m1 ridiculously fast

[stumblingfalk]

The thermal design (heatsink, airflow, chassis) of the M1 Air at least, definitely has its limitations and it will do what every other computer does if it runs too hot too long, it will throttle. Throttling means it clocks itself down, requiring less power which means less heat is generated. That's why the M1 Air is a mixed bag performance wise. For a lot of tasks it runs nice, cool and impressively smooth due the extremely efficient M1 package. The thernal design is capable of soaking up the heat generated from short bursts just fine, such as opening a program, a large file or things like that.

But if you put it under sustained heavy load, like a resource hungry game or encoding a long video, it will will reach its thermal ceiling and eventually start to buckle down. You can visually see this as progress bars slow down, fps drops and temperatures reach into the 90c range for a little while, before it suddenly starts to run cooler (classic thermal throttling).

Of course almost every mid to high end x86 laptop (except some large fat gaming rigs/desktop replacements) will do the same as their cooling solutions simply are not adequate to soak up 50-90 watts worth of heat from your typical Intel i7 CPU for extended time. Thats why some enthusiast pc laptop gamers spend a lot of time tweaking Windows, undervolting the CPU and modifying their laptops chassis for better airflow and things like that. To allow the CPU to run at max capacity as much/as long as possible with minimal throttling.

This said, the M1 Air performs admirably well and is an excellent device for a lot of people.

 

[Fomalhaut]

My M1 MBA and M1 iMac both get to over 90C when pushed hard for 30+ minutes. This only happens once in a while, but they both absolutely can get hot, depending on your workload.

Do you mean the CPU temperatures or is the actual machine hot to the touch? I think most people are referring to the latter.

 

[uller6]

Do you mean the CPU temperatures or is the actual machine hot to the touch? I think most people are referring to the latter.

I mean the CPU temperatures. The MBA is cold to the touch in normal operation, but the M1 iMac has a rather warm spot right where the CPU is in the chin.

 

[mi7chy]

M1 in MBA can draw upwards of ~20W when cool but drops to ~10W under sustained load since that's about as much heat the passive cooling can dissipate and with ~25% drop in performance due to throttling. If you need consistent sustained performance get a device with active cooling (fan) since the MBA is more for light short workloads like browsing.

 

[dandeco]

Yeah, I was pretty surprised myself. When it was first announced that future Macs would drop the Intel processors in favor of Apple's own ARM-based silicon chips similar to in their iOS devices, I thought this could lead to a decrease in performance, what with those being mobile devices that would generally be underpowered compared to Mac desktops and laptops. So imagine my surprise when the M1 chip came out and that the Macs using it were a lot faster and more powerful than their Intel predecessors! I had planned to buy a 13" Intel TouchBar MacBook Pro early this year because I figured it'd take quite a while for third parties to adapt to Apple Silicon and such, and I wanted to be able to run a Windows XP virtual machine in VMWare Fusion or whatever for if I want to play old PC games (like for my "Let's Play" commentary videos or any future convention panels I may host), while waiting until next year to get an Apple Silicon Mac desktop to replace my 2012 quad-core i7 Mac Mini. But when I heard about how fast and powerful the M1 Mac laptops were and how they cost less than the Intel MacBook Pros still on them market, how more third party software vendors were increasing support for Apple Silicon, how Apple would be steadily dropping support for Intel Macs, and how thanks to UTM I could still emulate a Windows XP computer on the Mac (along with running some iOS apps I like to use like GoldWave), I decided to take the plunge into the Apple Silicon world, and I was just blown away! I even opted for the M1 MacBook Air, the 512 GB model with 8-core graphics and the memory configured to 16 GB (because I do a lot of digital media work), after hearing how it was almost as powerful as the M1 MacBook Pro, compared to their Intel predecessors where the Pro would be far more powerful than the Air. Of course, come next year I'm still going to get a new Mac desktop to replace my 2012 Mini, and I am planning to go with either the higher-end Mac Mini that should be out by then (using the "M1X" chip or whatever it's called) or the rumored "Mac Pro Cube."

 

[ek9max]

I love the M1 platform so much. I picked up a Mac mini 16gb/1tb and a base model MacBook Air 8gb/256gb for my wife.

Love them so much that I picked up another MacBook Air for myself a couple months later.

 

[AgentMcGeek]

M1 in MBA can draw upwards of ~20W when cool but drops to ~10W under sustained load since that's about as much heat the passive cooling can dissipate and with ~25% drop in performance due to throttling. If you need consistent sustained performance get a device with active cooling (fan) since the MBA is more for light short workloads like browsing.

