200

m1 max
CPU ~2+ tflop
GPU ~10 tflop
watts ~70

 

[igor_kavinski]

Napkin numbers

xbox x
CPU ~2 tflop
GPU ~12 tflop
watts ~150-200

m1 max
CPU ~2+ tflop
GPU ~10 tflop
watts ~70

Interesting. I foresee a future die shrink leading to 40% power savings for Xbox Series X.

 

[IvanKaramazov]

I’m not sure why there’s so much skepticism about the GPU claims. The M1 in Anandtech’s own testing averaged about 39fps in Rise of the Tomb Raider on 1080p Very High. A quick hunt suggests laptops packing the 3070 and 3080 mobile tend to average from 110-130 on the same benchmark. Apple explicitly calls out the Razer Blade Advanced with 3080 mobile, which is the review in the last link.

The M1 Max has 4x the GPU cores, about 4x the bandwidth, and much more unified RAM for the GPU cores to work with. With GPU scaling being relatively linear it’s not at all unreasonable to expect the new MBPs to end up with similar FPS to the Razer Blade Advanced in a real-world game like this, yes?

 

[trivik12]

So M1 Max does not have fastest single core performance. Will ADL-P beat it in Multi-core as well. AMD will also have 6 series Ryzen with better numbers. But can either Intel/AMD come anywhere in the ballpark when it comes to performance per watt. Laptop fairly thin with 21 hours battery life and this level of performance is still fantastic. I am tempted to buy one though I have little need for this much performance

 

[trivik12]

That's not their business plan. They simply want to own the store where all purchases include 30% Apple tax. And the way to achieve this is to monopolize the best hardware and fuse it with their optimized software. Why on earth would they share anything, when the end goal is to get everybody into the Apple store?

This is why they need to be broken up. Then they can truly maximize the potential in all their markets. As an apple stock holder I would be happier if they let it loose rather than keep it close to chest. They could make phenomenal datacenter chips of this as well.

 

[Heartbreaker]

Ian said on Twitter earlier the the Max variant was 10.4 TFLOPs, which is around 3060 levels of performance.

TFLOPS is a poor comparison metric for a GPU from different families. For gaming I'd expect it to do better than that 3060. Probably similar to a 3070 Mobile or even 3080 mobile, while using less power.

Plus hard to compare a NVidia GPU which doesn't run on Apple products,with an Apple M-chip that only runs on Apple. For GPU-compute, it's really hard to say, you end up comparing CUDA to Metal applications.