2020 Air is thermal throttling - yes, really. (CS:GO & bootcamp & other tests inside)

[mick2]

I agree that the OPs results look strange. FWIW I have an identical machine 2020MBA i5/16/500.

I don't run bootcamp, and IMO any results under it will likely be suspect until drivers etc have settled. IIRC the same happened in the early days of the 16 MBP.

Under MacOS and using Intel Power Gadget, my machine idles at 33-34degC. Idle freq is 0.9Ghz.
Single core max test: Turbo Boosts to 3.47Ghz and stays there as long as the test is running. Temps here eventually reach the early 70'sdegC and stay there. Can't hear the fan at all whilst running.

All cores at 100% test: CPU freq starts at 3.2Ghz. Stays there for approx 60s as temps rise. When temps hit 99degC the turbo boost freq falls over the next ~3-4mins and levels out at 2.4-2.5Ghz. Then stays there. I can barely hear the fan if I put my ear to the chassis. No jet-engines or hurricanes. At no point does it even approach the Intel baseline of 1.1Ghz or lower (ie the definition of thermal throttling).

I am seeing nothing like the OPs complaint; sounds like his machine may simply be faulty.

 

[lllc]

So if some are cool it means maybe the heatsink is not properly attached, or its missing thermal grease. My fan dont start even it runs super hot. And the fan is not cooling the cpu at all anyway.

 

[thermals]

I agree that the OPs results look strange. FWIW I have an identical machine 2020MBA i5/16/500.

I don't run bootcamp, and IMO any results under it will likely be suspect until drivers etc have settled. IIRC the same happened in the early days of the 16 MBP.

Under MacOS and using Intel Power Gadget, my machine idles at 33-34degC. Idle freq is 0.9Ghz.
Single core max test: Turbo Boosts to 3.47Ghz and stays there as long as the test is running. Temps here eventually reach the early 70'sdegC and stay there. Can't hear the fan at all whilst running.

All cores at 100% test: CPU freq starts at 3.2Ghz. Stays there for approx 60s as temps rise. When temps hit 99degC the turbo boost freq falls over the next ~3-4mins and levels out at 2.4-2.5Ghz. Then stays there. I can barely hear the fan if I put my ear to the chassis. No jet-engines or hurricanes. At no point does it even approach the Intel baseline of 1.1Ghz or lower (ie the definition of thermal throttling).

I am seeing nothing like the OPs complaint; sounds like his machine may simply be faulty.

Thanks for uploading those numbers. I'm not sure which test is the single core one, but I ran the multicore test to see if anything is wrong with my Air.

• Your 33-34C idle temp: mine never seen sub 40C levels, it's idling at about 50C with the fan off. Idle freq is the same as yours
• All cores at 100% test: starts at max GHZ and starts to slowly decline over time, just like yours. But:
  • it doesn't even hold for 60s, drops to 2.4ghz in about 40s
  • fan is about 4500-5000 rpm, barely audible to me
  • cpu performance drops to about 2ghz and stays there, more or less

Attached screenshots with the details. Perhaps my Air is worse than the rest out there...
[automerge]1587568332[/automerge]

Just to be safe, I repeated the test again after I exited all the apps that were running in the background. Results were about the same.

Edit: I just realized that attachments were merged to my previous post, the 3rd image is the latest, repeated test. The first 2 is made of the same test, at the beginning and at the end.

 

[mick2]

I'd say your figures are a little worse than mine, but IMO most of the difference would be unnoticeable in practice:

- My idle is lower - am in the UK - could just be down to colder climate.
- I timed my all cores test; tbf its more like 40s before it hits 99degC. It decays to ~2.4ish sustained; yours decays to 2.1ish sustained (your 3rd screenshot). Outside of benchmarking this would be unnoticeable.

(btw you do need to close down all background apps etc before running a benchmark)

Both are pretty darn respectable speeds for a multi-core 100% sustained stress test.

As a comparison: my Thinkpad T480s i7-8550u Turbos to 4Ghz for about 8s then drops to about 2.2Ghz multi-core sustained, along with a fan that really *does* sound like a jet engine taking off. It hits this 2.2Ghz figure after max 30s of running sustained.

