Apple Silicon M1 Chip In MacBook Air Outperforms High-End 16-Inch MacBook Pro (macrumors.com) 174
The first benchmark of Apple's M1 chip shows that the multi-core performance of the new MacBook Air with 8GB RAM beats out all of the 2019 16-inch MacBook Pro models, including the 10th-generation high-end 2.4GHz Intel Core i9 model. "That high-end 16-inch MacBook Pro earned a single-core score of 1096 and a multi-core score of 6870," reports MacRumors. The MacBook Air with M1 chip and 8GB RAM features a single-core score of 1687 and a multi-core score of 7433. From the report: Though the M1 chip is outperforming the 16-inch MacBook Pro models when it comes to raw CPU benchmarks, the 16-inch MacBook Pro likely offers better performance in other areas such as the GPU as those models have high-power discrete GPUs. It's worth noting that there are likely to be some performance differences between the MacBook Pro and the "MacBook Air" even though they're using the same M1 chip because the "MacBook Air" has a fanless design and the MacBook Pro has an new Apple-designed cooling system. There's also a benchmark for the Mac mini, though, and it has about the same scores. The "Mac mini" with M1 chip that was benchmarked earned a single-core score of 1682 and a multi-core score of 7067.
There's also a benchmark for the 13-inch MacBook Pro with M1 chip and 16GB RAM that has a single-core score of 1714 and a multi-core score of 6802. Like the "MacBook Air," it has a 3.2GHz base frequency. A few other "MacBook Air" benchmarks have surfaced too with similar scores, and the full list is available on Geekbench. [...] When compared to existing devices, the M1 chip in the "MacBook Air" outperforms all iOS devices. For comparison's sake, the iPhone 12 Pro earned a single-core score of 1584 and a multi-core score of 3898, while the highest ranked iOS device on Geekbench's charts, the A14 iPad Air, earned a single-core score of 1585 and a multi-core score of 4647.
There's also a benchmark for the 13-inch MacBook Pro with M1 chip and 16GB RAM that has a single-core score of 1714 and a multi-core score of 6802. Like the "MacBook Air," it has a 3.2GHz base frequency. A few other "MacBook Air" benchmarks have surfaced too with similar scores, and the full list is available on Geekbench. [...] When compared to existing devices, the M1 chip in the "MacBook Air" outperforms all iOS devices. For comparison's sake, the iPhone 12 Pro earned a single-core score of 1584 and a multi-core score of 3898, while the highest ranked iOS device on Geekbench's charts, the A14 iPad Air, earned a single-core score of 1585 and a multi-core score of 4647.
Why only Geekbench? (Score:3, Interesting)
Every single performance comparison I've seen of M1 versus other CPUs uses Geekbench, either in its single-core or all-core (multi-core) version. This gives an extremely limited basis for comparison. Even when AMD is comparing itself to Intel, while it uses Cinebench as the top-line comparison, it generally provides comparison on other benchmarks as well. Why is Apple focusing so much on Geekbench?
Re: Why only Geekbench? (Score:2, Insightful)
Because that's how benchmarks for unreleased hardware works. A highly selective test that is cherry picked for most hype value.
There's no way the real world performance of an ARM system will beat out a high end x86 machine in everyday use with normal patterns of use.
Now, the pattern of use for an Apple user on the other hand... Those guys will probably buy these devices to do nothing but sit in Starbucks and show running Geekbench instances to random passers by.
Re: Why only Geekbench? (Score:3, Funny)
Geekbench is a standard. Iâ(TM)m sure others will surface. The thing ships out this month and you can run any test you want once they sit on the tables of your local Apple store. If the mobile space is anything to note, the eggheads at the Apple acquired PA Semi will have and continue to outperform their colleagues at Intel, which is probably why Apple decided to switch. The iPad has been getting faster and faster, catching up to the MacBook Pros, and for less heat and power. Apple have
Re: (Score:2)
Using a bad standard to measure performance of your CPU leads to a bad CPU.
