Apple would then be 1 die shrink ahead of everyone else at either 4nm or 3nm by now.If you look at the OpenFOAM CFD ranking you'll notice a trend with CPU cache size being king. 768MB > 256MB > 128MB > 64MB. M1 Ultra has yet to make it on the list but it has 48MB. Maybe someone will take a break from Geekbench and run OpenFOAM for MacOS.
https://openbenchmarking.org/test/pts/openfoam&eval=959768c2bfe5424e10ba636ac9e0a8759326f7a1#metrics
Lower time is better
1x 7773X 64-core (768MB) 120 secs
1x 7763 64-core (256MB) 206 secs
1x 7532 32-core (256MB) 238 secs
1x 7542 32-core (128MB) 321 secs
1x 7601 32-core (64MB) 331 secs
Upcoming Epyc Genoa-X will have 1GB+ cache and 96-cores. That would make for a nice single-socket desktop supercomputer.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
[CPU only] Apple M1/2(Max/Ultra) (TSMC 5nm) vs AMD Zen 4 (TSMC 5nm) - Technical Analysis
- Thread starter exoticSpice
- Start date
- Sort by reaction score
Chips and Cheese has written two posts about Zen4:
chipsandcheese.com
chipsandcheese.com
- frontend and execution engine
AMD’s Zen 4 Part 1: Frontend and Execution Engine
AMD’s Zen 4 architecture has been hotly anticipated by many in the tech sphere; as a result many rumors were floating around about its performance gains prior to its release. In February 2021…
chipsandcheese.com
- memory
AMD’s Zen 4, Part 2: Memory Subsystem and Conclusion
Please go through part 1 of our Zen 4 coverage, if you haven’t done so already. This article picks up where the previous one left off. To summarize, Zen 4 has made several moves in the fronte…
chipsandcheese.com
Last edited:
It is true that Apple Silicon is already a very wide architecture and we don't know whether making it any wider will bring any noticeable benefits. That said, nobody even comes close to building CPUs as wide, which probably means that Apple has some sort of method of utilising the ILP that might allow them to scale even further. At any rate, the future will show.
As to the rest of the argument... no, M1 didn't go the AMD route — it has only four performance cores while AMD has eight. The primary purpose of the E-cores in Apple Silicon is to improve energy efficiency and take the load off the main cores, and not to provide sustained performance like in Intel designs. And beyond that, Apple's main core is ridiculously efficient compared to Intel and AMD. They need 5-6 watts to provide performance at a level where others need 15 watts or more. So I'd say that Apple can still afford ramping up the clocks without losing the efficiency crown (of course, power consumption will go up as they do it).
Right now we have three main strategies:
1. AMD has cores that scale well from very low power consumption to very high power consumption. They can achieve high burst performance by clocking the cores very high (at the expense of efficiency) and they can achieve high throughput by clocking the cores low and having many of them
2. Intel has very fast (and power hungry cores) and slower (but more efficient cores), so they combine both. The P-cores are there for burst performance, and the E-cores significantly contribute to throughput when power is constrained
3. Apple has very efficient P-cores with a more conservative narrow dynamic clock range, which can provide good (but not industry-breaking) burst performance at very low power as well as excellent sustained performance. In addition, Apple has extremely low-power E-cores which are still decently fast, and uses those to run low-priority tasks (which can be used to either improve the power consumption or the performance of the main cores, or to give a bit extra sustained performance)
We'll have to wait and see where Apple goes from here. One thing is clear of course, they can't just rest on their laurels. M1 CPU architecture is incredible — it was released more than two years ago and still remains unsurpassed in key indicators. But it is a question of time until a new better architecture appears.
The opinion does not have merit. The figures are correct, the interpretation is foolish.
The primary implication, that Apple cannot scale beyond 8-wide (and more generally, can't increase IPC) is invalid. There remain a number of practical ways available to boost IPC (and to go beyond 8-wide).
The main thing that has happened, I think, is that schedules have, not exactly slipped, but had to be adjusted to various concerns.
The A15, as far as CPU goes, was mainly a substantially more energy efficient version of the A14. On the one hand, the very best designers had been working on the M1 rather than the A15; on the other hand, the team of future designers were undoubtedly working for the A16 CPU on a design targeting N3. But then N3 slipped schedule, not drastically but enough that Apple couldn't risk pinning the iPhone launch to it, and the A16 hd to follow the backup plan of a slightly improved A15, mainly taking advantage of N4 rather than N5 to slightly boost some cache sizes and frequency.