My paper thin MBA stays at 2.4Ghz speed whilst being almost silent, sipping battery and running rings around the i7-8550u performance-wise because its 2.4Ghz are 10th gen Ghz, not 8th gen. Same applie for your 2.1Ghz.

Bottom line: although our speeds are a little different, you're getting nothing close to the 'thermal throttling' you're claiming in your thread title, your OP, and the posts you've made across reddit linking to this thread. Just trying to add a little perspective here.

 

[nylon]

Just to be safe, I repeated the test again after I exited all the apps that were running in the background. Results were about the same.

Edit: I just realized that attachments were merged to my previous post, the 3rd image is the latest, repeated test. The first 2 is made of the same test, at the beginning and at the end.

These power gadget graphs are nothing like the first ones you posted. All seems normal according to these. No throttling under base clock. In fact they all show that the CPU can sustain a small boost over base clock at sustained CPU load.

 

[iMacDragon]

The only consistent result is in your first picture under windows 10 (apart from PkgPwrLimit0 indicating 100w) where CPU utilization indicates 100%. This only indicates that the power management driver has not been updated for the MBA under bootcamp.

Apple seem to not give one fig about power management under windows.. last three macbooks I've owned have had horrendous power management under windows compared to macos.

 

[mick2]

These power gadget graphs are nothing like the first ones you posted. All seems normal according to these. No throttling under base clock. In fact they all show that the CPU can sustain a small boost over base clock at sustained CPU load.

It's actually more than a small boost; OP is getting almost double the base clock with sustained multi core performance (1.1Ghz base, op gets 2.1Ghz sustained).

 

[thermals]

These power gadget graphs are nothing like the first ones you posted. All seems normal according to these. No throttling under base clock. In fact they all show that the CPU can sustain a small boost over base clock at sustained CPU load.

Keep in mind that this is a synthetic test, and far from everyday usage. The only thing that simulates open browser tabs, Office apps, and Slack /etc. is actually opening them. That's what I did in my first test.

The extra addition was a tool that I use for my regular job (also mentioned it in the first post), it basically simulates opening many websites quickly after each other, and that's when the CPU began throttling below base frequency. And that shouldn't happen, no matter how hard you push it. And I didn't even begin experimenting with more CPU-heavy stuff like video editing.

Edit: if you take a second look at the graphs, even the synthetic test was worse than the one from another person in this thread. So something might be wrong with my machine indeed, if those other numbers were correct.
[automerge]1587578010[/automerge]
If you still believe that the synthetic test provides valid results, take a look at this:

https://www.reddit.com/r/macbookair/comments/g63pgy

Performance is worse on average but the temps are still high. Keep in mind that the red line on the graph is the performance that the CPU would need, and the line below that is what the CPU can actually provide, so it's not because the system wouldn't need more power.

 

[nylon]

Keep in mind that this is a synthetic test, and far from everyday usage. The only thing that simulates open browser tabs, Office apps, and Slack /etc. is actually opening them. That's what I did in my first test.

The extra addition was a tool that I use for my regular job (also mentioned it in the first post), it basically simulates opening many websites quickly after each other, and that's when the CPU began throttling below base frequency. And that shouldn't happen, no matter how hard you push it. And I didn't even begin experimenting with more CPU-heavy stuff like video editing.

Edit: if you take a second look at the graphs, even the synthetic test was worse than the one from another person in this thread. So something might be wrong with my machine indeed, if those other numbers were correct.

Synthetic or not, I'm comparing to your original results showing the CPU throttling to below base clock under nothing approaching full CPU loads. So the first set of results (under MacOS) would indicate that you have faulty machine and the second set indicates you don't.

 

[thermals]

Bottom line: although our speeds are a little different, you're getting nothing close to the 'thermal throttling' you're claiming in your thread title, your OP, and the posts you've made across reddit linking to this thread. Just trying to add a little perspective here.

I'm not sure if you've read through the whole post, but in my first post the thermal throttling is there. Not nearly as bad as under Win 10, but the CPU clock falls below the base frequency under load. It's nice that the synthetic test performs better, but in real life I don't use the laptop to perform tests, I have it to actually use it, and that's what I was trying to simulate.

