Apple Silicon M1 – What to Expect, Based on A14 | Hacker News Reader

Apple Silicon M1 – What to Expect, Based on A14

anandtech.com

111 pointsGeekyBear37 comments

Apple clearly has an edge when it comes to silicon that others just won't be able to replicate anytime soon. From a cost perspective, just eliminating 4-5 chips by utilizing a System-on-a-Chip (SoC) allows them to create much simpler logic boards whilst saving hundreds of dollars. Not to mention saving the intel tax or being burdened by intel's lack of progress.

The big question remaining is how is this going to scale further up the mac pro line-up? I'd imagine a chiplet like architecture similar to AMDs zen3 is the way forward. But will they go all the way to 64 CPUs and beyond? Or will they just stick to a single SoC chip?

What cadence will they upgrade these chips with? I'd expect something closer to what the ipad has done with the AX chips, ie every 2 years. Questions.

Very interesting times. I want one, badly, just to have more than the meagre 2-4 hours of realistic batterly life i get on my 2017 MBP.

If this is the repeatable verifiable performance- Apple deserves kudos here. They’ve flipped over the table and created the new benchmark for performance and efficiency. I just wish the software side could keep up and start improving the developer experience.

From the comments:

"Firestorm is the fastest perf/W general computing uarch in the world & on the latest 5nm TSMC node, what ... else do people want?"

More than 16 GB RAM, more than 4 performance cores, and a discrete GPU.

They have only replaced lower-end Macbooks with 2 thunderbolt ports which they had also maximum of 16 GB of Ram before (I currently use 2020 model). They will absolutely increase the IO ports and RAM later but this is just a beginning of a new era.

The higher-end 13” Macbook Pro with 4 thunderbolts are still available.

No doubt there will be higher end designs in the future. The question is, how far will they go?

Will they build a discrete GPU, or continue their AMD partnership, or try to convince everyone that their integrated graphics are good enough? Will they build a massive multicore chip to challenge Threadripper? Will they go to higher TDP and push single core performance even more, or will they relax and just use medium TDP parts that still outperform Intel's best?

I'm excited to learn the answers to these questions in the next year or two.

>Will they build a discrete GPU, or continue their AMD partnership, or try to convince everyone that their integrated graphics are good enough?

Well, what if they are? There's no law/reason integrated graphics can't be, especially with significant pipeline changes and faster/nocopy access to memory...

There are several reasons actually. One is that GPUs have different memory tradeoffs than CPUs and prefer higher bandwidth at the cost of latency; unified memory is always a compromise, usually in favor of the CPU. Another is that top discrete GPUs are reticle limited, which means it's not possible to make the chip any bigger, so adding other SoC parts to the same chip would necessarily reduce the space dedicated to graphics. Yet another is that separating the GPU and CPU makes it easier to effectively cool both of them.

64GB in the core i7 Mac mini will be missed.

But these look like impressive upgrades for the Air and 13" MBP in terms of performance, heat, and battery life.

I just hope that the keyboard and trackpad work properly. ;-p

I sense a bit of a glass half empty vibe about the reaction to these machines.

All this will come in time - in the meantime this is a big step forward for machines of this class.

>I sense a bit of a glass half empty vibe about the reaction to these machines.

I think it's the classic tech nerd mindset ("No wireless. Less space than a Nomad. Lame.").

As if everybody and their dog had lower speced Airs with 2 4K monitors and 32GB before (as opposed to a small minority for the first and none for the second), and they will miss those with the new machine...

How about a higher TDP? Perf per W or J is dandy but absolute performance also has its place. How does this thing scale if you put a desktop-class cooler on it and how long will it be before we get to find out?

They seem to want to complete the transition in the next two years.

This could probably be merged with the existing discussion about the M1 chip, to not cause over-proliferation of threads. https://news.ycombinator.com/item?id=25049079

Personally I think M1's performance is big enough news to deserve a thread devoted to more hard data. Intel has fallen from 1st to 3rd place in single core performance in less than 1 month after holding the crown uncontested for a decade or more. It's a huge shift for the industry.

