No mention of MHz/GHz

[LonestarOne]

Agreed. Without the actual GHz rating then we're missing out on an ability to compare like for like.

How does an M1 chip with 7 GPU cores at clock speed X compare to an M1 with 8 GPU cores at clock speed Y? How does knowing the maximum clock speed tell you how fast the chip will run in a convection-cooled MacBook Air versus a fan-cooled MacBook Pro?

The only meaningful numbers are those that measure real-world performance. How fast can a machine apply a PhotoShop filter, render a Final Cut project, open a hundred browser windows, or draw frames for a video game.

 

[LonestarOne]

It would be interesting to know if there are any clock speed differences between the 3 new models using the M1 SoC.

It’s unlikely that the SOC runs at the same clock speed all the time, especially in a passively cooled MacBook Air. It’s also unlikely that all parts of the chip run at the same clock speed. You’re trying to reduce a very complex chip architecture to a single number. It’s like trying to compare the performance of two athletes by asking, “What are their heart rates?”

 

[LonestarOne]

According to CPU monkey the M1 runs at 1.8GHz base clock and boosts to 3.1GHz on all cores.

Assuming that is correct (a big assumption), it tells you nothing about real-world performance unless you know how often the chip boosts to 3.1 GHz and for how long. Which almost certainly varies depending on the user and what applications he runs.

 

[cmaier]

While some comparisons are meaningless, not all are. For example, I would think it would be very interesting to know if there are differences in clock speed between the M1's in the MBP, MBA, and Mini. These machines, after all, have different thermal envelopes. Plus efficiency is of course less of an issue for the Mini (not a non-issue, but less of an issue).

And it might also be interesting to know if there are differences in clock speeds between the M1 and the A14.

Since apple doesn’t have to market the chip speed, they have no need to reduce the performance to a single number. There is a max speed, a min speed, and then, depending on what you’re doing and thermal conditions, the actual speed. I suppose what people are looking for is the maximum speed, but that, too, may not mean much. The maximum speed on a MBP may be higher than that on the Mac mini, but maybe it only runs at max on the MBP 1% of the time vs. 10% on the mini. What matters is actual performance in real workloads.

 

[Fomalhaut]

Assuming that is correct (a big assumption), it tells you nothing about real-world performance unless you know how often the chip boosts to 3.1 GHz and for how long. Which almost certainly varies depending on the user and what applications he runs.

I think CPU Monkey is just quoting the base frequency of the efficiency cores (1.8GHz) and the performance cores (3.1GHz). We don't know if these are true or not yet. I'm not aware of any turbo features that could increase these, but I expect the clock-speeds would be reduced to adapt to current load.

 

[Fomalhaut]

It’s unlikely that the SOC runs at the same clock speed all the time, especially in a passively cooled MacBook Air. It’s also unlikely that all parts of the chip run at the same clock speed. You’re trying to reduce a very complex chip architecture to a single number. It’s like trying to compare the performance of two athletes by asking, “What are their heart rates?”

I understand that the base frequencies will be reduced to adapt to load (and therefore save power), in much the same way Intel & AMD does now. My question is whether the base frequencies (e.g. 1.8GHz efficiency cores / 3.1GHz performance cores) differ between the models. Do the actively cooled models get higher clock frequencies? Or is the cooling just to prevent/delay CPU down-clocking?

 

[Fomalhaut]

Since apple doesn’t have to market the chip speed, they have no need to reduce the performance to a single number. There is a max speed, a min speed, and then, depending on what you’re doing and thermal conditions, the actual speed. I suppose what people are looking for is the maximum speed, but that, too, may not mean much. The maximum speed on a MBP may be higher than that on the Mac mini, but maybe it only runs at max on the MBP 1% of the time vs. 10% on the mini. What matters is actual performance in real workloads.

Agreed, but it would be interesting to see how well clock speed is maintained under load with the cooled models compared to the MacBook Air. It would be a simpler baseline than application performance that potentially has a lot of variables (caching, application state, memory usage etc.).

e.g. if a cooled Mac Mini can maintain 90% of max frequency continuously and hold max frequency for 30 seconds, and a MacBook Air can only maintain 60% of max frequency continuously, or max frequency for 10 seconds, then this would give a yardstick for performance under specific loads. For a 15 minute video export, application build, or extended gameplay, this could be significant.

 

[cmaier]

Agreed, but it would be interesting to see how well clock speed is maintained under load with the cooled models compared to the MacBook Air. It would be a simpler baseline that application performance that potentially has a lot of variables (caching, application state, memory usage etc.).

I agree. But we won’t get that from a single “figure of merit” quoted by Apple‘s marketing. Lots of people will do lots of testing and report results.

 

[skitidetdu]

Those numbers doesn’t really say much anymore. I’m curious to see how m1 compares to A14

 

[Henk Poley]

That would be completely ridiculous. You just suggested that the M1 outperforms latest AMD Zen3 desktop CPUs by 30% in single core.