 

[nylon]

I'm not sure if you've read through the whole post, but in my first post the thermal throttling is there. Not nearly as bad as under Win 10, but the CPU clock falls below the base frequency under load. It's nice that the synthetic test performs better, but in real life I don't use the laptop to perform tests, I have it to actually use it, and that's what I was trying to simulate.

You amended the last post I replied to after I replied so I didn't see your Chrome tab methodology. To be honest Chrome is very poorly optimized on MacOS at the moment. Even on my 15" Pro it chews through battery like nothing else. Also your methodology of 40+ tabs bears no reality to real world usage.

Regardless of all this, your machine should not throttle under base clock at any point in MacOS. Your own graphs show synthetic tests putting more stress on your CPU than the non synthetic tests yet only your non synthetic tests show throttling below base clock. This logically does not make any sense. A synthetic test still puts a real world computational load on your CPU like in Cinebench. It's only called synthetic because it does not mimic normal usage. It's a stress test. Hence why I think you probably have a faulty machine.

 

[thermals]

Your own graphs show synthetic tests putting more stress on your CPU than the non synthetic tests yet only your non synthetic tests show throttling below base clock. This logically does not make any sense.

I agree with you on this completely, this does not make sense at all. But unfortunately this is what is happening.

Some open browser tabs are closer to everday usage than running any benchmark, I hope you agree with me on this - of course you can swap some tabs with Excel, Word, Slack, etc. to get even closer so there's always more room to find situations to measure performance.

But at the end of the day, if I simply do what I used to do on my previous Macs and the performance is no better, there's just no point spending this much money on a new Mac. Just like you said, it might be that there's something wrong with my 2020 Air, but seeing everyone else's experience in this forum I doubt it could be so much different that I could justify spending money on the new Air instead of getting an older Pro or maybe even an Air with a similar config.

If you like the new Air, that's completely fine. I just hope that this thread and the benchmarks are useful who was considering ordering the new Air and have a similar use case than mine.

 

[nylon]

I agree with you on this completely, this does not make sense at all. But unfortunately this is what is happening.

Some open browser tabs are closer to everday usage than running any benchmark, I hope you agree with me on this - of course you can swap some tabs with Excel, Word, Slack, etc. to get even closer so there's always more room to find situations to measure performance.

But at the end of the day, if I simply do what I used to do on my previous Macs and the performance is no better, there's just no point spending this much money on a new Mac. Just like you said, it might be that there's something wrong with my 2020 Air, but seeing everyone else's experience in this forum I doubt it could be so much different that I could justify spending money on the new Air instead of getting an older Pro or maybe even an Air with a similar config.

If you like the new Air, that's completely fine. I just hope that this thread and the benchmarks are useful who was considering ordering the new Air and have a similar use case than mine.

LOL! No one is forcing you to keep your MBA. If it's faulty then return it, or of you simply don't think its worth the money then return it. You have multiple options available to you. No one else has shown throttling below base clock in any scenario under MacOS. Only you. Your use case is 40 tabs on Chrome which is not even your own normal use case. It's test that you ran. The problem here is that you are trying to use your experience to insinuate that there is a problem with all the machines when in fact only your results seem to be the outlier.

 

[esphil]

Theres so many mixed reviews on these things between people. I don't know what to believe anymore. I haven't seen any negatives hardly on the site reviews like Amazon/Bestbuy. Perhaps there is a hardware issue/software/firmware issue between these machines? Perhaps newer BTO have issues fixed? I'm not sure. I'll thoroughly test drive mine and report back when it arrives. There is generally a explanation why something is occurring. There is no way a 2020 Apple laptop should be running worse than a 2014 version of it. Its not just 2 dudes at Apple coming up with a design and then starting production on it. The amount of R&D that goes into new products is immense. I thought this was a good video showing some workload.

 

[thermals]

LOL! No one is forcing you to keep your MBA. If it's faulty then return it, or of you simply don't think its worth the money then return it. You have multiple options available to you. No one else has shown throttling below base clock in any scenario under MacOS. Only you. Your use case is 40 tabs on Chrome which is not even your own normal use case. It's test that you ran. The problem here is that you are trying to use your experience to insinuate that there is a problem with all the machines when in fact only your results seem to be the outlier.

Wow, no need to be so aggressive. I only have 1 Macbook Air, and since these are not exactly handmade, so the machines should be similar, that's why I made the assumption that these are not as good as they can be. If you don't agree, that's fine, this is what forums are made for.