Replace an Anandtech deep dive containing hard data with a soft and fluffy piece based on a vendor presentation?

Where's the deep dive with hard data? This AnandTech article is a mashup of press release stenography and fiction.

Did you miss that the article has multiple pages?

Hmm, that's easy to miss - as easy as HN's multiple pages are. Here are the links:

https://www.anandtech.com/show/16226/apple-silicon-m1-a14-de...

Seems that only the last one was written after today's announcement though...

Have you considered adding a page selector at the top of comments when there are multiple pages as a feature of the site? Then you wouldn't have to post this on the articles with enough comments to necessitate this comment.

I didn't explain that well, but this doesn't seem like something someone should have to do manually.

Yes, but I don't want to do anything to make a temporary workaround more comfortable. See https://news.ycombinator.com/item?id=24042165

> Where's the deep dive with hard data?

>In the overall SPEC2006 chart, the A14 is performing absolutely fantastic, taking the lead in absolute performance only falling short of AMD’s recent Ryzen 5000 series.

The fact that Apple is able to achieve this in a total device power consumption of 5W including the SoC, DRAM, and regulators, versus +21W (1185G7) and 49W (5950X) package power figures, without DRAM or regulation, is absolutely mind-blowing.

https://www.anandtech.com/show/16226/apple-silicon-m1-a14-de...

There's not achieving that at 5W when running the benchmark though, more like 18W-20W. I question the obsession with perf/watt. It makes sense if you're running a data center, or if you need to do "light" work.

But almost everyone I know with a laptop spends the majority of the time working with it docked to power, and often a monitor. If you're using a laptop for work, and you're not running out of a coffee shop, do you want a Zen3 with 32-64Gb RAM, 8-16 cores and an RDNA2 graphics APU at 50W (Rembrandt), or do you want a a device that will play video for 18 hours?

Especially for gaming, Apple's GPU is still years behind AMD and NVidia. They're bragging about 22Gp/s fillrate when latest mobile discrete GPUs do 5-10x that.

Certainly, they have achieved a lot and are crushing Intel integrated chips at the low end, but there's a ton of marketing fluff and special case comparisons going on, and I look forward to seeing a real head-to-head matchup, of say a Zen3 based system against a MacBook on a heavy CPU and GPU bound benchmark like a triple-AAA title.

Years of synthetic benchmarks that exercise giant CPU caches leaves much to be desired.

Perf/watt matters even if you don't care about power consumption at all, because heat dissipation is usually the limiting factor for sustained performance. Higher perf/watt means higher perf for any given amount of cooling.

Sure, but people with high perf laptops have high-perf cooling solutions. I'm an engineer/developer and a high end gamer. I have an Alienware laptop (in addition to a Macbook Pro), with an RTX 2070 in it. It never overheats. I take it traveling with me. The M1 has a fillrate 10x slower, and I highly doubt it's competitive on Compute Shader performance either. Even if my RTX laptop were thermally down clocked, it's still got plenty of headroom over the M1 GPU wise.

The only time I worry about heat is if I have to have the laptop in my lap.

On top of all this, AMD's 6nm Rembrandt coming out next year is likely to have a model in the 20W range, which is on par with the M1's "performance" mode, but with much greater GPU performance.

I think the epitaph being written for x86 is a little premature, maybe for Intel, but AMD just delivered 20% IPC improvement, and a huge gfx architecture improvement, with a scalable chiplet architecture going forward.

It would be unwise to count them out.

The worst possible thing that can happen if you're using a laptop is for work is run out of battery (and no carrying a power brick around everywhere isn't great). Lots of people value a small form factor and no fan. Do you need 16 cores to run Word and Outlook? Probably not.

