They said they had the fastest NPU in a PC. Not the fastest on earth (one of the nVidia cards, probably). Not the fastest way you could run something (probably a 4090 as you said). Just the fastest NPU shipping in a PC. Probably consumer PC.
It’s marketing, but it seems like a reasonable line to draw to me. It’s not like when companies draw a line like “fastest car under $70k with under 12 cylinders but available in green from the factory”.
The fact is that a GPU from Nvidia is a much faster NPU than a CPU from Apple.
It is marketing as you say, but it's misleading marketing, on purpose. They could have simply written "the fastest integrated NPU of any CPU" instead. This is something Apple often does on purpose, and people believe it.
It just happens to be it’s good at neural stuff too.
There’s another difference too. Apple’s NPU is integrated in their chip. Intel and AMD are going the same. A 4090 is not integrated into a CPU.
I’m somewhat guessing. Apple said NPU is the industry term, honestly I’d never heard it before today. I don’t know if the official definition draws a distinction that would exclude GPUs or not.
I simply think the way Apple presented things seemed reasonable. When they made that claim the fact that they might be comparing against a 4090 never entered my mind. If they had said it was the fastest way to run neural networks I would have questioned it, no doubt. But that wasn’t the wording they used.
Yes, and so does the M4.
> It just happens to be it’s good at neural stuff too.
No, it's no coincidence. Nvidia has been focusing on neural nets, same as Apple.
> There’s another difference too. Apple’s NPU is integrated in their chip.
The neural processing capabilities of Nvidia products(Tensor Cores) are also integrated in the chip.
> A 4090 is not integrated into a CPU.
Correct, but nobody ever stated that. Apple stated that M4 was faster than any AI PC today, not that it's the fastest NPU integrated into a CPU. And by the way, the M4 is also a GPU.
> I don’t know if the official definition draws a distinction that would exclude GPUs or not.
A NPU can be a part of a GPU, a CPU or it's own chip.
> If they had said it was the fastest way to run neural networks I would have questioned it,
They said fastest NPU, neural processing unit. It's the term Apple and a few others use for their AI accelerator. The whole point of a AI accelerator is performance and efficiency. If something does a better job at it then it's a better AI accelerator.
There is no actual distinction. A GPU with Tensor cores(=matmul units) really does have an NPU just as much as a CPU with an NPU (=matmul units).
Nevertheless, Apple's bragging is a little weird, because both Intel and AMD have already announced that in a few months they will launch laptop CPUs with much faster NPUs than Apple M4 (e.g. 77 TOPS for AMD), so Apple will hold the first place for only a very short time.
It’s the fastest available today. And when they release something faster (M4 Pro or Mac or Ultra or whatever) they’ll call that the fastest.
Seems fair to me.
They explicitly mentioned in the event that the industry refers to the neural engine as a NPU
> faster than the neural processing unit of any AI PC today.
By comparing the non-existent neural engine in your typical AI PC, you could claim that the very first SoC with an "NPU" is infinitely faster than the typical AI PC
https://www.theverge.com/2023/12/14/23998215/intel-core-ultr...
So going by the intended usage of this marketing rubbish, the comparison Apple is making isn’t to GPUs. It’s to Intel’s chips that like Apple’s, integrate, CPU, GPU, and NPU. They just don’t name drop Intel anymore when they don’t have to.
Of course, Apple has had an NPU in their SoC since the first iPhone with FaceID.
https://www.tomshardware.com/pc-components/cpus/intel-shares...
It might be a marketing term by Microsoft, but that is just dumb, and has nothing to do with what Apple says. If this was in relation to Microsofts "AI PC" then Apple should have written "Slower than ANY AI PC." instead, as the minimum requirements for "AI PC by Microsoft" seems to be 45 TOPS, and the M4 is too slow to qualify by the Microsoft definition.
Are you heavily invested in Apple stock or somehting? When a company clearly lies and tries to mislead people, call them out on it, don't defend them. Companies are not your friend. Wtf.
This isn't a nice thing to say.
I don't own any Apple stock, at least not directly. I'm not defending Apple, just trying to understand what their claim is. Apple does plenty of consumer un-friendly things but they aren't dumb and they have good lawyers so they tend not to directly lie about things in product claims.
In this case they do straight up lie, without a question. There is no reasonable explanation for the claim. If they had some absurd meaning behind it then they should have put a footnote on it.
They could have written it correctly, but that sounds way less impressive, so instead they make up shit to make it sound very impressive.