 

[nylon]

Wow, no need to be so aggressive. I only have 1 Macbook Air, and since these are not exactly handmade, so the machines should be similar, that's why I made the assumption that these are not as good as they can be. If you don't agree, that's fine, this is what forums are made for.

I just responded in kind to your passive aggressive, "if you like it, that's fine" comment. It has nothing to do with 'liking' I'm responding to your data. I have no stake in your ownership of the machine which is why I have stated from my initial responses to you that I thought your machine was faulty and have stated that you should return it for any reason you wish, faulty or otherwise. Your thesis has been that the machine 'throttles' below base clock under cpu loads that do not come close to 100% CPU utilization (only for you) which would indicate that its faulty, not that the machine isn't as 'good as it can be' (which is a subjective assessment). Those are two different reasons.

 

[deeddawg]

Theres so many mixed reviews on these things between people. I don't know what to believe anymore.

Just remember - youtube reviewers exist to make money from people clicking / sharing / watching their video reviews.

Hyperbole / controversy generates far more revenue than boring/fair/balanced material.

Get your system. Get it set up and let it have a couple days to settle in - leave the lid up.

Then use it as you normally would use it.

If it works for you and you're happy, great. If it doesn't, return it and get something more suited to your needs.

 

[unoporfavor]

TLDR: 2020 Air is good for using Chrome and super lightweight tasks, but not at all for ANYTHING that stresses the cpu longer than 4-5 sec.

I was lurking here in the forums for almost a month now. I was waiting to upgrade my 2015 Air for years, and ordered the new Air right when I realized the keyboard seems to be reliable, finally. And the i5 was the icing on the cake - even if it's far from the Pros, I thought that would be good enough for years.

Oh boy how wrong I was.

Here's what I found out within 24 hours of having the new Air. Some of the charts are photos not screenshots, I'm sorry about that.

My config is i5/16gb/512gb SG.

I intended to use the Air as a work machine: this means 20-30 Chrome tabs, a few spreadsheets and Word docs open, Spotify is running in the background and occasionally an other tool that's specific to what I do (Screaming Frog if you know it).
I also wanted to be able to play CS:GO occasionally on the lowest possible settings. For the context, it's almost a 10 year old game, and the 2020 Air is a laptop with an above-average price, especially in the EU.

So no 4K video watching or video editing or Geekbench or anything hypothetical.

In the past 5 years I've used a 2015 13" MBP, the 2015 Air I mentioned, and a 2017 touch bar MBP. ALL OF THESE were able to handle my usual workload without the fans being too loud. The Air was noticeably slower, but quiet at least. CS:GO was running fine on the MBPs, but too slow on my old Air.

Now onto the 2020 Air. First I have to admit, the machine is fast until I do nothing else just open browser tabs. But in my opinion, the "not planned to use for heavy workload" does not mean that you shouldn't be able to do ANYTHING CPU-demanding at all.

Here's what I experienced:

First I set up bootcamp to be able to try out CS:GO in better circumstances. Again, this is something that my old 2015 Macbook Pro was able to handle (and it wasn't much more expensive back then than this Air now!) so I don't think I ask too much.

View attachment 907893


Note how the CPU drops below the base frequency. And not even on a constant level, but it does it like that. In-game this meant that I had 10-15 fps and 70-80 for a few seconds.

Here's what the result looks like with a better benchmark program:

View attachment 907894

Yup, 400 mhz. "But this surely must be some other error, not thermal throttling!"

Well, this is what the sensors said about it:

View attachment 907895

"Ok, but this is because of bootcamp, it must be better under MacOS" - that's what I thought at first.

By the way the bootcamp install was a clean install with fresh setup, bootcamp drivers installed as intended, so no idea what could be this wrong.

But I gave it a shot under MacOS as well:

View attachment 907896

No thermal throttling indeed. But do I want a CPU that "boosts" to 1.2 ghz from 1.1?

(I have the log files saved with more precise data, let me know if you want it)

To the common argument that says "bUt dis Is a LiGHt uSE lapTOP" -> light use does not mean that it should perform worse under load than my 5 years old MBP - for about the same price.

