Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

My Cinebench Throttling tests M2 vs M1 MacBook Airs

M

Mike Boreham

macrumors 601
Original poster
Aug 10, 2006
4,019
2,035
UK
Running continuous Cinebench test for 30 mins is a common way of assessing CPU throttling, which is a hot topic on the fanless M1 and M2 MBAs. I have seen claims that the M2 throttles more. I tested my M1 MBA (512/16) in Jan 2021, and just upgraded to a M2 MBA(2TB/24), and reran the tests out of curiosity. This chart shows the results:

Screenshot 2023-01-23 at 12.19.41.png


Conclusions

1. The M1 throttles by about 10% over 30 minutes. The M2 throttles by 15% over 30 minutes. So yes the M2 does throttle more (in this test), but....

2. The M2 is 15% faster than the M1 after first run. It is still 15% faster after 10 mins, and 9% faster after 30 minutes.

Notes about tests

- The M1 was on Big Sur at that time and the M2 is on Ventura.
- I did not take any special measures before testing, only to make sure machine had been idling for some hours before starting and quitting all running processes in the Dock, but not background processes. It looks like run#14 on the M2 might have been affected by a background task (TM?) kicking in.
- The average CPU core temp on the M2 was 105C after run1 and 95 after 19. I didn't record the M1 temps in 2021. There was a step in temps between runs 11 and 12 on the M2.
- Because the M2 is faster it did 19 runs in the 30 mins and the M1 did 16, so the x-axes are different.

Comment

I am not a great fan of benchmarking, and tend to only do it after getting a new machine, but in real life I would probably not notice a 15% slow down, and will occur very rarely on my machines.
Cinebench tests the CPU not the GPU and the M2 has 2 extra cores compared to the M1.
 
Last edited:
  • Like
Reactions: Makisupa Policeman, VaruLV, hxlover904 and 3 others
Thank you for the comprehensive benchmark. However,

This can be a hot topic only amongst people who have no idea this computer was not designed for this type of workload. It’s an interesting observation for the sake of curiosity but nothing more. Next, someone will perform the same test on the MBP and will claim for a few hundred bucks more you get like 5 times speeder computer.

Other conclusion: The M2 is faster by 9% after 30 minutes. Cool. It’s also like 20% more expensive. Dollar-per-productivity the M1 is better.
 
  • Like
Reactions: sorgo † and Alex Cai
Mike Boreham said:
Running continuous Cinebench test for 30 mins is a common way of assessing CPU throttling, which is a hot topic on the fanless M1 and M2 MBAs. I have seen claims that the M2 throttles more. I tested my M1 MBA (512/16) in Jan 2021, and just upgraded to a M2 MBA(2TB/24), and reran the tests out of curiosity. This chart shows the results:

View attachment 2146738

Conclusions

1. The M1 throttles by about 10% over 30 minutes. The M2 throttles by 15% over 30 minutes. So yes the M2 does throttle more (in this test), but....

2. The M2 is 15% faster than the M1 after first run. It is still 15% faster after 10 mins, and 9% faster after 30 minutes.

Notes about tests

- The M1 was on Big Sur at that time and the M2 is on Ventura.
- I did not take any special measures before testing, only to make sure machine had been idling for some hours before starting and quitting all running processes in the Dock, but not background processes. It looks like run#14 on the M2 might have been affected by a background task (TM?) kicking in.
- The average CPU core temp on the M2 was 105C after run1 and 95 after 19. I didn't record the M1 temps in 2021. There was a step in temps between runs 11 and 12 on the M2.
- Because the M2 is faster it did 19 runs in the 30 mins and the M1 did 16, so the x-axes are different.

