At most it would affect swap memory, but no one is benchmarking that.
No one who is dealing with large or many files is sticking them on a tiny 256GB internal drive, which is at least half full with OS and application files.
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.
Did Apple take a wrong turn?
- Thread starter Wando64
- Start date
- Sort by reaction score
No one who is dealing with large or many files is sticking them on a tiny 256GB internal drive, which is at least half full with OS and application files.
But we are benchmarking that, it is a key component of the performance. It influences benchmarks of the M2 with 256GB - the deltas in performance are there relative to 512GB and greater than they were with the M1 MacBook Pro.
With a few tabs open, the M1 MacBook Pro proves faster than the M2 MacBook Pro in a 5-minute 4K HEVC export test.
It doesn't end there, either, look at the thread.
I think ultimately Apple should be giving 16GB standard on the M2 MacBook Pro or M2 MacBook Air, even with a mild price increase on either one for the base model. There's just no excuse in shipping a base model with 8GB of RAM and a slow, small drive for 1199 or 1299, inflation and M2 or not.
For reference Asus' Zenbook with AMD's Zen 3+ 6800U is fitted with 1TB of a PCIE 4 SSD and 16GB of 6400MT/S LPDDR5, a 2.8K, 400 Nit OLED display with a 90Hz refresh rate, etc for $1299.
Yes, the M2 (or the expenditures behind Apple Silicon) and Apple's margin demands make us unlikely to see a similar price point for these specifications (and Rembrandt is on TMSC's N6 with less cache, etc) and an M2, but given the display on the M2 MBA or M2 MBP, I really don't think it's too much to ask for 16GB of LPDDR5 or 512GB SSD's at the base models, here. It's 2022.
Last edited:
Sure, and I've been a regular advocate on these forums for the need for better ST performance, since most productivity and scientific apps continue to be ST only.
But I think if you meant ST you should have said just that, instead of using language like an "M2 laptop approaches the performance of H-series workstations for some tasks" which, in the context of this discussion (which is about MT performance), and in the context of what workstations are (MT-focused machines), misleadingly implies the M2 approaches H-series workstations for some MT tasks.
It's like joining a discussion about off-roading, and saying "this car's performance approaches the capabilities of dedicated off-road vehicles in certain tasks", when what you really mean is just that it equals them on pavement.
I know you like to say how much better AS is than Intel's offerings, and I agree. At the same time, I think we should endeavour to stick to objective language in describing their differences, as neither of us work for Apple's marketing dept.
Also, I suspect that R will be faster for many numeric calculations on a 12900HK than an M2, at least for a while, until Apple (or someone else) creates an equally-performant replacement for the Intel Math Kernel Library:
Accelerate R Code with the Intel oneAPI Math Kernel Library
Improve R Performance without Modifying Code
Last edited:
Sure, and I've been a regular advocate on these forums for the need for better ST performance, since most productivity and scientific apps continue to be ST only.
But I think if you meant ST you should have said just that, instead of using language like an "M2 laptop approaches the performance of H-series workstations for some tasks" which, in the context of this discussion (which is about MT performance), and in the context of what workstations are (MT-focused machines), misleadingly implies the M2 approaches H-series workstations for some MT tasks.
It's like joining a discussion about off-roading, and saying "this car's performance approaches the capabilities of dedicated off-road vehicles in certain tasks", when what you really mean is just that it equals them on pavement.
I know you like to say how much better AS is than Intel's offerings, and I agree. At the same time, I think we should endeavour to stick to objective language in describing their differences, as neither of us work for Apple's marketing dept.
Oh, I agree with you absolutely, and yeah, I shouldn't have been using manipulative language. It's just that I get a bit annoyed reading all this doom and gloom based on multicore Cinebench with very little relevance to real-world experience.
Don't discount the AMX units
But of course, if you do heavy matrix manipulation, the base consumer M-series is not the machine you want to use.It's been a few months since I looked into it, and I haven't tried it myself yet since most of my work the past few months has been on Python projects, but IIRC Apple Silicon users can configure R to use Apple's Accelerate.framework, which contains a version of LAPACK and veclib optimized to take full advantage of the SOC's matrix math units. From what I remember, the speed benefits were pretty impressive.
EDIT: Just went through my comment history and found my research on this, hopefully it's helpful to someone: https://forums.macrumors.com/thread...-for-merged-mega-thread.2339841/post-31024607
Has anyone been able to find a link to Apple's own power consumption measurements for the M1, M2, and Intel versions of the 13" MBP?
By comparing the power consumption of the M2 to the power consumption the case and cooling were designed to handle (that of the late 2016 MBP with the high-end Intel chip) (2.9 GHz 6267U), we can give some context to the former. I did a a search, and can find those figures for the iMac, Mini, and Mac Pro, but not for the laptops.
