Apple M1 CPU

Jimmy Weirdarms · 17 Nov 2020 at 23:57

Cenuij said:
Yes quite true, I do recall though that IA64/EPIC was a better late than never hail mary against SPARC and DEC Alpha - I think it might have been the fastest Intel part at the time but the developer tooling and compilers were awful which meant it was d.o.a - I guess that's why they had to copy AMDs x86 64 instruction set.

The Itanic offloaded a lot, like decisions about parallelism, to the compiler and using it to full potential turned out to be hard. But TIL that they only got discontinued last year, with the last processor ("Kittson") only hitting the market in 2017. I assumed they were long dead!

And yeah, apparently when the opteron appeared and it was both 64-bit *and* back compatible with x86-32, intel had to respond with compatible xeons and the rest is history.

The Apple chip is looking epic. I want it in a laptop that can drive two external monitors. Guess I'm waiting a while...

Armageus · 18 Nov 2020 at 10:18

smilingcrow said:
They seem to be using LPDDR4X-4266 which is the same as Intel use in the current generation and possibly also AMD! It’s fast but it’s not that big a deal.

4266 is quite a big deal and at least explains why the ram upgrades cost so much.

Possibly the memory controller on the M1 just scales that much better?. Most Intel and AMD chips have tiny gains (~5-10%) even when moving between lower end memory and higher end memory.

I can't find any info as to what speed memory the A14/iPhone 12 uses other that it's also LPDDR4X

HACO · 18 Nov 2020 at 11:45

Armageus said:
4266 is quite a big deal and at least explains why the ram upgrades cost so much.

Possibly the memory controller on the M1 just scales that much better?. Most Intel and AMD chips have tiny gains (~5-10%) even when moving between lower end memory and higher end memory.

I can't find any info as to what speed memory the A14/iPhone 12 uses other that it's also LPDDR4X

Not sure I fully believe it, but this site says it's 2750MHz LPDDR5.

https://nanoreview.net/en/soc/apple-a14-bionic

0h_no_not_again · 18 Nov 2020 at 12:36

Donnie Fisher said:
Could that be because RISC has less possible branch prediction routes that it can go down, so with less options to choose from, it automatically becomes better at prediciton. ? By that I mean, does the ARM approach with a RISC instruction set simply have a lower number of possible next instructions compared to and x86 CISC approach ? ( I genuinely dont know )

I.e.
If ARM has an average of 5 possible branch choices, and x86 has 6 branch choices ... by the law of averages, the ARM will better it'll have a 20% chance of being right (at complete random choice) versus 16% for x86. ?

That is an excellent question! I don't believe the extra instructions available on x86 would have a sizable impact, as we're dealing with branch prediction and not the contents of the branches so much. So I would guess it is down to a superior cache and branch prediction.

I wish I could answer with certainty, but this goes beyond my knowledge of modern processors!

Spyhop · 18 Nov 2020 at 21:27

I think many (not all) gamer type people have a fear of the unknown when it comes to Macs. Because they've spent in some cases decades getting to know Windows exclusively, and know zero about MacOS concepts and how to use that OS effectively.

HACO · 20 Nov 2020 at 18:17

https://twitter.com/i/web/status/1329552544138485760

Pretty much ends the debate of whether it's the CPU or the browser that makes Apple devices (previously iOS) so fast at Javascript benchmarks. Spoiler: It's the CPU.

tommoT.Striker · 20 Nov 2020 at 20:01

Puget did some benchmarks for the M1
https://www.pugetsystems.com/labs/a...op-Workstation-for-Adobe-Creative-Cloud-1975/

Ok, so the Zen 3 CPUs are much faster, but the M1 is running the likes of photoshop using Rosetta apparently, thus the large performance gap. In any case would still feel quite snappy.

I'm getting a new Mac Book soon, Intel, the last of its kind, but still interested how ARM will runs Adobe apps. The native Xcode apparently flies when compiling.

HACO · 20 Nov 2020 at 20:41

tommoT.Striker said:
Puget did some benchmarks for the M1
https://www.pugetsystems.com/labs/a...op-Workstation-for-Adobe-Creative-Cloud-1975/

Ok, so the Zen 3 CPUs are much faster, but the M1 is running the likes of photoshop using Rosetta apparently, thus the large performance gap. In any case would still feel quite snappy.

I'm getting a new Mac Book soon, Intel, the last of its kind, but still interested how ARM will runs Adobe apps. The native Xcode apparently flies when compiling.

Strange comparison overall, zero mention of emulation, yet comparing Rosetta apps M1 in GPU accelerated tasks with their desktops with RTX 3080, but I guess that makes sense, they sell desktop PCs

They didn't run the M1 benchmarks themselves either.

