[ajross]

> First of all, Apple CPUs are not the fastest.

The cores are. Nothing is beating a M4/M5 on single CPU performance, and per-cycle nothing is even particularly close.

At the whole-chip level, there are bigger devices from the x86 vendors which will pull ahead on parallel benchmarks. And Apple's unfortunate allergy to effective cooling techniques (like, "faster fans move more air") means that they tend to throttle on chip-scale loads[1].

But if you just want to Run One Thing really fast, which even today still correlates better to "machine feels fast" than parallel loads, Apple is the undisputed king.

[1] One of the reasons Geekbench 6, which controversially includes cooling pauses, looks so much better for Apple than version 5 did.

[drob518]

For laptops at least, I appreciate not having fans that sound like a helicopter. I guess for Mac Mini and Mac Studio having more fan noise is acceptable (maybe a switch would be nice). One of the things that I love about my Air is there is zero fan noise all the time. Yes, it throttles, and 99% of the time I don’t notice and don’t care. Yes, I know there are workloads where it would be very noticeable and I would care, but I don’t personally run too many CPU bound tasks.

[zozbot234]

Bigger fans can move a lot more air while being less noisy, so if you care about a silent profile for any given amount of work the Mac Studio (or the Mac Mini if you don't need the full power of a Studio) is the best choice.

[TiredOfLife]

There is no non Apple desktop/server cpu with faster single core than apple m5 or even m4

[cosmic_cheese]

Same. It’s always disappointing when otherwise promising competing laptops turn out to be considerably more noisy if you’re doing anything more intense than using MS Paint.

It’s probably the single most common corner to cut in x86 laptops. Manufacturers love to shove hot chips into a chassis too thin for them and then toss in whatever cheap tiny-whiny-fan cooling solution they happen to have on hand. Result: laptop sounds like a jet engine when the CPU is being pushed.

[benjiro]

Even something like MS Paint can turn a laptop in to a aircraft.

The issue is actually very simple. In order to gain more performance, manufactures like AMD / Intel for a long time have been in a race for the highest frequency but if you have some knowhow in hardware, you know that higher frequency = more power draw the higher you clock.

So you open your MS Paint, and ... your CPU pushes to 5.2Ghz, and it gets fed 15W on a single core. This creates a heat spike in the sensors, and your fans on laptops, all too often are set to react very fast. And VROOOOEEEEM goes your fan as the CPU Temp sensor hits 80C on a single core, just for a second. But wait, your MS Paint is open, and down goes the fan. And repeat, repeat, repeat ...

Notice how Apple focused on running their CPUs no higher then 4.2Ghz or something... So even if their CPU boosts to 100%, that thermal peak will be maybe 7W.

Now combine that with Apple using a much more tolerant fan / temp sensor setup. They say: 100C is perfectly acceptable. So when your CPU boosts, its not dumping 15W, but only 7W. And because the fan reaction threshold is so high, the fans do not react on any Apple product. Unless you run a single or MT process for a LONG time.

And even then, the fans will only ramp up slowly if your 100C has been going on for a few seconds, and while yes, your CPU will be thermal throttling while the fans spin up. But you do not feel this effect.

That is the real magic of Apple. Yes, their CPUs are masterpieces at how they get so much performance from a lower frequency, but the real kicker is their thermal / fan profile design.

The wife has a old Apple clone laptop from 2018. Thing is for 99.9% of the time silent. No fans, nothing. Because Xiaomi used the same tricks on that laptop, allowing it to boost to the max, without triggering the fan ramping. And when it triggers with a long running process, they use a very low fan rpm until it goes way too high. I had laptops with the same CPU from other brands in the same time periode, and they all had annoying fan profiles. That showed me that a lot of Apple magic is good design around the hardware/software/fan.

But ironically, that magic has been forgotten in later models by Xiaomi ... Tsk!

Manufactures think: Its better if millions of people suffer from more noise, then if we need to have a few thousand laptops that die / get damaged, from too much heat. So ramp up the fans!!!

[cosmic_cheese]

And as a cherry on top, Apple uses custom fans designed to emit noise in less annoying frequencies and when multiple fans are in play, slightly varies their speeds to avoid harmonizing. So even when they do run, they're not perceived as being as loud at most speeds.

[ajross]

This seems like intuition fed from 8+ years ago and a HN culture of mac fetishism. It's just not like that. You can find some loud gaming laptops. And sure, occasionally there's an individual device with outlier fan behavior. But really the PC laptop world is pretty calm and boring these days. No, nothing goes "VROOOOEEEM". Chromebooks are even better.

[zozbot234]

You can mostly fix this by running your CPU in "battery saving" mode. CPUs should basically never boost to the 5GHz+ range unless they're doing something that's absolutely latency-critical. It's a huge waste of energy for a negligible increase in performance.

[saagarjha]

> Notice how Apple focused on running their CPUs no higher then 4.2Ghz or something... So even if their CPU boosts to 100%, that thermal peak will be maybe 7W.

If you max out your processor it will be more than happy to draw 20W+

[drob518]

Exactly. They shoot for the highest benchmark score and build something annoying to use on a daily basis.

[gpderetta]

It doesn't really make much sense to compare per-cycle performance across microarchitectures as there are multiple valid trade-offs.

Of course Apple did pick a very good sweet spot favoring a wide core as opposed to a speed daemon more than the competition.

[ajross]

> It doesn't really make much sense to compare per-cycle performance across microarchitectures as there are multiple valid trade-offs.

That's true in principle, but IMHO a little too evasive. In point of fact Apple 100% won this round. Their wider architecture is actually faster than the competition in an absolute sense even at the deployed clock rates. There's really no significant market where you'd want to use anything different for CPU compute anywhere. Datacenters would absolutely buy M5 racks if they were offered. M5 efficiency cores are better than Intel's or Zen 5c every time they're measured too.

Just about the only spaces where Apple is behind[1] are die size and packaging: their cores take a little more area per benchmark point, and they're still shipping big single dies. And they finance both of those shortcomings with much higher per-part margins.

Intel and AMD have moved hard into tiled architectures and it seems to be working out for them. I'd expect Apple to do the same soon.

[1] Well, except the big elephant in the room that "CPU Performance Doesn't Matter Much Anymore". Consumer CPUs are fast enough and have been for years now, and the stuff that feels slow is on the GPU or the cloud these days. Apple's in critical danger of being commoditized out of its market space, but then that's true of every premium vendor throughout history.

[gpderetta]

Oh. Apple won this and the last few rounds for sure. They definitely picked the right microarchitecture and delivered masterfully.

Early on personally I had doubts they could scale their CPU to high end desktop performance, but obviously it hasn't been an issue.

My nitpick was purely about using clock per cycle as a performance metric, which is as much nonsense as comparing GHz: AFAIK Apple cpus still top at 4.5 GHz, while the AMD/Intel reach 6Ghz, so obviously the architectures are optimized for different target frequencies (which makes sense: the power costs of a high GHz design are astronomical).

And as an microarchitecture nerd I'm definitely interested in how they can implement such a wide architecture, but wide-ness per-se is not a target.

[socalgal2]

What fast image / video gen software is available for Mac? I'm jealous of my friends with 4090s and 5090s

[llm_nerd]

I don't get your first line. When people talk about Apple's core speeds they're not talking about cycles per instruction or something, they're talking about single-thread performance on a benchmark like Geekbench. Geekbench runs various real-world code and it's the gross throughput that is measured, and it's there that Apple cores shine.