[E.g., for the iMac: https://support.apple.com/en-us/HT201918 ]
By comparing the power consumption of the M2 to the power consumption the case and cooling were designed to handle (that of the late 2016 MBP with the high-end Intel chip) (2.9 GHz 6267U), we can give some context to the former. I did a a search, and can find those figures for the iMac, Mini, and Mac Pro, but not for the laptops.
[E.g., for the iMac: https://support.apple.com/en-us/HT201918 ]
Last edited:
And yet, in the end, it finished the processing 10% faster than the M1. But it sounds so dramatic if you edit your video that way.
Regardless: The M-SoCs are not cool under full load. They are hot. The cooling systems that Apple installs are not always sufficiently dimensioned.
MacBook Air M2
Why would a MacBook Air be consistently used in a scenario that causes the CPU to run at full tilt? You’re buying the wrong computer if your workflow requires that.
Even Max Tech admits that this is not really how people are using MBAs. A quote from that video “If you do short bursty tasks, like most people do, you are not going to have any issue. Don’t be worried.”
and yet, those videos cause dozens of people to post here all worried that their new MBA will turn out to be slow and hot. In reality, look at what they had to do in that video to get it to reach those temperatures and to start slowing down? Is that how you actually use your computer? Are you a YouTube celebrity? If so, then the MBA is probably not for you. If not, then take a moment and take a deep breath of reality and stop fretting so much.
Why does it have to be cool under load? They require less power to deliver performance comparable or exceeding competing products, which is what allows Apple to ship them in a passively cooled system to begin with.
I really don’t understand the context here. Would M2 deliver higher performance if you put it in a larger, better cooled chsssis? Certainly. But this is also true for any other system. I mean, why limit a Rysen 6800U to 15-30W, if you can operate it at 100W and get the same performance as a desktop unit?
Hardware design is about balance and trade offs. MacBook Air is a passively cooled ultraportable. It’s 10W sustained power dissipation limit lets it compute very favorably agains 30W actively cooled designs from other manufacturers, which is an excellent result for Apple. What’s the problem? That they are leaving performance on the table? So what if they are? It makes sense for this particular machine and does not diminish it’s quality.
More so, I half expect to see this with M-series going forward. M1 was criticized as being mobile-focused part that wasn’t any faster in a desktop. What if Apple will start delivering more power-hungry chips that will need to be scaled back on mobile and can only deliver peak performance in the desktop chassis. Wouldn’t that be a legitimate design tactic? In the end of the day, the important thing is what users get from it. And so far M2 Apples to be 10-20% faster in the CPU and up to 30% faster in the GPU against, which is not bad at all.
Last edited:
People also don’t seem to understand the fact that chip temperature isn’t as simple as one number in your menu bar. When under load different parts of a single chip can vary in temperature by like 20 degrees. Temperature sensors are simply guidelines, just because one sensor says 100 degrees doesn’t mean the entire chip is 100 degrees, some parts could be 90 degrees, some transistors could be at 150 degrees, you simply don’t know because the transistors are smaller than a built-in temperature sensor.
so it is absolutely absurd to claim that because the m2 sensor says 108 degrees therefore it’s hotter than intel chips, or even m1 chips. We simply do not know where the sensor has been placed and how it relates to the reliability of the chip. It’s entirely possible, that maybe Apple simply moved the sensor closer to the actual cpu cores, which causes to read higher numbers than m1.
so it is absolutely absurd to claim that because the m2 sensor says 108 degrees therefore it’s hotter than intel chips, or even m1 chips. We simply do not know where the sensor has been placed and how it relates to the reliability of the chip. It’s entirely possible, that maybe Apple simply moved the sensor closer to the actual cpu cores, which causes to read higher numbers than m1.
So we are supposed to take the Intel readings (of 100C) at face value, but not the Apple Silicon readings (of 100+C)? If so that is a weird take.
In one of the other "OMG the MBA overheats when running benchmarks" threads a video was posted demonstrating that under a moderate video workload the 8GB M2Air is on average 10-20C cooler than the 16GB M1Air; under a heavy video workload both choked so while 16 or 24GB would help if that's your bread and butter buy a Pro. In fact, the M2Air never actually got hot enough to throttle while the M1Air did.
Last edited:
You have a weird take…
neither is “face value”, you can’t represent the thermal condition of a chip with one number period.
you can use the one number for one chip design, to determine its real time cooling requirements, you can’t compare it between any two different chip designs.
If only there was a way to sticky this and make every Macrumours member read this before posting.
A lot of "the sky is falling" doomsday proselytising when in reality, the people buying a MBA will never be in such a situation, and I think Apple designed the MBA with those limitations knowing this fully well.
Is a drawback a drawback if users will literally never encounter it?
You are right that CPU temp doesn't matter as long as it is below the TjMax set by the manufacturer. If anything all we are comparing is how well the cooling system can bring the temp away from that limit for a given workload.