Next iPhone faster than M1 Ultra Mac Studio

[MandiMac]

It just goes to show that single core performance is pretty irrelevant in a normal computing environment. I have no idea why some people think it's so important.

It depends on what you are meaning with "normal computing environment". Normal usage for light/medium users might include spreadsheets, web surfing and photo manipulation - all usages where single-threaded performance matters, let alone the responsivity of a system. Multi-threaded performance is when you're continously utilizing all cores for one job, like exporting video or rendering complex scenes in Cinema 4D or the likes. So there's a difference between the two.

TL;DR: Multicore performance is only one side of the medal. Even 100 weak cores can achieve high multicore performance. So singlecore performance is very important for everday tasks.

TL;DR2: Even when 8 average relocation workers are doing their job well, that doesn't diminish the worth of 1 really strong relocation worker.

 

[leman]

In order to do this on, let's say the Mac Studio SoC, Apple would have to do extra work testing, validating, and binning chips for higher frequencies. In addition, they'd probably have to make some changes to MacOS specifically to handle boosting in various situations. For example, no boosting if Low Power mode is on.

Could be. To me personally, it seems a bit weird that Apple would leave this din of low hanging performance on the table. They always went for highest ST performance they could get, so not leveraging everything they could from their hardware isn't like them, but hey, I am only guessing and speculating. What I write here is just a "weak opinion strongly held", and I surely have no intention to die on that hill

While Firestorm is a wider design (already mentioned by me), it can certainly boost much higher on a single-core. Even if the power is exponential, we're still talking very little power here. When running Geekbench5 ST, my M1 Pro mostly stays around 0.3w - 1w of power. Meanwhile, AMD and Intel boost to probably 10x or higher on their laptop chips.

From what I understand it's not that much about width (current x86 designs are wider than most ARM chips), but pipeline complexity. And Firestorm's peak ST power consumption is well known — it's around 5watt. I would guess that you get this extremely low reading because GB5 is extremely short, so the frequency sampler doesn't catch the full boost duration.

I've been overclocking chips for 15 years. Any chip can be overclocked and you'd likely see extra performance. IE. Back in the days, my AMD Athlon XP would run at 1.8Ghz but you could overclock it to 2.5ghz. I doubt Firestorm is an exception.

Your experience with Athlon does not have to translate to Firestorm though. But who knows, it's not like we can decide this in a test

 

[leman]

It just goes to show that single core performance is pretty irrelevant in a normal computing environment. I have no idea why some people think it's so important.

Quite the opposite. In "normal" computing environment, ST makes the difference between "this PC feels snappy" and "this PC feels sluggish", which is what most users care about. ST doesn't matter for specialised computing environment where you can distribute the world across many processing units.

Apple's strategy has always been about providing both ST and MT.

 

[Arctic Moose]

But isn't the M1 Ultra, like, 32 cores? It's way faster than the iPhone 13! (Maybe the same on 1 core, though.)

Although not obvious from the click-bait title, the post does say "single-core", i.e. when comparing performance of a single core. For single-core performance comparisons it really doesn't matter how many cores you have, and all else being equal, of course a processeror with more cores will outperform one with fewer in multi-threaded usage.

Personally, I am extremely interested in single-thread improvements because the metric I care about the most about is reduction of interface lag, and as I understand it, single-core performance is a significant factor. (I turn off all the motion and transparency I can, but I still find most computer interfaces incredibly slow.)

 

[Freeangel1]

Why you need all this power in a cell phone is beyond me..

It did absolutely nothing putting an M1 chip in an iPad. Let alone a cell phone.

 

[Apple Knowledge Navigator]

I think what’s interesting about all of this is thinking about how Apple will scale their SoC across devices in the future.

It’s clear that having A-series and M-series chips co-exist will be only be a temporary strategy, perhaps lasting for no more than another another 5 years. The reason is simple; Apple will want both their hardware and software to scale unilaterally.

A developer will create an app that easily scales between the simplest UI and demands - watchOS - to the most sophisticated - macOS - and between iPhone, iPad and entry level Mac, they will all use the equivalent of an M1. Higher-end Macs will continue to get the ‘pro’ variants.

It’s also the scale of economy that makes a lot of sense, as Apple would be investing their efforts into a simpler strategy.

But it will probably be at least a 3nm process before we get there.

 

[scottrichardson]

Aren’t the M1 cores the same as the iPhone 12? Rather than the 13? The 13 has the A15 cores which are faster than the A14/M1. So the iPhone is already faster in Single core performance, no? Correct me if I’m wrong.