Re: (Score:2)
Do you really think Apple's engineering team used Geekbench to judge its CPU? I am sure that they have their own internal benchmarks that are weighted towards the metrics that they think are important. It may be that Apple's marketing team uses it as it shows the CPU in good light, but my bet is that the design decisions were based on other criteria.
Re: Why only Geekbench? (Score:2)
Re: (Score:2)
Wasn't that the obvious implication of how I asked the question? I was hoping for someone to give an iron man approximation of Apple's logic, not just assume bad faith on Apple's part.
Re: (Score:3)
Your failure is assuming that Apple is behind it at all.
Re: (Score:2)
I have a Mac Mini 2012 (yep ... one of the abandon machines with Big Sur). It has an i7 and now it runs with Catalina and 10GB RAM (because one of its 8GB memory sticks failed and I had a 2GB around). And generally speaking it runs well.
However, my work is on Linux, so I have Linux virtual machines on that Mac, and a bunch of ARM based devices with a lot of different fruit logos printed on them.
What I think is that there is not unique working landscape that could define every possible usage. The M1
Re: (Score:3)
The M1 doesn't target a "high end x86 machine", it targets a low power notebook. Also, using Geekbench isn't "cherry picking", it's just basic and it's not Apple generating "hype value".
Re: Why only Geekbench? (Score:3, Interesting)
They're putting it in MacBook Pros. The laptop for power users. Well, what counts as a power user in the Apple ecosystem anyway.
Re: (Score:3)
You weren't expecting them to use video game benchmarks to show off their shiny new Macbooks, did you? The integrated GPU on the M1 processor would get thoroughly trounced by any similarly priced Windows laptop or desktop with a dedicated GPU.
Re: (Score:3)
Untrue. The way-less-speced GPU in the DTK is very, very fast and handles loads of work without even getting warm. M1 should be at least twice that.
Can you whip out a $1000 Nvidia card and beat it? Sure.
That's not the market for the M1.
Re: (Score:2)
How well you think it would do against a Ryzen 3400g?
Re: (Score:2)
I would imagine that M1 will beat 3400g in both CPU and GPU benchmarks.
Re: (Score:3)
I would expect the $150 GeForce 1660 from a comparably priced Windows gaming laptop or desktop to thoroughly trounce the gaming benchmarks of the new 13" MacBook Pro with an M1. That's not a fair fight, but then neither is benchmarking your new product with prerelease software and plugins that were specially designed with Apple optimizations against their 2-year-old hardware.
Re: (Score:2)
It would be incredible if the GPU in these things is at all comparable to similarly priced x86 laptops which will have a dedicated chip rather than an integrated one.
My guess is that Apple will have optimized it for video processing to keep video rendering decent, since a lot of low tier reviewers use Macs for that, but for games... eSports maybe.
Re: (Score:3, Interesting)
Benchmark are irrelevant.
Relevant is user experience.
I mostly play WoW Classic and eve online. It is basically impossible to build a laptop that does not run both in absurd high quality.
For ordinary day to day use, I use Mail, Safari, Chrome. And: Ecplipse. Seriously, who the funk cares if a windows laptop is "objectively faster" in a benchmark, when my Mac simply does what I want and does it so fast (e.g. compiling a huge Java project) that I can not even go a way to take a legit coffee break or write one
Re: Why only Geekbench? (Score:5, Informative)
Re: (Score:3, Interesting)
It's the corporate environment aspect. Corporate IT spent decades piling remote management software onto their Windows infrastructure boot processes. By the time everything loads and connects to the central server, it does feel like 30 minutes. Now what did these same people do when Apple hardware came along? Nothing... those machines are allowed to run without all of the boot-time crap because corporate IT only knows Windows and doesn't know how to "manage" Apple products.
Re: (Score:3)
While my corporate Mac that I use now has some shim shit installed on it, it is far less obnoxious than what was layered on in Windows 10.
Re: (Score:2)
I use my computer for many things, some of which are directly and discernibly dependent on CPU speed.
I expressly benefit from single core speed and multi-threaded performance.