But I thought that this must be just a CS:GO bug or something similar, so I did another test. Here's the performance under my regular workload, under MacOS:

View attachment 907904

Yep, it drops under the base frequency until the CPU cools down a bit. So it IS thermal throttling.

Geekbench tests are a lie. It can reach maximum performance for a few seconds, but it makes no sense if the constant performance is just about the bare minimum.

To sum up, I spent more money than the original price of my old 2015 13" MBP to get worse performance. I hate the touch bar but I'll send the Air back and wait for the 14" MBP.

I completely understand that I shouldn't expect Pro performance in an Air, and I would be fine with that. But I don't think I should be satisfied with bad performance under load. Just worse than Pro should be enough.

If you want me to test anything else you haven't found in Youtube reviews, let me know! This was just what I could experience within the first 24h. And of course, all the tests were done after updates and indexing finished.
[automerge]1587464905[/automerge]
This is by no means a professional test, this is just what I wanted to see in reviews but no one tested it like that. I hope some of you will find it useful.

Sorry you are just not comparing like-for-like -- a MacBook Pro that would have been twice the cost as the current MacBook Air (in net value terms) .
Other than they have the same sticker price but 5-6 years apart there is nothing here that is valid. Sorry bud
[automerge]1587787365[/automerge]

Sounds like I'll be sending mine back when I receive it. I guess I'll wait for the new mbp or get a refurb 16, this has been extremely disappointing overall reading all of these reports about this laptop after finally deciding to order a machine after using a iMac for the last 10 years. The fact that base clock speeds in 2020 are at 1.1ghz seems like a huge step back.

Try using it before looking to send it back. Honestly these people are so worried about temperatures and throttling they have an agenda before they start testing!
I have an i3 MBA. I was considering upgrading to i5 if there was any performance concerns but it's performed absolutely impeccably.
The battery life could be longer - coming from a MacBook 12", but the performance is amazing. Really fast at launching everything. EVen the front-facing camera that everyone criticises is far better than most of the people on the video calls.
I made over 4 hours CONTINUOUS video calling last night (Friday night), and it got warm, but never faulted and more importantly was silent or as near silent you had to put ear down to keyboard to hear the fan.
Brilliant equipment.
All I'm saying is do the testing yourself. None of this stuff you see is really quality empirical data.
Cheers mate

 

[esphil]

Sorry you are just not comparing like-for-like -- a MacBook Pro that would have been twice the cost as the current MacBook Air (in net value terms) .
Other than they have the same sticker price but 5-6 years apart there is nothing here that is valid. Sorry bud
[automerge]1587787365[/automerge]

Try using it before looking to send it back. Honestly these people are so worried about temperatures and throttling they have an agenda before they start testing!
I have an i3 MBA. I was considering upgrading to i5 if there was any performance concerns but it's performed absolutely impeccably.
The battery life could be longer - coming from a MacBook 12", but the performance is amazing. Really fast at launching everything. EVen the front-facing camera that everyone criticises is far better than most of the people on the video calls.
I made over 4 hours CONTINUOUS video calling last night (Friday night), and it got warm, but never faulted and more importantly was silent or as near silent you had to put ear down to keyboard to hear the fan.
Brilliant equipment.
All I'm saying is do the testing yourself. None of this stuff you see is really quality empirical data.
Cheers mate

I got mine today and I've been enjoying it so far. I didn't hear any fan sound and didn't feel it getting hot. I eventually got the temp viewer pg pro just to see, and I was really pushing it and even when it did hit 100, the laptop felt cool to the touch and the bottom only felt warm. No more warm than any other Mac laptop I've had. The main thing that has been messing with me is the screen size. I haven't used this screen size in many years, and I was debating if I want the portability of this with the 13 or get the 16. But so far the performance has been good with everything I do. I agree abou the camera too, it seems fine to me. The camera is all about lighting too, if you are in a dark room it will be grainy. I really like the wedge design though, that to me make a huge difference. I'm waiting to get a usb-c hub in the main to test with my screen because this was originally going to be for replacing my old iMac.

 

[RiaKoobcam]

TLDR: 2020 Air is good for using Chrome and super lightweight tasks, but not at all for ANYTHING that stresses the cpu longer than 4-5 sec.

Hey man, I found similar results to your Windows testing in my own testing. Of particular note was the power draw. I did a write up here if you're interested, but no, you're not crazy.