 

[Arctic Moose]

Aren’t the M1 cores the same as the iPhone 12? Rather than the 13? The 13 has the A15 cores which are faster than the A14/M1. So the iPhone is already faster in Single core performance, no? Correct me if I’m wrong.

Correct, the M1 Firestorm/Icestorm cores are shared with the A14. The A15 has updated Avalanche/Blizzard cores which operate at slightly higher clock speeds. (And another 3 billion transistors, 40% faster neural engine, updated image processor and bigger system cache.)

 

[leman]

I think what’s interesting about all of this is thinking about how Apple will scale their SoC across devices in the future.

It’s clear that having A-series and M-series chips co-exist will be only be a temporary strategy, perhaps lasting for no more than another another 5 years. The reason is simple; Apple will want both their hardware and software to scale unilaterally.

A developer will create an app that easily scales between the simplest UI and demands - watchOS - to the most sophisticated - macOS - and between iPhone, iPad and entry level Mac, they will all use the equivalent of an M1. Higher-end Macs will continue to get the ‘pro’ variants.

It’s also the scale of economy that makes a lot of sense, as Apple would be investing their efforts into a simpler strategy.

But it will probably be at least a 3nm process before we get there.

Not quite sure what you mean? Apple Silicon designs are already very modular and M1 is just a "bigger" A14, possibly with some additional desktop-related coprocessors and blocks (PCIe?) thrown into the mix. But the basic CPU/GPU architecture is the same. No wonder M1 was internally known as A14X. You might as well call A14 "m1 mini" or something

 

[bobcomer]

Normal usage for light/medium users might include spreadsheets, web surfing and photo manipulation - all usages where single-threaded performance matters, let alone the responsivity of a system.

I totally disagree with that statement -- there is no modern computer/OS where a single core will make any difference at all, no matter how fast, unless you have other cores to take up all the other tasks that need to happen to do all that -- servicing drivers/interrupts/devices. That's just not how a modern OS works. Where you get responsivity on a modern computer is having more cores and multithreading.

Multi-threaded performance is when you're continously utilizing all cores for one job, like exporting video or rendering complex scenes in Cinema 4D or the likes. So there's a difference between the two.

You're forgetting all the other tasks that need to happen outside that job.

 

[bobcomer]

Quite the opposite. In "normal" computing environment, ST makes the difference between "this PC feels snappy" and "this PC feels sluggish", which is what most users care about. ST doesn't matter for specialised computing environment where you can distribute the world across many processing units.

As I said in my other message I totally disagree, that just doesn't make sense with a modern OS and hardware. A computer has many other things to do than a single task and thread, just to service anything that single task needs and "snappyness" comes with being able to utilize more cores and not divert your main task.

Just go try using a single core machine again for one task and feel just how snappy it is in comparison.

 

[NotTooLate]

You would be amazed to see the clock speed and single core performance of an Epic 7763 for example (AMD server chip) , its slow as a snail vs their consumer lineup , also for years Apple A SoC`s were faster in single thread then most of CPU`s sold on the market , when ppl said it out loud they were getting the "its a phone chip , its a toy" or "benchmarks are useless" or "let me see them running on a real OS first".

 

[T'hain Esh Kelch]

Why you need all this power in a cell phone is beyond me..

It did absolutely nothing putting an M1 chip in an iPad. Let alone a cell phone.

It increases battery life since processes finishes faster, which is important in an iPhone, but not as much in an iPad.

 

[leman]

As I said in my other message I totally disagree, that just doesn't make sense with a modern OS and hardware. A computer has many other things to do than a single task and thread, just to service anything that single task needs and "snappyness" comes with being able to utilize more cores and not divert your main task.

Just go try using a single core machine again for one task and feel just how snappy it is in comparison.

You cannot delegate things like interrupt processing to other cores. What you are talking about is splitting work and prioritising work packages across multiple cores which is of course an important part of how modern OS works. But many key tasks like performing layout and drawing the UI are still fundamentally single-threaded, and there is no real way around it. Exactly the responsibilities of the "main task". No matter how many cores you have and how great your OS is at multitasking, you won't have a "snappy experience" if your main thread can't do the layout fast enough.

More accurately, smooth performance is determined by the ability to do certain types of user-oriented burst tasks within a certain amount of time. Good ST performance is critical here, as is the ability to quickly ramp up (e.g. some older Intel CPUs can be as fast as newer Skylake, but will feel sluggish because it takes them a while to activate turbo boost).