Comment

I am not a great fan of benchmarking, and tend to only do it after getting a new machine, but in real life I would probably not notice a 15% slow down, and will occur very rarely on my machines.
Cinebench tests the CPU not the GPU and the M2 has 2 extra cores compared to the M1.
Click to expand...
If you are doing any task regularly that take 30+ continuous minutes on a MacBook Air then you chose the wrong MacBook. That’s why all the hand-wringing over the M2 Air and throttling was so misguided. The M2 is a powerful SoC that can handle more than just simple office tasks but the MacBook Air was never meant to be used to run high-power sustained loads.

Edit:
georgivelev said:
Other conclusion: The M2 is faster by 9% after 30 minutes. Cool. It’s also like 20% more expensive. Dollar-per-productivity the M1 is better.
Click to expand...
🤦 Case in point. No the M2 is 15%-20% faster for everything you would reasonably buy an M2 MacBook Air to do.
 
Last edited:
  • Like
  • Haha
Reactions: ilikewhey, boss.king, now i see it and 3 others
Mike Boreham said:
Running continuous Cinebench test for 30 mins is a common way of assessing CPU throttling, which is a hot topic on the fanless M1 and M2 MBAs. I have seen claims that the M2 throttles more. I tested my M1 MBA (512/16) in Jan 2021, and just upgraded to a M2 MBA(2TB/24), and reran the tests out of curiosity. This chart shows the results:

View attachment 2146738

Conclusions

1. The M1 throttles by about 10% over 30 minutes. The M2 throttles by 15% over 30 minutes. So yes the M2 does throttle more (in this test), but....

2. The M2 is 15% faster than the M1 after first run. It is still 15% faster after 10 mins, and 9% faster after 30 minutes.

Notes about tests

- The M1 was on Big Sur at that time and the M2 is on Ventura.
- I did not take any special measures before testing, only to make sure machine had been idling for some hours before starting and quitting all running processes in the Dock, but not background processes. It looks like run#14 on the M2 might have been affected by a background task (TM?) kicking in.
- The average CPU core temp on the M2 was 105C after run1 and 95 after 19. I didn't record the M1 temps in 2021. There was a step in temps between runs 11 and 12 on the M2.
- Because the M2 is faster it did 19 runs in the 30 mins and the M1 did 16, so the x-axes are different.

Comment

I am not a great fan of benchmarking, and tend to only do it after getting a new machine, but in real life I would probably not notice a 15% slow down, and will occur very rarely on my machines.
Cinebench tests the CPU not the GPU and the M2 has 2 extra cores compared to the M1.
Click to expand...
Just asking, how do you make this chart? How do you obtain the data for each run?
I know that 3D mark wild life automatically produce these charts but I can’t see this option on cinebench.
 
georgivelev said:
This can be a hot topic only amongst people who have no idea this computer was not designed for this type of workload. It’s an interesting observation for the sake of curiosity but nothing more.
Click to expand...

Completely agree.


georgivelev said:
Other conclusion: The M2 is faster by 9% after 30 minutes. Cool. It’s also like 20% more expensive. Dollar-per-productivity the M1 is better.
Click to expand...

Value for money relative to the M1 Air is a whole different debate. No one in their right mind would upgrade or buy the M2 for just a 15% spec bump. Whether the other things matter to you is personal.

I upgraded from my M1 MBA to M2 MBA because I wanted 2TB internal storage, (my M1 was only 512GB). I was lucky to get a nearly new 2TB/24GB one from CEX for £500 less than Apple price. It had 2 hrs run time and was well within 60 days for Applecare+ which I had no trouble buying.
 
jdb8167 said:
If you are doing any task regularly that take 30+ continuous minutes on a MacBook Air then you chose the wrong MacBook. That’s why all the hand-wringing over the M2 Air and throttling was so misguided. The M2 is a powerful SoC that can handle more than just simple office tasks but the MacBook Air was never meant to be used to run high-power sustained loads.
Click to expand...

Completely agree. I haven't bought it to do 30+ continuous minutes at 100%, but I do like to test stuff particularly on controversial issues.
 
