Thursday, November 12, 2020

Apple M1 Benchmarks

Juli Clover (tweet, Hacker News):

In comparison to Macs, the single-core performance is better than any other available Mac, and the multi-core performance beats out all of the 2019 16-inch MacBook Pro models, including the 10th-generation high-end 2.4GHz Intel Core i9 model.

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.

Andy Somerfield:

Apple M1 chip benchmark vs. 6-core 3.7ghz 2019 iMac with AMD 580X in @affinitybyserif Photo - if I hadn’t measured the CPU number myself I wouldn’t believe it 😂A monster..

Steve Troughton-Smith:

Also — bear in mind that Apple’s other two M1 machines have active cooling, which could mean even higher benchmark performance. These are incredibly fast chips.

Previously:

Update (2020-11-16): Hartley Charlton (tweet):

Apple’s M1 processor often surpasses the graphics performance of desktop GPUs, including the Nvidia GeForce GTX 1050 Ti and AMD Radeon RX 560, according to a new benchmark submission spotted by Tom’s Hardware.

[…]

It is worth noting that GFXBench 5.0 benchmarks are primarily used to test mobile devices, and the GeForce GTX 1050 Ti and Radeon RX 560 are older GPUs. Nonetheless, the M1 contains integrated graphics and the fact that it can surpass 75W desktop graphics cards is impressive.

Colin Cornaby:

  • This is very good performance for an integrated GPU
  • This is not great performance vs desktop GPUs. The 560 is ancient history at this point.

Apple is still shipping it on iMacs, though.

Update (2020-11-23): Juli Clover:

The new M1 Macs are now arriving to customers, and one of the first people to get the new M1 13-inch MacBook Pro with 8-core CPU, 8-core GPU, and 8GB unified memory has run a much anticipated R23 Cinebench benchmark on the 8GB 13-inch MacBook Pro with 512GB of storage to give us a better idea of performance.

Ryan Jones:

A guy in Australia is testing M1 MacBook Pro vs iMac Pro by exporting H.265 10bit video...

$7k iMac Pro = 11m 30s
$1.2k MacBook Pro = 10m 20s

Oh, and the iMac Pro has 128GB of ram and the MacBook has 8GB.

Andrei Frumusanu (Hacker News):

The performance of the new M1 in this “maximum performance” design with a small fan is outstandingly good. The M1 undisputedly outperforms the core performance of everything Intel has to offer, and battles it with AMD’s new Zen3, winning some, losing some. And in the mobile space in particular, there doesn’t seem to be an equivalent in either ST or MT performance – at least within the same power budgets.

What’s really important for the general public and Apple’s success is the fact that the performance of the M1 doesn’t feel any different than if you were using a very high-end Intel or AMD CPU. Apple achieving this in-house with their own design is a paradigm shift, and in the future will allow them to achieve a certain level of software-hardware vertical integration that just hasn’t been seen before and isn’t achieved yet by anybody else.

Christian Selig:

Damn, @Dave2D coming in clutch with the Xcode build time tests on the new Apple Silicon Macs. That’s just nuts. I really want one to replace my aging Mac but am going to wait for the 16”.

Matthew Panzarino (tweet):

After a single build of WebKit, the M1 MacBook Pro had a massive 91% of its battery left. I tried multiple tests here and I could have easily run a full build of WebKit 8-9 times on one charge of the M1 MacBook’s battery. In comparison, I could have gotten through about 3 on the 16” and the 13” 2020 model only had one go in it.

Rik Arends (via Hacker News):

It’s hard to believe, a hot running fan blasting i9 loses by a factor 2 in a real world Rust compile from a passively cooled M1 macbook air. This has consequences.

Arnold Kim:

The benchmarks confirm that the new MacBook Air SSD is approximately twice as fast as the previous model with 2190 MB/s writes and 2675 MB/s reads. Apple did announce this performance boost when launching the new machine, and attributed the improved speed to the M1’s storage controller as well as new flash technology[…]

The 2019/2020 Intel MacBook Air was criticized for actually offering a slower SSD than the 2018 model. The 2019/2020 MacBook Air SSD speeds were 1319 MB/s reads and 1007 MB/s writes.