It would be silly to double down on a napkin estimation. But Rosetta has a +30% performance overhead on the A12X. So it would be a sensible sign-off criteria. It being roughly +30% faster than some off-the-shelf high end cores of the competition makes sense in that regard. There's not much performance difference between x64 desktop/laptop cores, just a TDP turbo limiter.

Kind of: "People, either the transpilation gets faster, or the silicon gets faster, but once we are at the performance level of our current laptops, we're going to announce the switchover."

Maybe I put the performance estimate slightly too high, I don't know 🤷‍♂️

 

[deeddawg]

A lot of tech media site readers writers simply don't get this.

Fixed that for you.

As you note: Since we entered the era of bursting, threading, thermally-variable clock speeds, and so forth, specifications like max clock rate largely ceased to be a truly meaningful comparison.

That hasn't, of course, prevented the marketing folks from taking full advantage of people's ignorance.

 

[dogslobber]

How does an M1 chip with 7 GPU cores at clock speed X compare to an M1 with 8 GPU cores at clock speed Y? How does knowing the maximum clock speed tell you how fast the chip will run in a convection-cooled MacBook Air versus a fan-cooled MacBook Pro?

The only meaningful numbers are those that measure real-world performance. How fast can a machine apply a PhotoShop filter, render a Final Cut project, open a hundred browser windows, or draw frames for a video game.

That is your opinion. I try to deal in factual information, of which there was none, just marketing fluff. Show us your numbers, Apple, and those who don't want to see them please look away.

 

[PeterJP]

I remember when the first ARM chip was launched by Acorn. It was 4MHz but did one instruction per cycle. Comparing it was futile: the Commodore 64 had a 1MHz 6502 but that one took sometimes over 10 clock cycles for a single instruction. When machines like the Amiga came out with a 16MHz processor afterwards, the discussion went completely bonkers. People could argue that either machine was 4x faster than the other with evidence to boot. But the proof of the pudding is in the eating. So let's wait for actual benchmarks, both in emulated and native versions, before we conclude anything, shall we?

 

[dogslobber]

I remember when the first ARM chip was launched by Acorn. It was 4MHz but did one instruction per cycle. Comparing it was futile: the Commodore 64 had a 1MHz 6502 but that one took sometimes over 10 clock cycles for a single instruction. When machines like the Amiga came out with a 16MHz processor afterwards, the discussion went completely bonkers. People could argue that either machine was 4x faster than the other with evidence to boot. But the proof of the pudding is in the eating. So let's wait for actual benchmarks, both in emulated and native versions, before we conclude anything, shall we?

When the Archimedes showed up we knew the ARM speed and could use that as a metric for performance comparison between ARM generations and within ARM generations. Tech people want to see the facts, just the facts. Unfortunately, Apple Marketing is trying to never provide that information so techies need to resort to benchmarking to extrapolate even same ISA metrics.

BTW, which 6510 instruction took 10 clock cycles to execute?

 

[PeterJP]

When the Archimedes showed up we knew the ARM speed and could use that as a metric for performance comparison between ARM generations and within ARM generations. Tech people want to see the facts, just the facts. Unfortunately, Apple Marketing is trying to never provide that information so techies need to resort to benchmarking to extrapolate even same ISA metrics.

BTW, which 6510 instruction took 10 clock cycles to execute?

Looking it up now, it seems like the longest was 8. I must have confused with the 386, which was horrible, particularly if you actually measure them in a real system like Michael Abrash.

I understand people can compare ARM generations correctly. But that's not the question here: why didn't Apple provide GHz to compare it to Intel macs. Despite the fact that ARM is more CISC-like now and Intel more RISC-like, it is still hard. Yes, Apple may have done it with PowerPC when it was to their advantage. Still not very useful. In that context, benchmarks are your best facts.

 

[KPOM]

Unless I missed it I find it quite odd they didn’t mention actual processor speed. Just X times faster.

I looked on the Apples tech specs for the Air no mention there either.

My guess is that Apple is binning chips. Those with 1 suspect GPU core get allocated to the base Air. The fastest chips get allocated to the Pro and Mac mini and the rest to the Air. Note that they are not offering different CPU speed options.

 

[chabig]

I would think it would be very interesting to know if there are differences in clock speed between the M1's in the MBP, MBA, and Mini

You’ll know next week. It’s just not part of Apple’s marketing message.

 

[deeddawg]

You’ll know next week. It’s just not part of Apple’s marketing message.

Indeed.

It could be they've binned the chips by max burst and allocated to the three product lines accordingly. Or they could all have the same burst and and settle into different clocks for sustained loads based on the cooling system. Or something else we don't yet know.

What I can assure you is that lots of folks will be interested, there'll be lots of benchmarking & analysis, the teardowns will be interesting, and there'll be tons of MacRumors Armchair Engineers thinking they'd have made different trade off choices.

 

[LonestarOne]

That is your opinion. I try to deal in factual information, of which there was none, just marketing fluff. Show us your numbers, Apple, and those who don't want to see them please look away.

They’ve shown numbers. They may not be the numbers you want, but they are numbers that are relevant in the real world. Gearheads want to know how fast the engine is spinning. People with work to do want to know how fast they can compile code or render 4K video. Dismissing that as “marketing fluff” is just arrogance.

