Apple is an existential threat to the PC

Because machine learning isn't unheard of in compilers nowadays, it's perfectly possible some parts of the transpiler are essentially fixed function ... highly fit functions it will just readily drop in when it recognizes a trace of code which so happens to occur in geekbench and unlikely to ever be used any where else. Not so much in an intentional effort to mislead either, I mean why do it online when you can throw millions more cycles at it offline? Storage and bandwidth are cheap. Today geekbench, tomorrow Office.
 
So, let's play Geekbench all night long?
James: “The same procedure as last year, Miss Sophie?”
Miss Sophie: “The same procedure as every year, James.”
Guessing we'll be 'playing it' for about 2 more days until people have their hands on them and can benchmark full apps?

Here's something that isn't geekbench then, GFXBench:

https://www.tomshardware.com/news/apple-silicon-m1-graphics-performance#xenforo-comments-3662289

Why are some people so impressed by Geekbench results? Every autumn/fall the same stuff: Yes, Apple has built a SoC for impressive Geekbench numbers. As an iPhone user for over 12 years: I am not impressed.
...ok? And do you find your iPhone slow when running apps?

The entire point with this thread now is that these results may actually have relevance to general purpose computing with the M1, this isn't complicated. This doesn't mean Apple's CPU's haven't always been great performers.
 
Lol, you serious, this is about phones. I think Apples are great, but so are some android phones.

Yes, why are you in such a hurry to downplay the performance characteristics of this 15W SoC?

I feel it’s a great starting point for Apple’s computers.

Do you think the good benchmark results will not carry over to a good user experience or actual use cases?
 
Cinebench 23 numbers are in, and they actually adress one of my main questionmarks - to what extent the fanless design of the MacBook Air causes throttling.
MacBook Pro CineBench 23 ST: 1498
MacBook Pro CineBench 23 MT: 7508
MacBook Air CineBench 23 ST: 1477
MacBook Air CineBench 23 MT: 6304

Only single numbers mind you, but still. Essentially no throttling single core, and 6304/7508 = 0.84 => a 16% drop with all cores running full blast for an extended time. That’s (in my book) really, really good, and I would never go for the fanned model out of performance concerns, although someone looking to replace his renderfarm with a MacBook might disagree. Only sustained graphics left to evaluate.
 
Yes, why are you in such a hurry to downplay the performance characteristics of this 15W SoC?

I feel it’s a great starting point for Apple’s computers.

Do you think the good benchmark results will not carry over to a good user experience or actual use cases?

I was soley talking about the mobile phones, as in 11 pro versus S20 where i think they are equally fast to most end-users. Ofcourse thats impressive for a 15w SoC.
It wont matter for me whats in my next console, PC or laptop. A potentional PS6 with Apple sillicon, or future PC's with it, aslong as it performs, i dont care. Competition in the hardware space is only good for all of us.

Anyway, a different question. I was looking into Geekbench IOS and Android results from all devices since 2013. Now, if geekbench is the best benchmark, then that means that an iphone 5s is basically on par with a SD810 xperia z5.
I still have these two phones, unused. When testing them side by side, the z5 is much and much faster in about anything. The 5s is running IOS12, the Z5 android 7.
Was the A7 really that far ahead?

https://browser.geekbench.com/ios-benchmarks
https://browser.geekbench.com/android-benchmarks
 
Cinebench 23 numbers are in, and they actually adress one of my main questionmarks - to what extent the fanless design of the MacBook Air causes throttling.
MacBook Pro CineBench 23 ST: 1498
MacBook Pro CineBench 23 MT: 7508
MacBook Air CineBench 23 ST: 1477
MacBook Air CineBench 23 MT: 6304

Only single numbers mind you, but still. Essentially no throttling single core, and 6304/7508 = 0.84 => a 16% drop with all cores running full blast for an extended time. That’s (in my book) really, really good, and I would never go for the fanned model out of performance concerns, although someone looking to replace his renderfarm with a MacBook might disagree. Only sustained graphics left to evaluate.
Possible source: And it's a 10 min run, so the result is very impressive.
 
Never understood why they designated the MBP 13" with the Pro moniker. It has always been like a dwarf with learning difficulties; It's not big (pro) and it's not clever (air). I get that it hits a sweet spot in the market, but they should just have run with MacBook, IMO, reserving Pro for the high end laptop/mobile workstation and Air for the super light categories.