Apple are clearly making trade offs and I suspect that gamers aren't a priority - for lots and lots of people having a laptop with this performance and power efficiency is going to be a compelling proposition.

> Do you need 16 cores to run Word and Outlook

No, but for that, I don't need much more than a Chromebook either. Are we really saying an Surface/Surface Pro can't run office for 99% of the use cases and workloads people use it for?

> if you're using a laptop is for work is run out of battery

Majority of people I know how have Macbooks have them plugged in when at the office or at home. And many offices have power-bricks in conference rooms pre-installed, especially with USB-C charging, your laptop can charge just by plugging it into your monitor, or in any standard USB-C port in a conference room.

Unless your office is Starbucks, let's face it, for most of corporate America, and home work during COVID, people are plugged in a majority of the time.

Yes I am saying that a Surface / Surface Pro is fine for most people - and that given the option of adding 8 cores or 8 hours of battery life most people will take the battery life.

Saying that you'll probably have a power supply nearby is fine - except for the cases when you don't - I can't count the cases when I've seen people scrambling to find power in a meeting and it's a pain and embarassing. People will take avoiding that over cores they don't need.

> Apple are clearly making trade offs and I suspect that gamers aren't a priority

Baldur's Gate 3 demos notwithstanding, Apple seems to have little interest in PC gaming, which is usually Windows/DirectX first with macOS being a distant afterthought or delayed, overpriced port. They also killed off 32-bit games, which further reduced the PC game landscape on macOS.

However, Apple has its own massive ecosystem of game devs on iOS, and it also has Apple Arcade.

If a PC game isn't on iOS, you may have to rely on cloud gaming if you want to play it on macOS in the future - which may not be much worse than the current situation of mediocre ports.

Apple is achieving that at 5W actually. 18/20W would totally blow up the phone. :)

Unless I'm missing something, those are estimated, extrapolated values. Chart title says "SPECint2006 Speed Estimated Scores".

A14 isn't competing with Intel/AMD at 5W.

They're not extrapolated; that's the actual performance of an iPhone 12. The "estimate" stuff is legalese.

Hopefully when they release these things, Anandtech can hook up an actual power meter, run the benchmarks, as well as some real software workloads (e.g. compilers, triple-A games, etc) and measure actual perf vs drawn power vs throttling.

It's one thing to run a benchmark when your device is cold, it's another to ask what the performance w/ passive cooling will be life after an hour of strenuous work.

The review site NotebookCheck has done a good service of consistently recording steady-state CPU performance of mobile devices, including in their latest MacBook Pro 2020 review, so I imagine they will have that data soon.

I do 90% of my work in coffee shops lol

I get that this was hastily written but I'm not even halfway done and I've already hit a handful of typos/grammar mistakes.

> Today’s event contained a whole ton of new official announcement, but also lacked (in typical Apple fashion) in detail.

"any detail" is probably what they meant

> This packaging style with DRAM embedded within the organic packaging is something that isn’t inherent new to the new M1 chip as we’ve already seen such a design featured on the A12X.

s/inherent/inherently

> We can see the new M1’s four Firestorm CPU cores on the left side. Notice the extremely large amount of cache – this was one of the surprises reveals of the event as the A14 still only features 8MB of L2 cache.

"this was one of the surprises reveals" not sure what they were going for, probably just remove "reveals" and it works?

Also they talk about the A14 in the title but then barely touch on it at all. It's like 3 sentences about it then a recap of Apple's last transition.

EDIT: I now see that there are multiple pages. I really don't like it when webpages do this and I like it even less when it's not clear at all that there are multiple pages. The title for the next "section" is 20px tall and not clear that it's a link to go to the next section (No "Next", arrow, etc) and the dropdown for switching between pages (this has got to be the worst way to do pagination IMHO) is 16px tall...

Note that the article has multiple pages. The really interesting stuff is on page 4, not page 1: https://www.anandtech.com/show/16226/apple-silicon-m1-a14-de...

Reply on news.ycombinator.com