The benchmark is irrelevant but user experience does matter, and your comparison between Mac and Windows fails miserably to really capture that.
For instance, my Windows tablet is 'press a button and it's on' fast, and boots in under 10 seconds when I restart it from scratch. Your Mac book will, if it complies with "company policies", t
Re: (Score:2)
Most "fast" windows laptop take up to half an hour to boot, because of "company policies" - security updates, scans, encrypted hard disks, slow connection to the AD service, or what ever reason.
Even my work laptop, which is about five years old, boots in less than a minute with all the myriad agents and things installed, and McAfee's pre-boot encryption prompt. My personal laptop boots faster than my external monitor can detect the signal and power on.
Re: (Score:2)
I am curious to know why you think that there isn't any way an ARM system can do better. Is there something fundamental that you see in its architecture that will prevent it from beating x86? I realize that all the ARM examples so far haven't done so, but that is primarily because they've been focused on low power applications rather than high performance.
Apple will be at a bit of a disadvantage as existing applications will be heavily optimized to run x86 rather than tuned for ARM (e.g. performance-sensiti
Re: (Score:2)
I don't think Apple would have "pulled the trigger" on the transition to ARM without having high confidence that they could do well against Intel. It isn't as if they were forced to make a transition. They are also hedging their bets - the higher-end will still be Intel for a while. If Intel keeps up, this might wind up being forever.
Apple is the one company that likely has both the ability, motivation and means to compete. The volume of CPUs that they ship is enormous (iPhone + iPad and now Mac) and this
Re: (Score:2)
Re:Why only Geekbench? (Score:4)
Maybe. It seems like the Geekbench scores are not very comparable across platforms. For example, the iPhone 12 Pro [geekbench.com] gets the same single-core score as the Ryzen 5 5600X [geekbench.com]. I find it difficult to believe that those two processors are equivalent for any compute-heavy task.
Re: Why only Geekbench? (Score:3)
Re: (Score:3)
I wasn't comparing the Ryzen to the M1, but to the A14. The simple physics don't make sense: How does a cell phone processor legitimately run as fast as a desktop processor and its roughly 20 times higher power budget? The differences in decoding instructions and register set size can't bridge that gap, unless you assume that AMD and Intel are secretly terrible at CPU design.
Re: Why only Geekbench? (Score:2)
Does a higher power budget make the clock faster than what it is, decrease cycles per instruction, decrease latency to caches and memory? No.
I think it does let you put more stuff on the chip, more cache, more parallelism, more bandwidth, those can all affect performance, but it depends on the workload right?
Power budget is like how much gas your car uses, when you should be thinking weight and power. Are you putting that energy to good effect or not is what it comes down to.
Re: (Score:2)
Are you saying that the A14 is like a go-kart compared to the Ryzen's heavy-duty pickup truck, and Geekbench simply measures the turning radius or top speed, making it a good basis for comparison?
I guess I'll accept that as Apple's argument.
Re: (Score:2)
I think Apple is better than you think at creating its CPUs. They create fast processors and the improvements have been more rapid than in the Intel world (mainly because they started off so far behind). The main difficulty you will find with a phone processor is that it cannot sustain the performance. So burst (as in a benchmark) it can do well. But because of cooling issues, it needs to be throttled after a while.
Re: (Score:2)
I think Apple is better than you think at creating its CPUs.
So let me get this straight, you believe that Apple licensed ARM and produced an ARM chip with a 53% IPC boost over AMD's wildly praised Zen architecture, and a 65% IPC boost over Intel's Gen10?
*breaths in deeply* bwhahahahahahahahahahahahaha
Re: (Score:2)
It's because Geekbench tests are highly artificial and focused heavily on the CPU core, so for example the massive amount of cache that a Ryzen 5600X has doesn't benefit those tests much. Also Geekbench is highly compiler dependent, i.e. the better the code that the compiler emits and the better tuned it is for a particular CPU the better it will run on that platform.