Alex Cai said:
Just asking, how do you make this chart? How do you obtain the data for each run?
I know that 3D mark wild life automatically produce these charts but I can’t see this option on cinebench.
Click to expand...
Cinebench prefs have a setting runtime. I selected 30 minutes, but in practice I had to re-launch Cinebench every 10 minutes. Then you have to sit over the machine and note the Cinebench score after each run by hovering the pointer over your machine in the list of computers on the LHS. After that its all Numbers, and annotating in Preview.
 
  • Like
Reactions: Makisupa Policeman and hxlover904
@jdb8167 and @georgivelev

Although I "agreed completely" with your comments that the MBA is not the right machine for 30 mins heavy duty tasks, it is an important (at least to me) fact that it is actually capable of doing heavy duty tasks on an occasional basis with only 15% throttling. I would not feel good about it being my only machine if I thought such tasks were impossible when needed. I will sometimes do this.
 
  • Like
Reactions: Mike82, hxlover904 and jdb8167
maflynn said:
The MBA is just not equipped to handle sustained loads, interesting to see how the M2 is handling it though.

Thanks for the informative graph and the grunt work needed to produce this
Click to expand...

But it is capable of doing such tasks when required. If needed regularly it is the wrong machine.
 
  • Like
Reactions: jdb8167
I have an MBA M2/1TB/24GB. I could have gone for the 2TB option as I'm paying it off over 24 months with Apple's credit partner in Australia (Latitude with the Creditline card), and while it was AUD $900 extra to go from 1TB to 2TB, which is an extra AUD $37.50 per month, the cost would have gone from the current $145 per month, to $182.50 per month, and external SSDs have never been cheaper, so I went for the 1TB model.

With the MBA M3, I'll max it out completely :)

I regularly edited video interviews on the MBA M1, and did so on previous MBA models with Intel processors, although those models by necessity of being powered by Intel processors, had cooling fans.

Going to a completely silent system is incredible, with vastly larger amounts of power than ARM-powered notebooks running the ARM version of Windows 11. As usual, it is better running Windows 11 virtually on Mac OS (be it on an ARM or Intel-powered system) than it is running it natively, although you'll always want gobs of RAM and the fastest Intel processors for the best experiences. Apple Silicon processors are already the best.

Anyway I was editing video interviews then, and I'm doing it now from the MBA M2. The larger screen is noticeable and welcome. The notch eats into the damn menu bar but I have Bartender for MacOS to help manage all of the right hand side running app icons to put a lot of them into a second row that you can pop up by clicking in the notch or menu bar area.

So being able to consistently edit videos on a regular basis without the MBA M2 breaking any noticeable sweat when rendering and really, run anything I want to without being hit with the kind of insanely noticeable and really bad performance penalties of the Intel Core M series of underpowered processors that was used in the super thin 12-inch MacBook from the middle of the last decade.

If you stayed in Safari and used the machine lightly, it was fine, but the moment you loaded Firefox, Chrome or Word, you could feel it really slow down.

Apple was trying to will the future into existence and if Intel had an appropriately powerful processor, it would have been a huge hit, but the processor was barely up to the job.

That 12-inch MacBook is now the 13.6-inch MBA M2.

And amazingly, the 8GB and 512GB config of the MBA M1 that I still have is amazing. The machine still works beautifully, I can screen record anything I want with Camtasia 2022 and edit it as needed, then export and upload to YouTube. The same goes with the M2. With no fan noise.

And tons of tabs open, textedit, preview, NordVPN, Word, Chrome, Messages, Whatsapp, Copied, etc etc. I'm always doing a lot with my Macs.

I have found it to be a great idea with the last four generations of MacBook Air to restart the machine if wanting to record something with Camtasia so everything is fresh, and that's of course where it's very handy to have more than one physical Mac. It's also great to have at least two phones, too.