True, but sustained benchmark performance between the MBA and MBP is not that much, usually around 5 to 15%. R23 5th consecutive run = 7100 vs 7700 points according to Dave2D. For comparison an 1165G7 stands at 5500 under similar conditions.

So I'm not so sure the MBP is worth the extra $$$ unless you do sustained workload everyday.

 

[doolar]

Another wow moment today. Anyone who's ever used Garmin Basecamp knows what a total dog that program is. It's better on PC, but only marginally so. On MacOS it's a laggy, choppy hot mess of a program.

Going on a three day motorcycle trip with my mate next week, so I had to venture into Basecamp. On my MBP 15" 2017 with a discrete GPU, it is just almost unusable, but since I have to, I've accepted it.

I'm not too surprised, but still a bit, running Basecamp on my M1 Air. Basecamp is of course not native and runs in Rosetta 2. But it is still a lot more responsive than on my MBP. It's just silly (but good silly) how powerful M1 is.

Anyways - just wanted to share my experience. Trip's planned and we're all set to go next week!

 

[dmccloud]

Can I get something straight here without being accused of some kind of dunning krugerism? This thread was featured on the front page and the pile-on is sort of interesting.

The initial claim was ‘the A14 is the base version of the M series chips’

The initial response (I think from cmaier) was ‘no you idiot, the cores in an A14 are completely different to the M1’. The reader is then told to ‘look at’ the A14 cpu cores and the M1 cores, in which case the difference would be plain as day. And in fact, the reader was informed:

‘Once again: the physical design of each firestorm/icestorm core in M1 is completely different than the physical design of the firestorm/icestorm cores in a14. Just look at them - you can easily see that the standard cell placements and the routes are completely different. They have completely different physical designs’

Then someone posted a link to an extreme tech post with photos of the M1 and A14 as ‘proof’ of that.

somehow the goalposts subsequently changed and the pile-on started pointing out that the layout and components of the A14 SOC are different from yeh M1. Which was never the initial claim.

anyway, I looked at the extreme tech photos and as far as I can tell, at least the firestorm cores are identical, but rotated and at a different scale (because the A14 is more zoomed in, being a smaller SOC).

Now, I’m confused. For all those playing at home, can anyone confirm whether or not the A14 cores are in any way different from the M1 cores, other than how many there are, and the ‘stuff’ that’s connected to them?

You're conflating the cores with the overall design of the A-series and M-series chips. While the Icestorm and Firestorm cores themselves are the same, it is the rest of the architecture within the M-series that differs from the A-series. Think of the cores of any ARM-based processor as pieces of a LEGO set. You could use the same pieces to build multiple items (SoCs), but each item will look and function in a different manner. How those common pieces connect to the rest of the object under construction can also vary widely from one project to another. Now in the case of the M1, there are even more differences between it and the A14 SoC when you look at the overall product. From the Extreme Tech article: "Second, the M1 integrates silicon that the A14 doesn’t, like the Apple T2 security processor, as well as support for standards like PCIe." Because the M1 SoC is laid out differently from the A-series and includes several items not contained within the A-Series SoCs, the way those cores are interconnected to the rest of the on-die components will also be different. When it comes to ARM-based processors outside of Apple, multiple models and even manufacturers can use the same processor cores (aka the Cortex M7 for example), but the overall performance of each model will not be identical because of the various customizations and additions made to the SoC. If you look at Qualcomm's product page, you will see multiple models of SoCs using the same CPU cores, yet they will have different base speeds and features added to them. The cores do not tell the entire story when it comes to these SoCs precisely because of the modularity of the architecture.

 

[tlab]

You're conflating the cores with the overall design of the A-series and M-series chips. While the Icestorm and Firestorm cores themselves are the same, it is the rest of the architecture within the M-series that differs from the A-series.

See, your post is what I thought the situation was until I read this thread. But I was responding to @cmaier who is an expert on CPU design. He said:

every transistor and wire in the a14 core is different than in the M1 core.

He then said:

A14 and M1 have completely different core physical designs. They are the same microarchitecture, but completely different from a circuit standpoint.

He then clarified:

Once again: the physical design of each firestorm/icestorm core in M1 is completely different than the physical design of the firestorm/icestorm cores in a14. Just look at them - you can easily see that the standard cell placements and the routes are completely different. They have completely different physical designs.