 

[Unregistered 4U]

If the M1 can be clocked higher but Apple chooses not to, even in situations where it's not constrained by heat and power, it's a sign that Apple is no longer interested in producing competitive hardware, and the future of Macs looks uncertain. If the M1 just doesn't run reliably much above 3.26 GHz, we can still assume that Apple is trying to make the best hardware it can.

If one looks at the current and every future Apple Silicon releases, one can predict with 100% certainty that Apple Silicon systems are only assured to be faster than the previous Apple Silicon system in that form factor. They are NOT assured to be faster than modded Intel or AMD systems at any time. The best hardware Apple can make is what they’ll release even if it means what they release has lower raw performance than those other chip vendors.

Mac apps are no longer that important, as it's not the 2000s anymore. Among the people I work, Keynote is pretty much the only Mac app they care about. Almost everything else is cross-platform, can be easily replaced, or runs in the cloud or in a browser. They don't want to buy new high-end MBPs or Mac Studios because those are Macs but because they want cost-effective performance in a convenient form.

Every browser that runs on macOS is a macOS app, so, yes, macOS apps are that important. And anyway, unless they’re doing some fairly heavy client side execution, anyone working through a browser isn’t all that interested in raw performance, so they could even use an iPad if the app was coded properly.

 

[yabeweb]

The iPhone 13 already has the same single-core performance as the M1 Ultra.

Hence the next iPhone 14 is going to be faster in single-threaded operation than the most expensive M1 Ultra.

That is going to be funny.

It's all funny, until you realize you can't run Pro apps on an iPhone

When Final Cut and similar runs on an iPhone, then it'll be funny, right now you are only comparing benchmark numbers, with do very little for people who actually work.

 

[Unregistered 4U]

However I’d like to add that raw performance is very far from the only metric of importance for computers used professionally.

Oh, that’s a certainty. I don’t think there’s very many people at all that are really that concerned about raw performance. BUT, for those that are concerned, and “raw performance”… where, with every release of EVERY chip, they’re looking at how it can be overclocked, what kind of cooling would be required, what would be the cash outlay… is the MOST IMPORTANT THING they’re concerned with (because those tiny percentage gains make a difference to them), they’re not using macOS with Apple Silicon.

People using the highest performing (but still less raw performance than the other vendors) macOS systems for awhile already know that performance is better with another rig, but the macOS experience is what they prefer. There will always be those primarily engaged in waste water ejection competitions that will bemoan or cheer that Apple’s chips are not the best at raw performance, but neither Apple nor their customers are concerned. They’ll be making and getting work done using the best tools they have available that run the OS of their choice.

 

[Unregistered 4U]

This is entirely untrue. One of the main reasons I use a Mac is because it’s the fastest portable computer (by a wide margin) for my type of work out there.

Then, doesn’t that mean the work you do requires macOS?

 

[leman]

Oh, that’s a certainty. I don’t think there’s very many people at all that are really that concerned about raw performance. BUT, for those that are concerned, and “raw performance”… where, with every release of EVERY chip, they’re looking at how it can be overclocked, what kind of cooling would be required, what would be the cash outlay… is the MOST IMPORTANT THING they’re concerned with (because those tiny percentage gains make a difference to them), they’re not using macOS with Apple Silicon.

People using the highest performing (but still less raw performance than the other vendors) macOS systems for awhile already know that performance is better with another rig, but the macOS experience is what they prefer. There will always be those primarily engaged in waste water ejection competitions that will bemoan or cheer that Apple’s chips are not the best at raw performance, but neither Apple nor their customers are concerned. They’ll be making and getting work done using the best tools they have available that run the OS of their choice.

The thing is, you are technically correct, but your criteria are so narrow that it’s almost not useful anymore. Nobody is looking for “pure raw performance” (and who is runs custom hardware or specialized setups). But one might look for best performance in a certain class of devices, balanced by some other factors.

For example: my baseline criteria are a portable machine with excellent screen and usability as well as an above average battery life. I need to be able to work untethered and on different locations, so a desktop is out of a question. Within this class of devices, an Apple Silicon laptop is so far above competition that it’s not even funny.

Then, doesn’t that mean the work you do requires macOS?

No, the tools I work with are available on all major platforms. But M1 is considerably faster on my average workflows and the unmatched ergonomy of a MacBook and macOS further increase productivity.

 

[Unregistered 4U]

The thing is, you are technically correct, but your criteria are so narrow that it’s almost not useful anymore. Nobody is looking for “pure raw performance” (and who is runs custom hardware or specialized setups). But one might look for best performance in a certain class of devices, balanced by some other factors.