The whole article was quite defensive on their parts, as if they needed to make the point that their desktop PCs are good. Did they not expect their Ryzen 5950X with RTX 3080 to come up on top in multithreaded and GPU heavy tasks?

tommoT.Striker · 20 Nov 2020 at 21:37

I do like their articles, their benchmarks are pretty good. They've been accused of biased this and that way, especially of being pro intel. But they're a business and they want you to buy their stuff!

4K8KW10 · 20 Nov 2020 at 22:50

HACO said:
Strange comparison overall, zero mention of emulation, yet comparing Rosetta apps M1 in GPU accelerated tasks with their desktops with RTX 3080, but I guess that makes sense, they sell desktop PCs They didn't run the M1 benchmarks themselves either.

The whole article was quite defensive on their parts, as if they needed to make the point that their desktop PCs are good. Did they not expect their Ryzen 5950X with RTX 3080 to come up on top in multithreaded and GPU heavy tasks?

Let's compare their crappy desktops with the Supercomputer Fugaku - Supercomputer Fugaku, A64FX 48C 2.2GHz, Tofu interconnect D, Fujitsu
RIKEN Center for Computational Science
Japan

:rolleyes:

deusexlumina · 21 Nov 2020 at 06:10

Eugh. I'm torn abut this one. On the one hand, the performance is very impressive. Even on that Puget review above, the M1 CPU sometimes hitting half the speed of the 5950X under emulation, with a significant core deficit, is nothing short of incredible.

On the other hand. Apple now have a node advantage and they packed a metric butt-ton of transistors into the SoC, so it *should* be good compared to e.g. intel 14 nm. We don't know what clock speeds these can reach and if there is a wall for the architecture, and we don't now what multi-core scaling will be like as they try to go for higher performance variants.

I'll keep my eye out for the M2 i guess. Interested to see where this goes.

Donnie Fisher · 21 Nov 2020 at 13:45

From the outset the is a common thought of it being a short first Gen product by a lot of reviews. So much so it's almost like that's been a leaked motion by Apple.

I agree the M2 will likely show where is heading ... Or what ever they use for the iMac

benparker · 21 Nov 2020 at 16:00

Do you think M2 will be releasing next year?

HACO · 22 Nov 2020 at 00:29

PostgreSQL benchmarks on M1 (8GB ram), compared to Intel macbooks:

https://info.crunchydata.com/blog/postgresql-benchmarks-apple-arm-m1-macbook-pro-2020

TLDR: To beat it, they had to bring in desktop Ryzen CPUs with 64GB of ram, lol.

pete910 · 22 Nov 2020 at 10:11

Does rather well in some test here, rosetta looks impressive. I presume rosetta is like wine for Linux ?

https://www.phoronix.com/scan.php?page=article&item=apple-mac-m1&num=1

Armageus · 22 Nov 2020 at 10:24

pete910 said:
Does rather well in some test here, rosetta looks impressive. I presume rosetta is like wine for Linux ?

Rosetta allows apple x86 apps to run on the arm M1 processor.

Wine is x86 windows on x86 Linux.

Similar but one is translating between the same os but different architecture, the other is translating different os calls on the same architecture

pete910 · 22 Nov 2020 at 12:16

Armageus said:
Rosetta allows apple x86 apps to run on the arm M1 processor.

Makes sense, I thought as much but wasn't entirely sure. Cheers Armageus.

muon · 23 Nov 2020 at 01:01

So this is the real deal then?

This single threaded performance on R23 is as good as Zen 3 (if not better clock for clock).

Blows my Zen 2 chip out of the water.

If they decided to use 8 big cores instead of big.litttle it would also be competitive at multicore.

HACO · 23 Nov 2020 at 12:24

muon said:
So this is the real deal then?

This single threaded performance on R23 is as good as Zen 3 (if not better clock for clock).

Blows my Zen 2 chip out of the water.

If they decided to use 8 big cores instead of big.litttle it would also be competitive at multicore.

Pretty much the real deal. Now that benchmarks are out, it's effectively handling everything people have thrown at i, and the performance results are very consistent. Slightly ahead of Zen 3 at FP, slightly behind at INT.

Bigger chips (8 Big, 4 Little) are very likely for larger Macbooks and a higher end Mac Mini, and I think we'll see an all-big core variant for at least the larger iMac (maybe 16 big?), with a dedicated GPU.

Grantelicious · 23 Nov 2020 at 12:24

Who cares? Apple will continue to lock their OS down until it is like IOS, so no one will want to use them. They'll continue to fall behind on the graphics side so they'll only ever be used for mobile games. I don't get why everyone is wetting their knickers over them, it only seems impressive because Intel has done nothing for 10 years.