This is all very different from 'the ice storm cores and fire storm cores in an A14 are the same as those in an M1, but they're just components in a larger SOC design, which you can put together like lego'. This was his refutation of the idea that an M1 is more or less an A14X, because the A12X had the same cores as the A12 (which is similar what you are saying about the M1 vis-a-vis the A14, though obviously the M1 has more 'additional' SOC components that the A12X didn't).

So I'm trying to understand whether cmaier is in fact right about the firestorm and icestorm cores, or whether you are right. The photos I've seen of M1 and A14 seem to show icestorm / firestorm cores that are identical (just rotated and at different zoom levels). But cmaier definitely knows his stuff so I'm confused.

 

[Joelist]

At a very high level you might be able to call M1 an A14X but there are a lot of differences between M1 the SOC and the Apple A Series SOCs. M1 has more RAM, more GPU cores, more cache, different controllers for the laptop elements and thunderbolt, even some special blocks for running pieces of x86 code that are used to help Rosetta2 perform better.

 

[quarkysg]

This is all very different from 'the ice storm cores and fire storm cores in an A14 are the same as those in an M1, but they're just components in a larger SOC design, which you can put together like lego'. This was his refutation of the idea that an M1 is more or less an A14X, because the A12X had the same cores as the A12 (which is similar what you are saying about the M1 vis-a-vis the A14, though obviously the M1 has more 'additional' SOC components that the A12X didn't).

So I'm trying to understand whether cmaier is in fact right about the firestorm and icestorm cores, or whether you are right. The photos I've seen of M1 and A14 seem to show icestorm / firestorm cores that are identical (just rotated and at different zoom levels). But cmaier definitely knows his stuff so I'm confused.

My very high level understanding of this is that both the A14 and M1's CPU cores have the same logical design but very different physical design. The M1 have different components compared to the A14, e.g. larger GPUs, larger caches, e.g. and when all these components are layed out physically on a die, the designer has to consider various issues such as trace path, voltages, signal propagation delays, etc. between all components and has to compensate or tweak the placements of the transistors between the A14 and M1. I don't think one can take the A14 CPU core physical design and put it into the M1 die and expect it to work without any changes.

 

[cmaier]

My very high level understanding of this is that both the A14 and M1's CPU cores have the same logical design but very different physical design. The M1 have different components compared to the A14, e.g. larger GPUs, larger caches, e.g. and when all these components are layed out physically on a die, the designer has to consider various issues such as trace path, voltages, signal propagation delays, etc. between all components and has to compensate or tweak the placements of the transistors between the A14 and M1. I don't think one can take the A14 CPU core physical design and put it into the M1 die and expect it to work without any changes.

Correct. (Other than the last sentence. If they are using an SoC design methodology, then the design would probably work, though possibly with more latency.

 

[MrGunny94]

M1 in MBA can draw upwards of ~20W when cool but drops to ~10W under sustained load since that's about as much heat the passive cooling can dissipate and with ~25% drop in performance due to throttling. If you need consistent sustained performance get a device with active cooling (fan) since the MBA is more for light short workloads like browsing.

I multi task with my Air for the last month and 0 issues. I do a sort of Dev Ops work mixed with admin tasks at the OS level via SSH and use like 20 tabs on total due to web apps and what not.

And for memory :

Never went past the 50% CPU Mark.

 

[JanErik75]

This thread quickly derailed into a M1 vs A14 debate, just because of some guy's pointless misunderstanding/semantics over what a 'base chip' is. Everyone knows that the other guy meant base chip as in this first generation Apple Silicon M chip.

Back on topic, the M1 is a game changer for sure, and things will only get better from here. I personally can't wait to see what Apple has in store for the remaining product line, especially the 27" iMac and the Mac Pro.

My late 2013 27" is getting long in the tooth. It can't even run Big Sur.

 

[vs40]

Never went past the 50% CPU Mark.

Which app keeps those stats for 30 days?

 

[imp3rator]

Which app keeps those stats for 30 days?

iStat Menus

 

[MrGunny94]

Which app keeps those stats for 30 days?

iStat Menus it's pretty good!

 

[eoblaed]

do we have a mind blown smiley ?

?

 

[pshufd]

My late 2013 27" is getting long in the tooth. It can't even run Big Sur.

You might take a look at the thread on running Big Sur on unsupported hardware in the Big Sur forum.

Monterey isn't supported on my 2014 MacBook Pro but I got it up and running with the directions in the Monterey forum.