Anyway have had the MBA M2 for months now and I love it. Of the 24GB available, 6.7GB was free, so the more RAM you have, the more your Mac has space to breathe without needing to turn to your SSD, and pressing the free memory button in the Parallels Toolbox memory management app called "Free Memory", and it went to 9.6GB free before settling at 8.81GB of free memory as I type.

That is without Camtasia 2022 open at this time. Plus I normally connect via Zoom and use the Camo Studio software to record video of me through the iPhone 14 Pro Max's main camera, when doing video interviews being recorded by Camtasia 2022, so there's always those programs running at the same time so I can record the video I want to record, and get better quality video than the improved 1080P webcam atop the M2 MBA, which the notch holds - without adding Face ID capabilities (that will hopefully come in the M3-powered model).

So to conclude, I do push my M2, or at least I think I do. I'm doing doing Cinebench, just the computing that I do, and yes, the supposed problems from the M2 throttling are unquestionably overblown to the max, simply to generate a reason to watch a series of YouTubers and read a series of bloggers.

Even throttled, the M2 produces better performance than the M1 at all times, so how are you supposed to even notice this if you are coming from an Intel or M1-based MBA?

And because of the discovered slowdown of read/write speeds on a single 256GB chip instead of the previous 2x 128GB chips, or 2x256 or larger SSDs, the above statement is for MBA M2 models with a 512GB or larger drive, don't get the 256GB only model.

There have been no benchmarks that I've seen of the M2 Mac mini as yet, but I'm guessing its 256GB base model with 8GB RAM is going to have the same slower read/write SSD speeds of the MBA M2 8GB/256GB - we'll soon see what the reviews and benchmarks say. '

Anyway - thank you for the chart and the work you've done! Bring on the 3nm M3 processor architecture and family - it will be fascinating to see the improvements Apple will deliver at that time, while the M2 Pro and Max inside of 14 and 16-inch MBPs and the M2 Pro Mac mini take up the next round of Apple attention, and whatever new scandals the YouTubers dream up or sniff out in amplifying something from molehill to mountain status in the quest for controversy-driven essential must-see clickbaitery.

Cheers,

Alex
 
  • Like
Reactions: Burnincoco
Very informative work. Anyone who doesn't see the relevance of tests like this are of course welcome to scroll-by this thread, without leaving a comment that the results are meaningless and not representative of the 'light tasks' people buy a two-grand computer for. But they won't. I fact I can hear their knuckles cracking already.
 
  • Like
Reactions: ilikewhey
Mike Boreham said:
I am not a great fan of benchmarking, and tend to only do it after getting a new machine, but in real life I would probably not notice a 15% slow down, and will occur very rarely on my machines.
Click to expand...
This is likely true for most of the benchmarks that are posted to those score repositories. :) The computer may see some excitement in it’s early days, and then it’s just… checking the network to see if an email has been pushed. Load another web page. Play another YouTube video!

For anyone that had grand plans of executing heavyweight long running applications on there MacBook Air... this confirms that one of the other systems are better suited to that task.
 
Mike Boreham said:
Completely agree. I haven't bought it to do 30+ continuous minutes at 100%, but I do like to test stuff particularly on controversial issues.
Click to expand...
You know what would be fun? Getting a few ice packs and running it while it’s sitting on them, swapping out as needed. It could be that it IS suited for the task but, as with many things, one must prepare!
 
  • Haha
  • Like
Reactions: Burnincoco and Mike Boreham
Great stuff! Assuming the sustained performance of the chassis is roughly 15-17 watts, this is the performance of the base M-series when severely thermally restricted. We can also see that M2 outperforms M1 by 9% at same restricted power consumptions (assuming that the tests were conducted under similar conditions). Not bad at all for a minor architectural tweak, especially considering that the memory subsystem likely consumes more power.
 
georgivelev said:
Thank you for the comprehensive benchmark. However,

This can be a hot topic only amongst people who have no idea this computer was not designed for this type of workload. It’s an interesting observation for the sake of curiosity but nothing more. Next, someone will perform the same test on the MBP and will claim for a few hundred bucks more you get like 5 times speeder computer.

