Thursday, November 12th 2020

Apple's M1-Based MacBook Air Benchmarked

When Apple announced that they are going to switch their Mac lineup from Intel-based x86 processors to the custom "Apple Silicon," everyone was wondering how the new processors will look and perform. To everyone's luck, Apple has just a few days ago announced its first Apple Silicon custom processor for MacBook. The M1, as the company calls it, is their first processor designed for higher-power and performance tasks The M1 features eight CPU cores (four high-performance and four-high efficiency) paired with eight cores dedicated to the graphics. On the die, there is also a 16-core neural engine made to accelerate machine learning tasks found in the new applications.

Today, we are getting the first GeekBench 5 CPU benchmarks that showcase just how far Apple has come with its custom design. What we have is the M1 processor found in MacBook Air. This Mac model features a passive cooling system, cooling a CPU with a base frequency of 3.2 GHz. The system scored 1719 points in the single-core result, and 6967 points in the multi-core result. The single-threaded results measure itself with some of the highest-end offerings from Intel and AMD, while the multi-threaded results are very good given the mix and match of small and big cores.
Source: GeekBench 5
Add your own comment

116 Comments on Apple's M1-Based MacBook Air Benchmarked

#26
Vya Domus
Dredi
How long does it take to launch candy crush?
Even that'd be more relevant, although it would test storage more than CPU performance.
Posted on Reply
#27
Blueberries
Dredi
Uh, ok. What would you deem relevant? Boot up time? How long does it take to launch candy crush?

Please elaborate.
Those are Dual / Single-Threaded processes so the GB5 ST benchmark is probably a pretty accurate representation for both of those operations.
Posted on Reply
#28
Dredi
Vya Domus
Even that'd be more relevant, although it would test storage more than CPU performance.
So you couldn’t figure out any relevant metric. Nice. Bootup and app launch are just a combination of IO performance and decompression.
Posted on Reply
#29
z1n0x
What's with the compiler differences between platforms?

Linux and Windows version use LLVM/CLANG 9.0, while on Apple LLVM/CLANG compiler used is v12, that's yet to be released.
Posted on Reply
#30
Vya Domus
Dredi
So you couldn’t figure out any relevant metric. Nice.
Lol okay dude, you know what, you're right SPECint is totally relevant. On my way to run some schur decompositions, I can't get enough of them.
Posted on Reply
#31
windwhirl
Fourstaff
Real life benchmarks uses a mixture of workloads etc etc. Everyone have their preferred reason to disregard a certain benchmark, but once you have enough of them there is really not much left but to accept.
At this point, I'd say Blender or any other type of application where you can test a workload in a systematic way would be best. But that also means waiting for those tools to be compiled and optimized for Apple Silicon.
Posted on Reply
#32
Dredi
Vya Domus
Lol okay dude, you know what, you're right SPECint is totally relevant. On my way to run some schur decompositions, I can't get enough of them.
They are a great indication of how the chip performs in various algorithms, and all software is just some IO and algorithms. If you can’t spot the ones relevant to your use case from SPECint and SPECfp it’s your problem. At anand they saw around 19% difference in spec between zen2 and zen3 and the same difference persisted in the ””real world”” tests.

SPEC is an industry standard for a reason.
windwhirl
At this point, I'd say Blender or any other type of application where you can test a workload in a systematic way would be best. But that also means waiting for those tools to be compiled and optimized for Apple Silicon.
There is a raytracing subtest in SPECfp (povray). Should correlate well with blender.
Posted on Reply
#33
Ashtr1x
Look at AT, same bs. GB and SPEC results x86 is dead all random trash talk. There is no overlap at all with Apple SoC and Intel/AMD business, the world wide marketshare of Macs and how much revenue does it contribute to the Apple is clear. Both are under 10%, and Apple has to pay a ton of money for Adobe to make software on their platform and optimize it. They also pay Intel a ton of cash, imagine paying that much while their whole R&D is being spent on 23% share of Services and the CPU for the iPhone sitting at 56% of their revenue cut, a simple cut of money to Intel nets them a ton of billions saved add those pesky lawsuits for the VRM throttling on Macs and heatup, when their userbase don't even run heavy loads... That's the reason Apple switched which many fail to get.

