I only use Safari. I was running in full screen. I do not find benchmarks to be useful. Actually using the device for regular stuff is the best way. I'm happy to run any YouTube suggested to test.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
2020 Air with i5 runs HOT!
- Thread starter wirewick
- Start date
- Sort by reaction score
Safari already includes some sort of cross site tracking prevention by default. Previously, I was using uBlock Origin as my content blocker but currently using AdGuard with all features turned on. I have a DNS-over-HTTPs client (cloudflared) installed on my Mac (through system wide network settings) to route my encrypted DNS queries and it also includes DNSSEC for validation. I previously experimented with Pi-Hole but my current setup feels good enough wherever I am.
It'a not a benchmark. Watching 4K/8K YouTube videos in safari will cause YouTube to automatically downgrade them to 1080. In other words, YouTube will only send the 1080P version so you're not stressing the CPU anymore than watching any other 1080P only videos. Try watching those same videos under chrome. Even on the MBA screen you'll be able to see a quality difference.
For me, 10.15.5 fixed that problem. Data clearly backed that up, not to mention fan noise.
What I meant to ask was, were you monitoring fan speed and CPU temps before and after the update?
It runs full blast to maintain the highest clock it can *at that temp* with the cooling solution it does have (whether it is a "good" cooling solution or not is besides the point).
100C operating temp whilst boosting (i.e., above base-clock) is in spec for the CPU, thus "normal". If you don't believe it, look up the CPU specs on ark.intel.com and read up on how intel's various CPU boost technologies operate.
I'm not saying it is desirable, I'm saying the CPU is designed to operate at temps up to 100C and it will self-throttle if needed to drop below that.
Apple are running the fan to try and run it faster than it would if fan-less at that temperature. The MBA maintains faster than base clock at 100C; the base of 1.1-1.2ghz is all you are guaranteed from intel and the MBA runs faster than that under sustained load.
Apple could have made the MBA entirely fan-less (Just like the 12" Macbook was) and it would still run at 100C under load, just at a lower clock speed.
Yes, the MBP is faster. It has a higher wattage CPU and a higher spec cooling system to suit.
The MBA uses a 9 watt CPU that Apple have basically TPD-upped (as per intel spec) to 10 watts and included a fan to help it run faster than base clock. This means better performance than they would have otherwise got without a fan, and without the higher TDP rating. But fan noise and heat.
Last edited:
The i3 MBA will stay above its base clock speed in a sustained CPU/GPU load.
However, the i5 MBA will throttle down after awhile to about 400MHz, according to the Notebookcheck review.
That is not my experience, my i7 will maintain around 2 Ghz with the fan going full blast at 100% CPU. If your i5 is failing to maintain base clock or faster, RMA it.
My MBA i5 happened to have 100 degree problem too.
I did an easy custom heatpipe mod to solve the problem.
Here is the original post.
Last edited:
Nor mine. My i5 maintains 2.1-2.2Ghz (multi core) and 3.3-3.4Ghz (single core) under sustained loads, after a gradual ramp-down to those speeds over the course of a minute or two.
The links above from Notebookcheck appear to be from Windows running on the MBA. And the review was posted soon after release of the MBA, running early-doors unoptimised BootCamp drivers. Not exactly a typical usage scenario from which most people would realistically draw the 'throttles to 400mhz' conclusion above.
What kind of sustained load did you put on your machine, and how long did you test it?
Notebookcheck found the i5 MBA went down to about 400 MHz in around 23 minutes of CPU/GPU load. I think they benchmark in Windows because if they use the same tests on all notebooks (Windows and Mac), people can get an idea of performance when comparing machines. I do not believe any software has changed the MBA's performance since release. It makes sense that the CPU will throttle down to avoid going over its maximum allowed temperature.
Last edited:
Intel Power Gadget iirc. I ran a couple of benchmarks the day after I got the machine, was happy with what I saw, then uninstalled them and got on with using it for real life stuff. It's been genuinely great, for both work and play.
As I said, hardly a typical usage scenario for a MBA; in terms of producing a meaningful benchmark it's frankly absurd. If you're going to quote worst-case stress test results at least quote the ones done using the machine's native OS, where it has the corrrect drivers and optimised CPU P-states and fan curves. Otherwise it's like benchmarking an Nvidia 2080Ti running nouveau VGA drivers, then declaring it a slow, buggy POS.
That probably explains the difference between what you saw and the tests in the Notebookcheck review.
There is no indication that the MBA is missing any drivers, is missing the high-performance CPU P-states or has a poor fan curve in Windows.
While I agree performance can be better in some benchmarks under macOS, optimization in macOS does not explain the magnitude of difference here. The i3 MBA does not thermal throttle under its base clock in Notebookcheck's tests. The i3 MBA sustains performance at 1.5 GHz after over an hour of CPU/GPU load. However, under the same conditions, the i5 MBA throttles to about 400 MHz in under 23 minutes. CPU temperature aside (both run at 99-100 degrees C), this indicates to me that the chassis and cooling is generally sufficient for the i3, but not the i5.
To return to the OP's issue, watching full-screen, 1080p60 video for awhile (without hardware decoding) can put a sustained load on the CPU and GPU, so this throttling might explain why he had trouble.