Other conclusion: The M2 is faster by 9% after 30 minutes. Cool. It’s also like 20% more expensive. Dollar-per-productivity the M1 is better.
Click to expand...
Clearly, if that is all you're concerned with and don't care about the better display, speakers, etc. The M2 was a redesign for which you got a lot. Now if you don't care, well you don't care, and on a dollar per productivity basis, you could do way better getting an older used machine.
 
Mike Boreham said:
Running continuous Cinebench test for 30 mins is a common way of assessing CPU throttling, which is a hot topic on the fanless M1 and M2 MBAs. I have seen claims that the M2 throttles more. I tested my M1 MBA (512/16) in Jan 2021, and just upgraded to a M2 MBA(2TB/24), and reran the tests out of curiosity. This chart shows the results:

View attachment 2146738

Conclusions

1. The M1 throttles by about 10% over 30 minutes. The M2 throttles by 15% over 30 minutes. So yes the M2 does throttle more (in this test), but....

2. The M2 is 15% faster than the M1 after first run. It is still 15% faster after 10 mins, and 9% faster after 30 minutes.

Notes about tests

- The M1 was on Big Sur at that time and the M2 is on Ventura.
- I did not take any special measures before testing, only to make sure machine had been idling for some hours before starting and quitting all running processes in the Dock, but not background processes. It looks like run#14 on the M2 might have been affected by a background task (TM?) kicking in.
- The average CPU core temp on the M2 was 105C after run1 and 95 after 19. I didn't record the M1 temps in 2021. There was a step in temps between runs 11 and 12 on the M2.
- Because the M2 is faster it did 19 runs in the 30 mins and the M1 did 16, so the x-axes are different.

Comment

I am not a great fan of benchmarking, and tend to only do it after getting a new machine, but in real life I would probably not notice a 15% slow down, and will occur very rarely on my machines.
Cinebench tests the CPU not the GPU and the M2 has 2 extra cores compared to the M1.
Click to expand...
Idk if you’ve also tested it but I’m very curious about battery

Which of the 2 was the most efficient ?
 
Love-hate 🍏 relationship said:
well running the same tests, which dropped the most battery percentage?
Click to expand...

Yes I guess that would be interesting! I don't have my M1 MBA anymore so can't compare. If someone wants to repeat the M1 I will re-do the M2. Background software might be different but the great bulk of battery usage would be by Cinebench so wouldn't matter too much.
 
  • Like
Reactions: Love-hate 🍏 relationship
Mike Boreham said:
Great. I am going to repeat now. Since it takes a while to move off 100% I am going to start at my normal max of 70%. Can you do the same? Anything lower than 100% should be OK.
Click to expand...

Here is my M2 MBA test: 23% battery charge used during 32 minute test.

This is from iStat if you use it.

Interestingly on todays run the total drop in R23 score was 12.1%. Yesterdays test it was 14.7%

Look forward to M1 number if you (or anyone) can do it.

Screenshot 2023-01-24 at 07.04.13.png
 
Unregistered 4U said:
You know what would be fun? Getting a few ice packs and running it while it’s sitting on them, swapping out as needed. It could be that it IS suited for the task but, as with many things, one must prepare!
Click to expand...
I’ve done 30 minute Cinebench runs with both my M1 and M2 MBA with a laptop cooler (two fans on a plastic shelf) and I found that the M1 doesn’t throttle (or very little) and the M2 still does throttle to an extent but maintains a higher performance than the M1. If you frequently run high intensity tasks but still want the portability advantages of the M2 MacBook Air, a laptop cooler will greatly improve sustained performance.

Edit: More details here: https://forums.macrumors.com/threads/how-hot-does-your-m2-mba-run-doing-what.2356187/post-31437451
 
  • Like
Reactions: rmadsen3 and Mike Boreham
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back