It's going to be interesting to see the excuses when real benchmarks appear showing actual application performance.
Re: (Score:2)
How long does the Cinebench test run? We're talking about the Air here, so they might have skipped that benchmark for thermal reasons.
Re: (Score:2)
Re: Why only Geekbench? (Score:3)
Itâ(TM)s a combination of commonly used synthetics like n-bodies and real world use cases like compiling with clang and manipulating images.
Anandtech have found that itâ(TM)s results correlate pretty strongly with SPEC results, so itâ(TM)s a pretty good benchmark.
Re: (Score:2)
Re: (Score:2)
Who says it is Apple behind these initial tests? Geekbench is well known and easy to run.
Re: (Score:2)
Apple has repeatedly referenced Geekbench. Maybe someone else is violating their NDA and releasing M1 benchmarks early, but I would think any third party would be more interested in a benchmark different from what Apple has used for its comparisons.
Re:Why only Geekbench? (Score:4, Informative)
Re: (Score:2)
Nobody seriously reviewing CPUs uses Geekbench. Its results do not translate to real-world performance. For some reason it's taken hold in the mobile space, probably because real application and game benchmarks are difficult to do properly on mobile devices and because it tends to favour Apple.
Also the comparison here is flawed. The Macbook Pro suffers from severe thermal throttling. When running sustained multi-core loads the Intel CPU hits 99C and throttles to protect itself. A comparison with a better de
Re: (Score:2)
Re: Why only Geekbench? (Score:5, Informative)
Because none of the other benchmark suites have released binaries for the macOS/ARM platform yet.
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
https://developer.apple.com/do... [apple.com]
Re: (Score:2)
Re: (Score:2)
This is the same company that when making the transition from PowerPC to Intel touted on their website you should upgrade because its faster. While on the same webpage touting that you should by a Mac because PowerPC is so much faster than their Intel based competitors...
Re: Why only Geekbench? (Score:2)
Consumer Gear (Score:2)
Re: (Score:2)
Also, you can't run Docker.
Re: Consumer Gear (Score:2)
Rosetta 2
Re: (Score:3)
Rosetta 2 is closed source and no one has actually used it yet, so there is a lot to still look forward to here RE compatibility - we simply don't know.
Frankly, if Apple has developed an efficient way to transpile X86 binaries to run on ARM, I am surprised they are not trying to monetize that outside OSX. A lot of people would buy that for Android and AWS.
Re: (Score:3)
Anyone that has the DTK has been using Rosetta 2. Works quite well. Just download random Mac App's and they work (i.e. Intel based ones that have not been released as Intel/Arm binaries.)
Re: (Score:2)
Rosetta 2
You're saying run Docker using Rosetta 2?
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re:Consumer Gear (Score:5, Insightful)
With 16GB, it would be paging.
So what does your kid do that requires 32GB?
Discord, Twitch, Fortnite, Youtube, and 47 tabs of Chrome. All at once. That's what kids do on a PC.
I just upgraded my nephew to 32GB because his machine would grind to a halt every time it tried to flush a few gigs to disk.
Heck if people close Chrome, most people I know don’t need more than 8GB.
I don't know who you know, but they sure as hell aren't the people I know.
And as for gaming- of course. A game doesn't recommend 32GB, when it commits 8GB tops. But games aren't the only thing a PC does, and they're rarely the only thing running.
Re: (Score:2)
Re: (Score:2)
I don't know who you know, but they sure as hell aren't the people I know.
It sounds like you know computer nerds. 8GB is still very much the stock standard professional grade PC that is selling by the 100s of thousands to corporations around the world. It's also very much the standard laptop size. It is right now the most sold size for PCs.
To be clear its not enough for me either, but then I do more than just open a couple of tabs at a time. The girlfriend is in the next room editing her DSLR photos on her 8GB machine, and that's already more of a workload than most people stress
Re: (Score:2)
Of course, that also includes god knows what else... probably an HP shitware printer app running, etc.