Ken Case:

Building one of our iOS apps from scratch, 3x, on Intel Macs vs. M1 Macs:

  • Intel MBP (2020, 32GB): 9:58, 10:08, 9:59
  • Intel mini (2020, 64GB): 9:17, 8:20, 8:23
  • M1 Air (16GB): 5:49, 6:14, 6:17
  • M1 mini (16GB): 5:44, 5:36, 5:34

See also: Wojtek Pietrusiewicz.

David Smith:

This 13" Silicon MacBook Pro is the first computer I’ve ever used that is faster than my “Distraction Threshold”…where the time between hitting Build+Run and the app launching is too short for my mind to wander, which has been fantastic for my focus the last few days.

John Voorhees:

The Pixelmator team says ML Super Resolution can be accomplished up to 15 times faster on an M1 Mac than was possible before.

Steve Troughton-Smith:

Real ‘wow’ moment; try changing the resolution on an M1-based Mac, or attaching an external display — no blanking, just instant

Tyler Stalman:

Currently editing 45 megapixel photos w/lots o’layers on the new #M1 MacBook Pro 13”, 8GB of ram

Zero lag, no slowdowns, no fan noise. I’d never guess this machine wasn’t way higher end and more expensive

And Photoshop is still running in emulation 😵

Paul Hudson:

My Unwrap project (13k lines Swift, 10k Obj-C, more) was 19.5s on Intel vs 11.7 on M1.

AudioKit (39k C, 27k Swift, 12k C++) was 73s on Intel vs 31 on M1.

Paul Haddad:

Performance wise the MBA M1 is 💯. Haven’t experienced any throttling, easily faster than any real Macs I’ve ever used.

It’s not however the fastest OS X experience I’ve had, a beefed up desktop PC hardware (10+ cores) can beat it.

Helge Heß:

A fresh debug build of Noze.io takes 20secs on the M1 Mini, and 35secs on the 6 core i7 (other specs the same, 16/512). So almost twice as fast on compilation. Price (net): M1 €1058, i7 €1660 (Jan 2019)

Valtteri Koskivuori:

So @jcs helped benchmark my C-ray renderer on the Apple #M1 Chip. This image was rendered by the M1 MacBook Air in 29.44 seconds! For reference, my 2019 16" MBP (i9-9750H) did it in 38.74s seconds and my 2018 13" MBP (i5-8259U) did it in 53.14s. That new Air is ridiculously fast!

Juli Clover:

We’ve seen endless speed tests, but we thought we’d pit the M1 13-inch MacBook Pro against the model that it’s replacing, the 2020 13-inch MacBook Pro with 1.4GHz quad-core Core i5 processor, Intel Iris Plus Graphics 645, and 8GB RAM.

Maxwell Swadling:

MacBook Air finally arrived! It runs my z3 solvers about 20% faster than an i7. And a lot colder! Nice! CMD+F1 to enable screen mirroring is much faster now! Pretty much instant.

David Heinemeier Hansson:

You don’t sit around thinking “oh, browsing the web is slow on my decked-out iMac”, but then you browse with the M1, and you’re like, DAMN, I can actually feel that +50%. 224 on Speedometer 2.0 is just 🤯

Savannah Reising:

Take a look at the performance improvements we saw when running Astropad and Luna on M1[…]

Luke Miani:

Quick Minecraft test... even the MacBook Air running at 10 watts, without a fan, through a translation layer, is running 60fps at native res without getting warm at all. Apple Silicon is nuts lol

Francisco Tolmasky:

An sad but inescapable conclusion from the impressive launch of the M1 is just how much Apple squandered the potential of the iPad. The iPad has had amazing performance for a while, so why is the M1 a game changer? Because it’s finally in a machine we can actually do things on.

Update (2020-11-30): Frank Reiff:

One thing about the M1 equipped MacBooks that nobody has made a sufficiently huge fuss about is that that just like an iPad they are connected to Wifi even before you open them up.. no 5 seconds scanning for networks. Wifi is just on.

It sounds minor, but it’s actually huge.