Well, there seem to be some substantial difference after all at least when it comes to power management (and where there is one issues there may be more).
Investigation: Why is Bootcamp so much worse on the 2020 Air?
I've owned a 2020 i5 MBA and currently own a 2020 i3, and Bootcamp has been a nightmare on both compared to the 2019. Mostly in terms of battery life - you can watch it tick down before your eyes. But there are also issues in regards to realising the Ice Lake/Iris Plus performance boosts, making...
forums.macrumors.com
Beside that, as a Mac buyer, Im more interested how it performs under OS X than any other OS.
The MacBook Air internal design was made for the switch to Apple (ARM) CPU, and so I will wait. Imagine a MacBook Air with the same performance as Intel, but with no fan needed, and only getting mildly warm to the touch! And with a GPU that runs circles around the Intel integrated one. The move to Apple's own SoC can't come quickly enough for me.
Well, there seem to be some substantial difference after all at least when it comes to power management (and where there is one issues there may be more).
Investigation: Why is Bootcamp so much worse on the 2020 Air?
I've owned a 2020 i5 MBA and currently own a 2020 i3, and Bootcamp has been a nightmare on both compared to the 2019. Mostly in terms of battery life - you can watch it tick down before your eyes. But there are also issues in regards to realising the Ice Lake/Iris Plus performance boosts, making...
Beside that, as a Mac buyer, Im more interested how it performs under OS X than any other OS.
That thread is talking about how power draw at idle is higher under Windows, because some idle sleep states of the processor are missing in Bootcamp.
This thread is talking about video playback taxing the processor, during which the processor uses high-performance sleep states, which are available in both Bootcamp and macOS.
Last edited:
This is an assertion without any argument. *What* explains the difference?
You mean, other than the fact that the CPU throttles to 400Mhz? And that it simply does not do this when run under MacOS, and by some order of magnitude? I'd suggest that this absolutely *is* an indication that the premise of testing under Windows as per Notebookcheck is a flawed one. If the Bootcamp performance is not unnaturally skewed, then why not repeat figures from CPU tests when running under MacOS? They should be equally valid, no?
Last edited:
I am not sure I understand.
I thought you used the built-in tests in Intel Power Gadget. Notebookcheck did not do that. They used Prime95 to test the CPU and FurMark to test the GPU at the same time. This approximates a simultaneous CPU and GPU load.
Different benchmarks can perform different operations on the CPU and the GPU, so they can provide different results. In addition, the results might change if the tests run for different periods of time.
The benefit of running the same benchmarks on multiple notebooks is people can make comparisons between them. For example, we can see how the i5 MBA performed in these benchmarks compared to the i3 MBA or even MBA competitors, like the Dell XPS 13 and the Thinkpad X1 Carbon, in their respective reviews.
So, when Notebookcheck ran Prime95 and FurMark at the same time for about 23 minutes, the i5 MBA CPU throttled down to about 400 MHz. The i3 MBA sustains performance at about 1.5 GHz after over an hour under the same conditions. This is unlikely to be the result of an undiscovered Bootcamp bug that affects the i5, but not the i3. It is more plausible for the reason to be the i5's higher power draw generates more heat than the i3. So, the i5 taxes the cooling system more than the i3.
The reason we cannot run Prime95 and FurMark at the same time on macOS is because FurMark is a Windows program.
The OP's issue was with the heat and fan noise when playing 1080p60 video. This makes sense given the benchmark results. Video decoding and playback without a hardware decoder can lead to a prolonged CPU and GPU load, which leads to high temperatures and fan noise.
Someone said earlier that the MBA stayed above base clock under sustained load, even though it ran at 100 degrees C. I responded with the benchmark results for the i3 and the i5.
You responded with Bootcamp Windows benchmarks....meanwhile 2 users who actually own the machines in question - replied with their own benchmarks made under MacOS that suggest the benchmarks you're quoting are an order of magnitude wide of the mark (and I have speculated that this is because they're made while running under Bootcamp).
Last edited:
Notebookcheck uses a standard stress test on notebooks they review. They do not do this in macOS because the GPU benchmark program is FurMark, a Windows program.
I do not know how those two benchmarks in macOS compare to running Prime95 and FurMark at the same time. Further, I do not know how long those tests ran. This is why it is valuable to run the same benchmarks on machines we want to compare.
Last edited:
Just my humble opinion, but running benchmarks in an OS I'll never run on my MBA kinda seems pointless.
What's kinda even more pointless is running heavy sustained CPU/GPU load benchmarks on a computer that's clearly intended for intermittent bursty CPU/GPU utilization.
If someone needs to frequently/continuously perform heavy sustained CPU/GPU loads, they need to buy a MBP not a MBA.
The fact that screwdrivers make poor hammers doesn't make them bad - it just means you bought the wrong tool for the job.
What if the load is caused by video playback? If the video decoding is done by a CPU without a hardware decoder, it will load the CPU. Depending on how big the video window is, rendering will load the GPU.
The OP is just trying to play 1080p60 video without high temperatures and fan noise. That should not require a MBP.