For anyone who needs or wants Windows and is considering an MBA 2020, I'd warn them off it. The performance hits and insanely high power draw (even compared to the 2019 MBA aren't worth it). Full write up here.

 

[Spectrum]

I don't use Bootcamp, but the idea of having a machine that is already throttling under current workloads doesn't seem great for a machine I'd like to perform well now and into the future.

I really want to replace a 2014 13 inch i7 MBPro. I really don't like the Touch Bar on the new MBPro models, so in many ways, the Air seems spot on. But if I want a machine to last 5-6 years, it looks like I am going to have to wait for the new 13/14" MBPro (and sadly will therefore have to live with/tolerate the Touch Bar.)

What I really don't like is the trend for intel to be launching modern CPUs with incredibly low base/TDP frequencies. I'd essentially be replacing a 2014 dual core model with a "guaranteed" TDP/CPU combination of ~25W/3 Ghz, with a 2020 quad core model with a "guaranteed" TDP/CPU combination of ~10W/1.2 Ghz.

 

[thermals]

Hey man, I found similar results to your Windows testing in my own testing. Of particular note was the power draw. I did a write up here if you're interested, but no, you're not crazy.

For anyone who needs or wants Windows and is considering an MBA 2020, I'd warn them off it. The performance hits and insanely high power draw (even compared to the 2019 MBA aren't worth it). Full write up here.

I just saw your update, thanks a lot for testing it!

And now that the MBP refresh is out there with worse specs than the Air, I really hope that the “get a Pro if you want Pro performance” crowd will calm down a bit. Proper cooling is a must.

 

[deeddawg]

And now that the MBP refresh is out there with worse specs than the Air, I really hope that the “get a Pro if you want Pro performance” crowd will calm down a bit. Proper cooling is a must.

Why would you think the advice to buy the computer best suited to your workflow would be any less valid today that it was last week?

What's available on the market is what's available on the market.

Vote with your wallet. Choose whichever is best suited to your workflow.

My MBA remains quiet and cool - spent the evening rating/labelling/culling through 600 photos, with a bit of email and facebook/macrumors/online-shopping. Never got warm on my lap and remained silent.

 

[nochese]

But I thought that this must be just a CS:GO bug or something similar, so I did another test. Here's the performance under my regular workload, under MacOS:

View attachment 907904

Yep, it drops under the base frequency until the CPU cools down a bit. So it IS thermal throttling.

Geekbench tests are a lie. It can reach maximum performance for a few seconds, but it makes no sense if the constant performance is just about the bare minimum.

To sum up, I spent more money than the original price of my old 2015 13" MBP to get worse performance. I hate the touch bar but I'll send the Air back and wait for the 14" MBP.

I completely understand that I shouldn't expect Pro performance in an Air, and I would be fine with that. But I don't think I should be satisfied with bad performance under load. Just worse than Pro should be enough.

Your macbook air is performing as intended. A 3D game needs both CPU and GPU processing. The chip prioritizes GPU power when gaming. It only has an energy budget of 10 W. Even a macbook pro with better cooling and a 28 W chip will drop below base CPU frequency while gaming with the integrated graphics. I've experienced this myself.

On the other hand, you only had one picture of your regular, non-graphics heavy workload. As far as I can see, everything is working as it should. You're getting an average frequency that's above the base frequency of your i5 chip. The only places it dips below 1.1 GHz is because the requested workload drops in those instances, not because, as you claim, it is throttling. Intel promises base frequency performance; turbo is a temporary bonus. That's the reason why these 15 W parts have base frequencies that are so low. Also, we see the the core temperature is high, but that's standard operation for intel's laptop CPUs under load for the past decade basically.

In your screenshot, the utilization hovers around 50%. I'd be curious to see what happens under an even heavier workload, like a code compilation that uses 8 threads or something similar. If you can show that it can't maintain 1.1 GHz under 100% CPU, non-AVX workload, then that would show an issue.

 

[AirBud]

I don’t consider video games to be a “workload”. Games by design try to get as much power as they can, even old games (Starcraft) end up running at 100s of FPS if allowed

If high performance during video games while maintaining low temperature is important to you, you should be buying a laptop with an actual GPU onboard (16” MBP).