Then on top the x86 to ARM 32 bit is going to be axed soon, Apple hates backwards compatibility a lot. And no more Virtualization, the CPU is not having VM at all, and no eGPU as well, which brings even more haziness to the equation. Add heavily soldered platform, no more repairs at all, mega BGA garbage. Apple cleverly moved away from x86 by gating the Software behind their walled garden. No Phoronix tests, no real world tests are possible at all, some crappy browser benchmark.

All that SPEC scores are useless to measure the real world performance, AT benches the 5950X top of the line CPU with that tool and gets 50W power consumption on a single thread. There are no benches for rest of the CPU threads and cores at all, same for Apple chip only 1 big Firestorm core, rest are not at all benched how can we when the tool itself is only for that purpose. And if we look at A13 benchmarks the old articles also show same "reaching desktop territory" yet there was no software to show it nor the HW top rank in the Android vs Apple pricing (more on this pricing below). Now the Cinebench came with R23 for M1 support, will show some proper synthetic bench than these useless SPEC and GB. Still I'd like to see how the CPU performs with RPCS3 emulation can it even handle ? And how the Blender like workload behaves on the same CPU, add more benchmarks like high FPS gaming too, & other software productivity tools like MATLAB.

I will even add the more stupid tests here, consider watching those Youtube tests of iPhone vs Samsungs and Onepluses etc.. how much of the SPEC graph differs in the AT benches and then the real world performance where Application response times, execution windows for multiple apps and in game FPS benchmarks, Apple always lost. All in all this SPEC is like zooming in one part of the picture and then deciding it's the best, Plus I will even mention if the Apple iPhone CPU is very fast vs Android's Qualcomm QSDs and Exynos, Kirins why the phones pricing is similar ? On Desktop side I saw Intel getting shredded in the X299 market and Mainstream market where they slashed the prices significantly due to lack of performance vs competition, best example is Nvidia's Turing vs Ampere pricing, the performance dictates the market in consumer side, this whole fiasco of world's fastest bs claims is very stupid imho.
Posted on Reply
#34
Dredi
Ashtr1x
All that SPEC scores are useless to measure the real world performance
Compression is useless? Decompression is useless? Raytracing is useless? Physics algorithms are useless? Sure..
Posted on Reply
#35
Ashtr1x
Dredi
Compression is useless? Decompression is useless? Raytracing is useless? Physics algorithms are useless? Sure..
Show me where the work is done, I want to see the work not some silly synthetic score and determine. Show me a graph that Gamers Nexus and TPU showcases while Decompression and Compression with 7-Zip software, for Raytracing, I would like to see DXR new 3DMark benchmark or any game, SOTTR has a pure Vulkan based release, not sure if it has Vulkan based Raytracing or not, but it is out now, for gaming, BioShock 2 Remastered is also available on mac OS, Physics Algorithms ? what is that, MATLAB simulation ? well I don't see them either.

Guess what ? All of them have to be recompiled and updated with ARM from x86 code plus if the executable is 32bit then say goodbye in the future unless the developer updates it, also this M1 chip doesn't have eGPU output as well, meaning it won't even meet min spec for the SOTTR and BioShock 2 on Mac. Not even Adobe has software ready, it's due in 2021. And also Raytracing ? You sure, running some SPEC load on a CPU for Raytracing of all, Nvidia has 3 years of AI advancement (on the side note, Apples Siri is utter trash vs Google Assistant, Apple is lacking severely in that dept.) and still they couldn't nail the RT, AMD barely started and suddenly Apple is going to takeover them ? It's not even funny anymore.
Posted on Reply
#36
Fouquin
pat-roner
Say what you want about Geekbench, but I found this quote from Andrei interesting;

There’s been a lot of criticism about more common benchmark suites such as GeekBench, but frankly I've found these concerns or arguments to be quite unfounded. The only factual differences between workloads in SPEC and workloads in GB5 is that the latter has less outlier tests which are memory-heavy, meaning it’s more of a CPU benchmark whereas SPEC has more tendency towards CPU+DRAM.

I can only say that I'm really excited about the new M1 chips. As a lover of technology I'm all for new technology, and seeing people hate on a product they will never buy is beyond me.