Both of these were not "Apple runs out of steam"; they were adaptions to new circumstances. The Mac silicon launch was a one time event, but it came pretty much along with both covid and N3 being a quarter later than expected, and so what should have been a one year slowdown in the *performance* improvements of iPhone (energy was substantially improved by A15) turned into two years of non-impressive performance improvements. But none of this reflects (IMHO) a fundamental inability to boost IPC (or to improve the CPU and the SoC in other ways); it's just a perfect storm of bad luck. In a way covid was probably responsible for all of it; everything slowed down because of covid, but nothing derailed. (My suspicion, when all is said and done, is that even N3 slipping by three months will ultimately be learned to be the result of covid; interaction between TSMC and partners in Europe or Japan was forced to slow down because people couldn't travel freely, and a week delay here, a week delay there, and eventually you're off by three months.)
There are two points to make:
- can you make Apple's cores wider without hitting a fundamental limit. Yes. The most difficult part right now is the Rename stage, but you can do a lot to make the core "effectively" wider while keeping Rename 8-wide.
- will making the core wider get you higher IPC? It is true that blindly adding resources does not boost IPC much. This is why I'm not especially excited by the claims being made for the next Intel core which are presented as a list of ROB now has this size, physical register file now has that size, and so on.
But I said *blindly adding*, that is adding new resources without changing the algorithms that utilize those resources. Apple has been very aggressive in changing these algorithms as their cores have grown, and there is no reason to believe this will change. The main thing is that each time you change these algorithms you usually need more transistors. Hence my point already made – the "real" next design was queued up for N3, but with N3 unavailable that design had to be shelved for now.
The most interest question going forward is whether the next round of M2 Pro and Max use the M2 CPUs (ie the A15 Blizzard and Avalanche) or whether they are on N4 and use the A16 CPUs (ie Sawtooth and Everest, which are mostly like the A15 CPUs but slightly tweaked for N4) or whether M2 Pro and Max are actually on N3 and use the (as yet unreleased) next generation CPU design...
Has any company scaled beyond 8-wide? Is there any limit of how wide a pipeline can be?
Thanks Maynard! Are you aware of any newer Apple parents that could foreshadow what they might be planning for future CPUs?
What people haven't realized yet is that we don't need a new architecture. All we need is a power refinement to the existing ones. If AMD manages to have 12-hour battery life with comparable (or better) performance , does it really matter it's not ARM?
In fact, I would say that a slightly worse power-efficient AMD would still be in favor of X86 / X64, considering the library of applications is so much larger.
If Apple wants to have a more serious chance and not reply on always having a processing advantage (because it'll probably eventually vanish), they need to either invest more heavily in virtualization and be more permissive with it, or find new ways to play along with existing programs. E.g, by creating an OpelGL -> Metal compatibility layer, or maybe by integrating WINE / Rosetta in the OS. But that step is probably too "extreme" for their current corporate mindset.
I don’t follow this. If you are talking about the availability of existing applications, you are talking about Windows-compatibility and not x86 vs. ARM. The ability to run Windows programs was always a small niche for Macs and they did just fine. I would even argue that nowadays windows-compatibility is much less important than just few years ago.
Apple built the Mac on the principle that it’s different from other PCs. That’s also why Macs are successful. Focusing on emulating other OSes would be a sure way to quickly kill the Mac altogether.
UTM Windows 11 is very good. I feel that I don't need a Windows laptop to run Windows programs anymore. Parallels is even better though I don't need strong graphics performance. VMware and VirtualBox are both essentially in beta; so if you want or need Windows; it's available with some limitations. But probably good enough for most that need to run a couple of Windows programs.
Exactly. If it’s an occasional legacy windows app one needs, Apple Silicon with Parallels can probably run it just fine. If it’s something much more complex, you want a Windows PC anyway. There is no incentive whatsoever for Apple to cater to Windows users (aside the lackluster attempts to sell them services)
That was when Macs had a clear edge over Windows. Sure, it was on a narrow field (audio and video editing); but now, that advantage has narrowed.
What can you do with a Mac that you can't do with a PC? With e.g, a handheld / portable AMD setup, you are getting a similar performance, slightly worse battery life, but a much more versatile system. You have more games and tools available, and don't risk Apple forcing your system to be obsolete (e.g, due to phasing out the browser, or blocking access to the store because your system is too old).
The advantage indeed has narrowed in the past as average quality of laptops has improved and Apple was not the sole provider of exclusive high-end laptops.
Which is why Apple started making their own chips.
Higher performance and better battery life are two main things I’d say. As to versatility, depends what you do I suppose. MacOS is much more versatile for my work.
These are not "things you can do". Performance is not higher either, since if you are willing to sacrifice power efficiency, there are X86 / ARM setups that far surpass Apple setups.
Tell us why your what your machine can do that a X86 machine cannot.
In a compact laptop form factor? Which setups are those? I mean, my 2.2kg Apple laptop fits Bayesian models and builds software almost as fast as a large workstation, and I can go an entire day on a single charge. There is nothing even remotely comparable from any other laptop maker.