For example: my baseline criteria are a portable machine with excellent screen and usability as well as an above average battery life. I need to be able to work untethered and on different locations, so a desktop is out of a question. Within this class of devices, an Apple Silicon laptop is so far above competition that it’s not even funny.

It’s narrow to match the “Apple has GOT to compete” refrain. Apple really never has to ever release anything that has better raw performance than AMD/Intel ever again. Apple Silicon ONLY has to be more performant than the prior Apple Silicon. In the real world, like you said, very few are looking ONLY at the fact that an Intel chip can be juiced and cooled to provide 4% better performance and making a system choice. That bleeding edge competition is still ongoing, but it’s between AMD and Intel.

No, the tools I work with are available on all major platforms. But M1 is considerably faster on my average workflows and the unmatched ergonomy of a MacBook and macOS further increase productivity.

I DID miss that previously. So “best performance” is not the most important metric, it’s just one metric AFTER the most important metric which would be “portable machine”. My assumption is that there’s a desktop Intel/AMD system out there that WOULD outperform your portable machine for those folks where “best performance” is more important than “portable machine”.

 

[ADGrant]

I totally disagree with that statement -- there is no modern computer/OS where a single core will make any difference at all, no matter how fast, unless you have other cores to take up all the other tasks that need to happen to do all that -- servicing drivers/interrupts/devices. That's just not how a modern OS works. Where you get responsivity on a modern computer is having more cores and multithreading.

A desktop or mobile application processes all user events and UI updates on a single thread. The faster that single UI thread can run the faster the computer will seem to the end user.

Web apps are not excluded from this limitation. Web browsers also have to process UI events in a single thread and Javascript single page apps also run in a single thread.

Even workloads that can be distributed among multiple cores can have parts of the workflow that bottleneck on a single thread which is why these workloads don't always scale perfectly.

 

[Ziperix]

The trolls in this thread claiming apple is behind in performance compared to other manufacturers is amusing

 

[leman]

It’s narrow to match the “Apple has GOT to compete” refrain. Apple really never has to ever release anything that has better raw performance than AMD/Intel ever again. Apple Silicon ONLY has to be more performant than the prior Apple Silicon. In the real world, like you said, very few are looking ONLY at the fact that an Intel chip can be juiced and cooled to provide 4% better performance and making a system choice. That bleeding edge competition is still ongoing, but it’s between AMD and Intel.

If Apple Silicon would be slower than similar x86 machines, I would at least contemplated moving to Linux. I don't think that you conclusion that Apple never again needs to compete agains x86 chips is entirely justified. People do choose machines based on multiple factors, which also includes performance. For example, there appears to be a not an insignificant amount of developers who purchased an M1 based laptop recently because of it's incredible performance in many development workflows.

I DID miss that previously. So “best performance” is not the most important metric, it’s just one metric AFTER the most important metric which would be “portable machine”. My assumption is that there’s a desktop Intel/AMD system out there that WOULD outperform your portable machine for those folks where “best performance” is more important than “portable machine”.

To be honest, I am not quite sure that any desktop machine out there would outperform Firestorm on the data transformation stuff I do. Intel does have higher performance, but my workloads are limited by memory bandwidth and cache, where M1 reigns supreme.

 

[bobcomer]

You cannot delegate things like interrupt processing to other cores. What you are talking about is splitting work and prioritising work packages across multiple cores which is of course an important part of how modern OS works. But many key tasks like performing layout and drawing the UI are still fundamentally single-threaded, and there is no real way around it. Exactly the responsibilities of the "main task". No matter how many cores you have and how great your OS is at multitasking, you won't have a "snappy experience" if your main thread can't do the layout fast enough.

More accurately, smooth performance is determined by the ability to do certain types of user-oriented burst tasks within a certain amount of time. Good ST performance is critical here, as is the ability to quickly ramp up (e.g. some older Intel CPUs can be as fast as newer Skylake, but will feel sluggish because it takes them a while to activate turbo boost).

Yeah, I shouldn't have mentioned interrupts, but I still disagree, it just doesn't make sense to me given what I know about how OS's work on modern hardware.

I do wonder what you mean by main task though as I never have just one main task and what I do takes more cores to feel "snappy".

 

[bobcomer]

A desktop or mobile application processes all user events and UI updates on a single thread. The faster that single UI thread can run the faster the computer will seem to the end user.

There's more going on than that.