PCMasterrace amirite?/s
GeekBench is also an OS benchmark, and macOS is the fastest OS on GeekBench. There are some options hidden in macOS for GB5 for example that improve performance on any architecture, even Ryzen.
Posted on Reply
#37
Chrispy_
Rahnak
Geekbench is worthless for comparing between different architectures.
QFT.
We need actual application performance for apples-to-apples comparisons - and even then, the recompiled-for-ARM version of AfterEffects for example still won't be apples-to-apples as the x86_64 variants will be the lowest-common-denominator version that'll run on a huge range of different OSes and architectures. Still, it'll be interesting to see the Native ARM version on apple vs the Windows in like-for-like application testing.
Posted on Reply
#38
okbuddy
smash 5900x? cinebench r23
Posted on Reply
#39
Dredi
Ashtr1x
Show me where the work is done, I want to see the work not some silly synthetic score and determine. Show me a graph that Gamers Nexus and TPU showcases while Decompression and Compression with 7-Zip software, for Raytracing, I would like to see DXR new 3DMark benchmark or any game, SOTTR has a pure Vulkan based release, not sure if it has Vulkan based Raytracing or not, but it is out now, for gaming, BioShock 2 Remastered is also available on mac OS, Physics Algorithms ? what is that, MATLAB simulation ? well I don't see them either.

Guess what ? All of them have to be recompiled and updated with ARM from x86 code plus if the executable is 32bit then say goodbye in the future unless the developer updates it, also this M1 chip doesn't have eGPU output as well, meaning it won't even meet min spec for the SOTTR and BioShock 2 on Mac. Not even Adobe has software ready, it's due in 2021. And also Raytracing ? You sure, running some SPEC load on a CPU for Raytracing of all, Nvidia has 3 years of AI advancement (on the side note, Apples Siri is utter trash vs Google Assistant, Apple is lacking severely in that dept.) and still they couldn't nail the RT, AMD barely started and suddenly Apple is going to takeover them ? It's not even funny anymore.
I think you are mixing raytracing as a cpu benchmark with computer games. The spec test is more like blender.

7z performace correlates almost 1:1 with the ””syntethic”” spec compression/decompression tests. Could you be more specific on why you think that spec does not translate to real world performance? It does take some time for SW to be compiled to work natively on the new platform, but you wouldn’t buy it for productive work anyway before the SW support is there.
Posted on Reply
#40
Rahnak
Chrispy_
QFT.
We need actual application performance for apples-to-apples comparisons - and even then, the recompiled-for-ARM version of AfterEffects for example still won't be apples-to-apples as the x86_64 variants will be the lowest-common-denominator version that'll run on a huge range of different OSes and architectures. Still, it'll be interesting to see the Native ARM version on apple vs the Windows in like-for-like application testing.
I would be cool if you could install Windows 10 (ARM) or Linux on these new machines to properly test, but I doubt Apple will make that possible. I'm sure the walls are even higher in this new garden.
Posted on Reply
#41
bummpr
And the IRONY Champion of the World (drumroll please)..."imho"
Ashtr1x
Look at AT, same bs. GB and SPEC results x86 is dead all random trash talk. There is no overlap at all with Apple SoC and Intel/AMD business, the world wide marketshare of Macs and how much revenue does it contribute to the Apple is clear. Both are under 10%, and Apple has to pay a ton of money for Adobe to make software on their platform and optimize it. They also pay Intel a ton of cash, imagine paying that much while their whole R&D is being spent on 23% share of Services and the CPU for the iPhone sitting at 56% of their revenue cut, a simple cut of money to Intel nets them a ton of billions saved add those pesky lawsuits for the VRM throttling on Macs and heatup, when their userbase don't even run heavy loads... That's the reason Apple switched which many fail to get.

Then on top the x86 to ARM 32 bit is going to be axed soon, Apple hates backwards compatibility a lot. And no more Virtualization, the CPU is not having VM at all, and no eGPU as well, which brings even more haziness to the equation. Add heavily soldered platform, no more repairs at all, mega BGA garbage. Apple cleverly moved away from x86 by gating the Software behind their walled garden. No Phoronix tests, no real world tests are possible at all, some crappy browser benchmark.

All that SPEC scores are useless to measure the real world performance, AT benches the 5950X top of the line CPU with that tool and gets 50W power consumption on a single thread. There are no benches for rest of the CPU threads and cores at all, same for Apple chip only 1 big Firestorm core, rest are not at all benched how can we when the tool itself is only for that purpose. And if we look at A13 benchmarks the old articles also show same "reaching desktop territory" yet there was no software to show it nor the HW top rank in the Android vs Apple pricing (more on this pricing below). Now the Cinebench came with R23 for M1 support, will show some proper synthetic bench than these useless SPEC and GB. Still I'd like to see how the CPU performs with RPCS3 emulation can it even handle ? And how the Blender like workload behaves on the same CPU, add more benchmarks like high FPS gaming too, & other software productivity tools like MATLAB.