I don't understand the question. What can an x86 machine do that an ARM Mac cannot? It's not like there is a fundamental capability discrepancy no matter which platform you look at. Sure, available software and application interfaces differ, but that's hardly an area that promotes constructive discussion.
If you are instead asking why macOS is more versatile for me, well, it's because I happen to like how macOS does things. It offers a bunch of little tools and features that I use daily that other systems lack, and it has a good selection of polished, high-quality software. It also integrates very well with various open-source tools, given that it's one of the primary platforms for developing and using FOSS. Add to this my comments about the hardware platform (workstation-level performance in my domain in a compact laptop form factor and excellent battery life), and it should be easy to follow why I much prefer a modern Mac both privately and professionally.
All you have told us is, "I prefer Macbooks because they have a better battery life." Which is fine. But a "versatile" system means it can easily be adapted to many cases. You haven't told us how it is more versatile for you; just that it has better battery life.
In fact, you have told us the opposite:
You can easily match the extra battery life of a Macbook with a power bank. It's a bit clumsy, but you'll be able to have access to everything a X86 / X64 / AMD machine offers. And I can list quite a lot of stuff that it can do that a Macbook cannot, starting from the copyright nightmare games EA offers that you can't easily run on a VM with a Macbook.
The reasons I like my 2021 MacBook Pro:
- I can use it on a blanket or carpet. I'm using it on a blanket right now.
- No fan noise (have never heard it in a year of use). That's nice on Zoom calls and doing video recordings.
- Speakers:
- Trackpad (I stopped carrying an external mouse after getting the 2021 MacBook Pro)
- Battery life
- Efficiency for video editing
- Magsafe
- Sturdiness (it's a chungus)
- Keyboard (I prefer mechanical but this is nice on a laptop. It is clicky, firm, and not mushy side to side)
Things that I'd like Apple to improve on:
- Get rid of the notch
- Another USB port, preferably USB-A (not going to happen)
- Lighted Apple logo on back (not going to happen)
My previous laptop was the 2015 MacBook Pro and the 2021 was the first that I deemed good enough to upgrade to. The improvements are quite remarkable - so many for one release; and I haven't had any problems with mine that were hardware-related.
I am firmly in the Apple Ecosystem and live off of iCloud Notes. You can run it in a web browser, and I tried doing that for a couple of months, but it's an inefficient way to run it as your browser downloads all of your notes when you restart it.
The definition of "versatility" will change depending on which criteria one uses. That's a stupid game to play. That's why I am talking about versatility for me specifically. What I need in a computer is that it's fast at building and testing software, fast at running Stan models and fast at running code that transforms and manipulates data. ARM Macs happen excel at this thanks to their hardware architecture. MacOS also offers a much of additional features (in regards to available software, automatic, integration and file management) that I find particularly convenient and that massively improve my workflows. Finally, with an ARM Mac I can have all of this in a mobile compact package that I can use either untethered or at the desk. Can't do this with a Windows laptop that drops the performance by up to 50% when you disconnect the charger.
All of this shouldn't be that difficult to follow, right? Your needs might be different. If you care about running EA games (I absolutely do not), then it's clear that a Mac is not a suitable tool for you.
And sure, I could use a powerbank and a Windows laptop. Or carry a diesel generator with me. I could also use a RaspberryPi and NixOS. Or a Threadripper workstation that I would wheel around in a cart. I could also quit my job and become a farmer. But I don't want to do any of these things. Apple Silicon gives me what I need with very little compromises.
Depending on how much you want to spend, you can find X86 laptops today with up to 15 hours battery life.
And if you want all the power, but still lots of battery life, you could purchase something fits in the palm of your hand, but that will give up you up 25 extra hours of battery life. It's not an extremely large compromise.
But they won't have the performance I need. I need both. In the x86 world I can get either mobility (and I said mobility, not portability!!) or performance. Not both at the same time. And then there is macOS, which — again — is more convenient for my workloads.
You could get a GPD or Steam Deck, which are handheld PCs, and simply get a power bank. GPD is very powerful, by the way. It comes with a Ryzen 7 processor and up to 32 GB RAM. Most AAA games will run buttery smooth. And if you want an external monitor, you can connect it to your TV / Macbook / iPad.
And since it is so powerful, it doesn't run a "lite" OS with odd limitations. It runs full-blown Windows 10, meaning you could run more intensive applications in it.
You seem very focused on games. Most Mac users care little for games or they would already be using a different platform (or a second platform).
You misunderstand.
While I do like games, the hardware in those systems isn't good just for games. You could definitely run audio and video production or scientific applications there if you wanted to, and even connect an eGPU if your heart feels like it (and if you have the money to shed!). They're that powerful.
If you want to game, get a Windows or Console. I play online chess for gaming. Any computer in the last 10 years is fine. Any computer with a high-end CPU in the past 15 years is fine.