It’s the term Microsoft, Intel, AMD, and Qualcomm decided to rally around. No need to get upset at Apple for using the same term as reference for comparison.
Ps. Nvidia also doesn’t like the term because of precisely what you said. But it’s not Apple that decided to use this term.
The marketing hype is overblown, but for many (most? almost all?) people, the MacBook is a much more useful choice.
Under full throttle/load, even though the M2Pro is rated as less-performant, it is only using 105W — the tower/GPU are >450W!
I generally like Apple products, but I cannot stand the way they present them. They always hide how it compares against the directly previous product.
So while they're significantly slower than even casual gaming GPUs, they're technically the fastest NPUs on the market.
It's marketing speak.
https://www.intel.com/content/www/us/en/newsroom/news/what-i...
In that context it seems fair to make the comparison between a MacBook and the PC version which is closest on perf/watt rather than absolute performance on a space heater.
The PC version of accelerated AI workloads, in 2024, is a GPU with optimized matmul cores. It's the most powerful and most efficient way of accelerating neural network loads right now. Comparing to a suboptimal implementation and making it sound like you're comparing to the industry is misleading.
If they are referring to a specific marketing term for a single company, they should do so explicitly. Otherwise, it's just being misleading, because it's not even including Apple's main competition, which is AMD and NVidia, and using a generic sounding term.
You need to understand that the presentation, and product, wasn't made for the perf/watt crowd. That’s also why they’re so successful with consumer electronics: 9X% of the population is not part of the perf/watt crowd.
The average user who is comparing this against a PC for AI stuff would prefer a faster PC with a (much!) larger battery and the same autonomy at the end of day in that task but that's much faster every time. If I need to do intensive enough on-device AI workloads for efficiency to matter and on battery, why would I take an iPad with a 30Wh battery, 2x perf/watt, and 1x speed over a laptop with a 65Wh battery, 1x perf/watt, and 10x speed? The latter will give me a better experience, it will be more responsive and last longer.
So at the end of the day, Apple's comparison is extremely misleading at best, and a straightforward lie at worst.
If your primary use case is local AI on battery, you're probably into some small hundredths of a percent of the population here, who would be better served on some budget cloud compute (self deployed even) rather than heating up your lap.
> why would I take an iPad with a 30Wh battery...
Because it fits between the pages of the book in your backpack, has a nice pencil, and can *also* do ok with local AI.
There is no other use case where the power efficiency of the NPU matters, so you should direct this criticism at Apple.
> Because it fits between the pages of the book in your backpack, has a nice pencil, and can also do ok with local AI.
So if the only avantage is the form factor, why did Apple decide to compare it against PCs?
I'm saying this would not be an average user. The average user would not be comparing AI capabilities, they open an app that uses AI, with a high chance of not even being aware of it.
> So if the only avantage is the form factor, why did Apple decide to compare it against PCs?
What should they have compared it to? There's only one other tablet on the market, with an old chipset and a fraction of the performance, that just received its first real video editor late last year [1].
I agree though, they could/should have made clearer comparisons, along with Intel, AMD, Samsung, Nvidia, and everyone else. But, as with all general audience marketing presentations from all tech companies, vested technically literate users will look to the benchmarks, after release.
[1] https://r2.community.samsung.com/t5/Galaxy-Store-Apps-more/L...
What is the point here? If the average user doesn't care about comparing AI capabilities, then clearly Apple wasn't targeting that user in their AI capability comparison.
> What should they have compared it to? There's only one other tablet on the market, with an old chipset and a fraction of the performance, that just received its first real video editor late last year [1].
It's very clear that Apple is trying to make the case that the iPad Pro is a replacement for a laptop. In fact, that was their core marketing claim when it came out : "What's a computer?" [1], and arguably that's the point of their latest ads where they market it as a general purpose computer. Given the price point and the marketing, yes, they should be comparing it with a computer when making the case it's a good replacement for a computer. And when they do, they should not be misleading to the point of falsehood.
1: https://www.tomshardware.com/pc-components/cpus/intel-says-l...
Neural processing unit is basically a made up term at this point so of course they can have the fastest in the world.
Previous iPads had the best screens in a tablet even if they weren't oleds. Now that they finally use oleds, they have the best oled screens and the best screens in a tablet.
I always wonder what crazy meds Apple employees are on. Two RTX 4090s is quite common for hobbyist use, and that is 1321 TOPS each, making two over 69 times more than what Apple claims to be the fastest in the world. That performance is literally less than 1 % of a single H200.
Talk about misleading marketing...