 

[Serban55]

So the M1 can maintain under active cooling system from the Mac mini, up to 3.1 ghz 4 cores(i guess the 13" mbp will also since both has an active cooling with 1 fan)
The macbook air after sustain load will clock their perf 4 cores to 2.4 ghz
Some dev apps will let you have some access to let the 4 cores go up if you want but you will face the heavy load heat issues that intel macs had
So, by default these 2 13" Macbooks are made to run colder by around 10-12C than the intel counterparts...remember under heavy load an Macbook air or even the 16" MBP you could run the cpu at 93C under sustain load, and that translated to around 38-40C around upper keyboard keys "touchbar" or FN key on the normal macs
Apple wants now to be able to have a far better experience under heavy load, so they dont want to let the cpu reach even 80C and that would translate to an 30C top key temp and thats something that you cannot appreciate enough after years and years of windows or macs heated and melting your fingers on/or your lap

 

[dogslobber]

They’ve shown numbers. They may not be the numbers you want, but they are numbers that are relevant in the real world. Gearheads want to know how fast the engine is spinning. People with work to do want to know how fast they can compile code or render 4K video. Dismissing that as “marketing fluff” is just arrogance.

I'm not sure why you're upset. Some of us simply want to see the actual tech specs for the chips and we'll draw our own conclusions. Apple can create all the PowerPoint slides with their numbers they want but it's all marketing fluff until proven otherwise. Nobody ever kept their job in tech by listening to a company's marketing dept.

 

[dogslobber]

So the M1 can maintain under active cooling system from the Mac mini, up to 3.1 ghz 4 cores(i guess the 13" mbp will also since both has an active cooling with 1 fan)
[snip]

Where are you sourcing this info from? Thanks.

 

[LonestarOne]

I'm not sure why you're upset. Some of us simply want to see the actual tech specs for the chips and we'll draw our own conclusions. Apple can create all the PowerPoint slides with their numbers they want but it's all marketing fluff until proven otherwise. Nobody ever kept their job in tech by listening to a company's marketing dept.

Who said I’m upset? Do you know what “Bulverism” is?

Having worked in an engineering group that spent a great deal of time measuring the performance of supercomputer applications, I can only roll my eyes at the “benchmarks” some “techies” believe.

All you’ve done is throw out insults. Saying “marketing fluff” over and over again is not a logical argument.

 

[NotTooLate]

I think Apple needs to provide some sort of performance index for comparisons among their own Apple Silicon machines. Like we need to have some sense of how many "x" Macbook Pro will be at peak vs. Macbook Air. Such as "1.2x" or "1000" vs. "800", similar to how benchmarks are done. How about an "Applemark" scale?

How is a person supposed to know whether it's worth it to buy a Pro over an Air, leaving aside battery size?

(And please don't reply saying you can't compare M1 to Intel Hz. Got that message from the many replies already. That's not my point)

The problem is that you been gauged for so many years by CPU vendors to believe that there is a big difference between getting a 3.2Ghz and 3.4Ghz of the same CPU so they can charge you the extra bucks , I think that if you have the M1 in all of those machines , you dont really need to look for the exact frequency , it will be optimized to the machine , you will be able to look at benchmarks of your target SW and compare between the Pro vs Air , but for all intended purposes it wont help you much just knowing the Ghz numbers , more over as ppl mentioned ,even if its the same , the Air will throttle , its just physics , so you wont be getting the same sustained performance if thats what you are looking for (long 100% utilization workload).
I think the fact you have M1 in the Air and the Pro makes them similar in peak performance even if one is clocked 100-200Mhz higher at the boost peak , you just need to decided if you like the other Pro features and care about long sustained workloads.
I am honestly shocked we dont get that M1 at 3.0Ghz and M1 at 3.2Ghz for an extra 200$ , its the easiest money grab in the silicon world , Intel are doing it for years.

 

[NotTooLate]

I'm not sure why you're upset. Some of us simply want to see the actual tech specs for the chips and we'll draw our own conclusions. Apple can create all the PowerPoint slides with their numbers they want but it's all marketing fluff until proven otherwise. Nobody ever kept their job in tech by listening to a company's marketing dept.

I always know it is you posting when i read one of your comments , for some reason you are very anti apple with 99% of your comments , you want to see actual tech specs? like what ? do you even know what you are looking for .... "draw your own conclusion" ...... do you think when you read the intel/AMD slides on their release presentations its not FULL of marketing ?
You got M$ sizes , you got a die shot , you got core count and some IP that are on the SoC , Thunderbolt ,Wifi6 , USB4 , DRAM capacity , SSD capacity , Battery life numbers and more what "tech specs" are you looking for ?

"it's all marketing fluff until proven otherwise" - fair , if you still in deny of whats coming next week then its on you , your time is up with this drum , you beat it this entire time in this forum , time to look for some salad to go with that hat you are about to eat , this CPU is going to crush the competition in front of it , and maybe we can finally see a positive post from you ! it will be nice and show some class !