Re: (Score:2)
I can't speak for anyone else but I'm finding now that 16GB isn't enough for gaming. I can just barely load Rust in 16GB, for example. And I can't run the browser at the same time. For a desktop box including gaming, I now want 32GB. That seems egregious to me too, but it is what it is.
Re: (Score:2)
Re: (Score:2)
Re: Consumer Gear (Score:5, Insightful)
My work laptop has 32GB right now and I'm a software engineer. If you use docker, plus 8 copies of chromium (vscode, teams or slack, etc), plus 2 browsers, plus your IDE, you need at least 16GB. That's the low end. I'm using 24GB on my iMac right now. I have 64GB in my desktop PC.
There are workloads where you don't need more, but I honestly need 24GB for some tasks. If apple offered 24 or 32GB models, I'd be happy.
Also consider that the intel mac mini can support 64GB but the new one can only do 16. That's a big cut.
Then you get into the yesteryear sized SSDs. 1TB should be standard at this point. They're cheap now.
no docker and how long before the it's app store o (Score:2)
no docker and how long before the it's app store only?
Re: (Score:2)
no docker and how long before the it's app store only?
Most people can see the writing's on the wall for that. First you could run whatever you like, then they blocked that and you had to go into the security options and set it to 'run apps from anywhere', then they removed that option and hid it behind an 'right-click + run' process. The people who tell you this isn't going app store only wouldn't know the water they were sitting in was slowly boiling if you gave them a temperature gauge.
Re: (Score:2)
maybe not app store only but
app store only + some 3rd party devs only that need stuff the app store sand box does not work with. Stuff like office, adobe , other big apps.
Re: Consumer Gear (Score:4, Informative)
Re: (Score:2)
I am just a system admin. With a bunch of web browsers, terminals and email open, along with a Windows 10 VM for the occasional document that does not work in LibreOffice I am sitting at 24GB. So for me 16GB would be inadequate and it's been that way for some time. I upgraded my desktop at work about 4 years ago to 32GB and it made a big difference to responsiveness of the machine as it stopped swapping.
I have just put together a new system at home (Asus PN50 with 4800U) and went for the full 64GB. For the
Re: (Score:3)
No wonder software never feels any faster despite massive hardware speedups.
Re: (Score:2)
Which is cool- don't get me wrong. This machine looks like its designed for the lower-power users.
I'm a senior network and software engineer. I remember the last time I used a 16GB machine. It's like having 1.5 hands tied behind my back. Never again.
Re: (Score:2)
Yeah, I once spent two years working on a PC with 16MB of RAM, it was hell. You'd hit "compile" then go and sit on the support desk answering calls for ten minutes before coming back to test the app.
Re: (Score:2)
I work as a full time coder, and my 32GB MBP swaps. Chrome & Slack take most of the memory, with the fleet of containers running in Docker and the IT mandated spyware^U security software taking the rest. How a simple VPN can use ~240MB I'll never understand. Software in 2020 is insane.
Re: (Score:2)
Chrome & Slack take most of the memory
Yet you bitch about the VPN. The VPN isn't your problem, the shitty browser and browser based chat program you use are your problem.
How many "news" posts about this. (Score:5, Interesting)
Seriously, it is great in Geekbench. It may be good for some people. It may be bad for other people. But it is probably enough "news" posts about it? I'll just quote my own post from the previous story about it [slashdot.org], since we are copy pasting stories anyway...
Basically the M1 is better than all of Intel desktop & laptop CPUs and most AMD laptop CPUs at Geekbench and some other benchmarks when run natively on ARM. Sure, that is irrelevant to myself, as I use x86 VMs daily for development and they would be significantly slowed down on an ARM chip, but you can't really call it an outright lie, just classic marketing hyperbole, certainly not deserving a rant like this.