I will even add the more stupid tests here, consider watching those Youtube tests of iPhone vs Samsungs and Onepluses etc.. how much of the SPEC graph differs in the AT benches and then the real world performance where Application response times, execution windows for multiple apps and in game FPS benchmarks, Apple always lost. All in all this SPEC is like zooming in one part of the picture and then deciding it's the best, Plus I will even mention if the Apple iPhone CPU is very fast vs Android's Qualcomm QSDs and Exynos, Kirins why the phones pricing is similar ? On Desktop side I saw Intel getting shredded in the X299 market and Mainstream market where they slashed the prices significantly due to lack of performance vs competition, best example is Nvidia's Turing vs Ampere pricing, the performance dictates the market in consumer side, this whole fiasco of world's fastest bs claims is very stupid imho.
Posted on Reply
#42
Chrispy_
Rahnak
I would be cool if you could install Windows 10 (ARM) or Linux on these new machines to properly test, but I doubt Apple will make that possible. I'm sure the walls are even higher in this new garden.
I have faith that the Hackintosh crowd will be trying hard to break those walls. They do it for funzies.
Posted on Reply
#43
Dredi
Rahnak
I would be cool if you could install Windows 10 (ARM) or Linux on these new machines to properly test
Why the hell would you need windows to test these? I’m sure there will be more than enough multi platform apps to make good conclusions about the performance differences.
Posted on Reply
#45
Rahnak
Dredi
Why the hell would you need windows to test these? I’m sure there will be more than enough multi platform apps to make good conclusions about the performance differences.
You don't need it, obviously. But it would put the machines on more of an equal ground. If Apple ever released drivers for the M1, of course. Which they won't.

Mac OS will be optimized for this SOC in a way Windows/Linux can't really ever come close to and that alone already gives it a leg up.
Posted on Reply
#46
windwhirl
Rahnak
You don't need it, obviously. But it would put the machines on more of an equal ground. If Apple ever released drivers for the M1, of course. Which they won't.

Mac OS will be optimized for this SOC in a way Windows/Linux can't really ever come close to and that alone already gives it a leg up.
Indeed. Apple only worries about their software running on their hardware and screw everything else. We should see even tighter integration in the future.

Windows/Linux on the other hand have to work on nearly all hardware launched in the last 10 to 20 years.
Posted on Reply
#47
Dredi
Rahnak
You don't need it, obviously. But it would put the machines on more of an equal ground. If Apple ever released drivers for the M1, of course. Which they won't.

Mac OS will be optimized for this SOC in a way Windows/Linux can't really ever come close to and that alone already gives it a leg up.
Would it not make more sense to bench the systems with the operating systems their users use 99% of the time? No one will buy these new machines to run windows. Installing windows would just make the test more synthetic in nature.
Posted on Reply
#48
Searing
Rahnak
Geekbench is worthless for comparing between different architectures.
wrong, go back to 2010 where you come from

Geekbench is great for different OS and ISA comparisons, and is also not memory dependent anymore

next week there will be a lot of surprised people who have had their heads in the sand for the last 5 years

trends matter, Apple has been improving more than Intel every year for 5 years, that is why they are ahead now
Posted on Reply
#49
illli
Sandbo
I had been discouraged by the fact that paying a premium for mac is giving me overpriced, outdated spec, guess this has changed now.
No, they are still overpriced. The $699 mac mini only comes with 8Gb ram (lol) and 256Gb ssd (lol again).
Oh, you might think you can upgrade this yourself? NOPE. Everything is soldered onto the mobo.
The $899 mini is STILL the same 8Gb ram (pathetic) and 512b ssd (so basically +$200 for 256Gb upgrade)
Posted on Reply
#50
Vayra86
Vya Domus
Here is why Geekbench is dogshit :



Every time other SoCs inch closer to Apple's the app gets updated and a chasm appears again between their chips and everyone else's.

In 4.4 there was less than 10% difference between A13 and Exynos 990, in 5 that somehow became a colossal 50%. This happens every single time a new version appears, without exception. Only an idiot would take these numbers for granted and not realize that this benchmark is always optimized specifically for Apple's chips.
Can we get a community effort going to just straight up shitlist this bench from any news on TPU? Or alternatively, demand a neutral bench is run alongside it...
Posted on Reply
Add your own comment