I wouldn’t be surprised if it turned out that an awful lot of things that people love about iPads have nothing to do with multi-touch.. it’s quite startling how different an instant-on Mac feels from a.. Mac and how similar it feels to an iPad.. and how that’s not a bad thing.

Nadim Kobeissi:

At the risk of making this a very boring review, all I have to say is basically that all of the above is so far holding as true. This computer is nuts. Compiling tons of things in the background doesn’t slow down Safari web browsing, somehow. I haven’t had to plug it in once since I fully charged it up two days ago. Performance is so quiet and cool that I feel like my terminal compiling a bunch of things is actually an SSH into a much stronger workstation located somewhere else. I actually discovered that I’ve had an instinct of measuring my MacBook’s CPU usage by feeling the heat on the strip of aluminum right above the Touch Bar, and I can’t even do that anymore now. Because even if the M1 MacBook Pro has been running at 100% on all cores for ten straight minutes, you’ll barely feel it getting warm.

This just doesn’t feel like an entry-level 13” laptop. It feels like a workstation that’s magically bottled within a 13” laptop chassis.

Update (2020-12-16): Daniel Lemire (Hacker News):

The new laptop is faster in these specific tests. Yet the differences are all over the map.

I do not yet understand why the fast_float library is so much faster on the Apple M1. It contains no ARM-specific optimization.

Daniel Lemire:

In some cases, the ARM-based MacBook Pro was nearly twice as fast as the older Intel-based MacBook Pro.

[…]

But it does not follow that the 128-bit ARM NEON instructions are generally a match for the 256-bit SIMD instructions Intel and AMD offer.

[…]

As you can see, the older Intel processor is slightly superior to the Apple M1 in the minify test.

Howard Oakley:

This article looks at one specific feature – SMB Target Disk mode – assesses how fast it is, finds a nasty problem with it, and concludes by considering M1 performance more generally.

Howard Oakley:

On my T2-equipped iMac Pro, writing a 10 MB file occurred at 1.8 GB/s, rising to 2.6 GB/s for 10 GB (a thousand times the size). Reading the same files delivered 6.2 GB/s for 10 MB, and 1.5 GB/s for 10 GB. Those results are broadly consistent with those found using Blackmagic.

On my M1 Mac mini, results were quite different. Writing occurred at 2.4 (10 MB) to 2.9 GB/s (10 GB), which is at the upper end of the T2’s performance, but reading was significantly faster at 10.8 (10 MB) falling to 3.1 GB/s (10 GB).

Michael Potuck:

YouTuber Martin Nobel shared a useful video of the process to run an ARM Windows virtualization on Apple Silicon as well as a real-world look at the overall impressive performance considering it’s an unofficial workaround.

Impressively, the Martin’s M1 Mac mini benchmarked much higher than Microsoft’s Surface Pro X…almost doubling the single-core score, and coming in almost 2,000 higher in the multi-core score.

Update (2020-12-24): Howard Oakley:

If you feel that you can explain these very high transfer rates, I’d love to learn how my M1 Mac mini is achieving them without breaking the laws of physics.

Update (2021-01-06): Daniel Lemire (Hacker News):

The latest Apple laptop processor, the M1, has apparently a lot of memory-level parallelism. It looks like a single core has about 28 levels of memory parallelism, and possibly more.

Howard Oakley:

Returning, at last, to the original comparison that I wanted to make, the M1 SSD is similar in performance to that in the iMac Pro. Their write speeds are almost identical when comparing median values, although using linear regression the iMac Pro is faster at 3.0 versus 2.6 GB/s. However, the iMac Pro has poorer read performance, at 2.2 GB/s, compared with the M1 at 2.8 GB/s. Whether those will translate into any perceptible difference in performance isn’t clear.

The differences between the internal SSDs and the external encrypted SSD are stark: typically 300 MB/s read and 295 MB/s write, around 10% of the internal SSD transfer rates.

[…]

I’m happy now, as I hope you are, that Stibium is producing test results which appear consistent and reliable.

1 Comment RSS · Twitter

Could some of the nice performances of the M1 Mac be related to the CPU (and so the software) not suffering from the Spectre, Meltdown, etc. vulnerabilities?

Leave a Comment