Apple has done much worse too. I am old enough to remember the switch from PowerPC to Intel CPUs. For a while they were selling both, so they had a section on the Apple website dedicated to showing off how much faster the PowerPC Macs are compared to PCs - conveniently using Pentium 4 (instead of the much faster back then AMD CPUs) and enabling AltiVEC on the Mac benchmarks but no SSE on the P4. AT THE SAME TIME, the "new" Intel Mac pages linked to a different section of the website, where there were benchmarks showing off how much faster the new Intel Macs were compared to the PowerPC Macs, to warrant the upgrade! If they could pull that off with a straight face, they can pull off anything.
PS. In the meantime I try to post stories like Cisco bungling an entire country's e-education system on launch day [slashdot.org] that seem quite big nerd news to me, and the editors decline it. Not enough space I guess after all the Apple news...
Re: (Score:2)
Boom (Score:3)
Goes the dynamite. These many-core designs were always the future of computing. It took us a lot to start making these risks but ARM opened us up more I believe, since we returned to these more fundamental changes in architecture. Many of the gains will likely be hard to measure, such as the gains from neural cores for AI applications. As a SD, playing with these sounds like a lot of fun. It will be interested to see how they approach dedicated graphics as the graphics cores currently seem very minimal. APUs from AMD have some different functionality with AMD graphics cards, but I have never personally experienced or played with it. I think Apples best route will be making a chip with a larger Graphics core count but I also have never seen a comparison of core typewith respect to heat generation. The restriction may be heat but otherwise they could Either have a single chip with 4x to 10x the graphics cores prevent more radical put two chips on the same board. The latter probably is more a unsound approach.
No matter what, I hope this starts to silence some naysayers. Jia you Apple.
Re: (Score:2)
Missing half the point (Score:2)
Re: Missing half the point (Score:2)
But this particular article (out of the many slashvertisements that have been posted about this laptop) is about benchmarks - so people are calling out the obvious marketing bullshit.
It's this mostly due to thermal throttling? (Score:3)
Re: (Score:3)
Mac books tend to have pretty bad thermals
Yes, but the burn scars on my thighs show just how good they are at getting that heat out of that shell.
Re: It's this mostly due to thermal throttling? (Score:2)
No - other manufacturers i9 mobile systems score similarly. Itâ(TM)s worth noting that the M1 is beating even AMD and Intelâ(TM)s fastest desktop chips at single threaded workloads. This thing is just a beast of a chip.
These are positioned as entry Mac replacements (Score:3)
But interesting is that these are positioned as entry level Macs in the range. Apple's M2 must be quite a bit faster still (and probably with 32gb on chip as well) to position them for the more upmarket Macs.
Regardless how relevant geek benchmarks are, I am pretty confident that these Macs will for most people be a total pleasure to use with recompiled software.
I am still enjoying a maxed out early 2014 Air for my home use. Hope it works well for another couple of years, so I can skip to the next generation of Apple Silicon Airs.
Yeah but how well does it run Dwarf Fortress? (Score:2)
Remember that this is the slowest Apple Silicon (Score:2)
They came out with a Air, the Mini, and the tiny Mac Book Pro. Apple's entry machines.
M1 isn't the end all. I would expect that the iMac gets the next tier up, and they will finish
the transition on the Pros with something Intel and AMD will have a hard time besting for a while.
That said, it's not magic and others could copy/catchup/contract with TMSC.
Re: (Score:2)
and the Macbook Pro is only "pro" in name.. not a single pro feature in it.
It doesn't even have enough USB-C connectors let alone useful stuff.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Power consumption is the real news here. (Score:2)
Given even cheap laptops can handle the vast majority of everyday computing activities without much effort, the focus on the CPU seems ... disingenuous.
Heck, I'm using a 2010 mac pro cheese-grater and even this has a multi-core score of 5100 - and it's 10 years old. Granted, it uses about the same amount of power as a small third world country ... so, you know, comparing Apples to Apples ... hmm.
But I'm willing to bet this old banger wipes the floor with the new MacBook Air on GPU benchmarks.
Even *that* is
Re: Sounds legit (Score:2)
Apple is one of only a very few manufacturers Iâ(TM)ve ever heard of getting caught doing the dirty like detecting geekbench and setting everything to Max clock. Do you have a reference to them doing that?