I would expect Apple to build a higher end SOC with more M1 cores for upcoming iMacs and MBP 16s. Something like 12 M1 cores, binning for power and defective cores to get 12/10/8/6 cores iMac SKUs and high core count MBP 16". They are almost at performance parity with Zen3 and they are way ahead on performance/power. Not taking advantage of this to build a high core count SOC would be a wasted opportunity.

Cheers
 
Never understood why they designated the MBP 13" with the Pro moniker. It has always been like a dwarf with learning difficulties; It's not big (pro) and it's not clever (air). I get that it hits a sweet spot in the market, but they should just have run with MacBook, IMO, reserving Pro for the high end laptop/mobile workstation and Air for the super light categories.

I would expect Apple to build a higher end SOC with more M1 cores for upcoming iMacs and MBP 16s. Something like 12 M1 cores, binning for power and defective cores to get 12/10/8/6 cores iMac SKUs and high core count MBP 16". They are almost at performance parity with Zen3 and they are way ahead on performance/power. Not taking advantage of this to build a high core count SOC would be a wasted opportunity.

Cheers

It's expected that Apple will release a higher end SoC, and the rumour/leak is they have a 12 core on the way (with 8 performance + 4 efficiency cores). Presumably it would also have support for discrete GPUs or Apple may also have their own discrete GPU. But given that it's just the start of the 2 year transition period announced by Apple and the limits of their design teams in bringing new SoCs to market, it was never likely that all the designs would come at once. Either we see it in a quarter or so, or in a year where they keep the M1 and also introduce a higher end M2 or M2X. Devs also need some time to get the software ecosystem ready, especially for the pro apps, and this is why we see the low end, basic consumer variants released first.
 
Given Apple volumes and lack of need for very fine grained mix and match differentiation I'm not sure what a discrete GPU would get them. They can just make a SoC for every performance level they want.

They would find it hard to compete with PCs with multiple reticule buster CPUs plus multiple reticule buster GPUs ... but only the Mac Pro has to do that, I still think a NUMA approach makes the most sense there. You're already kinda breaking the unified memory architecture the moment you split the CPU and GPU, in fact I would say that would be much farther removed from "standard" Mac than a NUMA Mac Pro made from integrated SoCs.
 
Last edited:
I was soley talking about the mobile phones, as in 11 pro versus S20 where i think they are equally fast to most end-users. Ofcourse thats impressive for a 15w SoC.
It wont matter for me whats in my next console, PC or laptop. A potentional PS6 with Apple sillicon, or future PC's with it, aslong as it performs, i dont care. Competition in the hardware space is only good for all of us.

Anyway, a different question. I was looking into Geekbench IOS and Android results from all devices since 2013. Now, if geekbench is the best benchmark, then that means that an iphone 5s is basically on par with a SD810 xperia z5.
I still have these two phones, unused. When testing them side by side, the z5 is much and much faster in about anything. The 5s is running IOS12, the Z5 android 7.
Was the A7 really that far ahead?

https://browser.geekbench.com/ios-benchmarks
https://browser.geekbench.com/android-benchmarks

It was the first 64-bit Arm SoC and still does reasonably well. Apple have very strong web / browsing performance and they can leverage vertical integration with hardware and software. Something other vendors can't in the same extent.

When you say the z5 is much faster, what do you mean? It feels faster and more responsive or in certain benchmarks?
 
That's some impressive performance out of the smallest SoC in the Apple Silicon family.

What I find funny is just how irrelevant ISA truly is ... x86 or ARM, it matters almost bugger all. Even for power, most of the M1 advantage is that it's a SoC. Apple's main advantage is being able to tailor the design. PC's main disadvantage is how fragmented the market is.

Of course this parity will be destroyed as more and more of the margin and investment moves away from PCs ... and even if Apple doesn't get into the server market to challenge the last stronghold for x86, Amazon will with Graviton.

PCs are facing an existential threat.
 
When you say the z5 is much faster, what do you mean? It feels faster and more responsive or in certain benchmarks?

Generally faster in games and apps. Tested some stargazing apps and games like pubg. Im not intrested in playing on mobile, just like to compare.

I havent benchmarket any of those phones. The z5 (compact) does feel much faster in general use, yes.

The iphone is two years older, but runs two years younger OS (ios12 vs android 7.1.1). Thats just a shame i think.

I also have a z1 compact (SD800), the 5s trade blows with that one, with that i mean opening apps, memory management, in-app use etc.
The 5s is smoother, the z1 opens apps faster, but it seems the 5s runs then abit better.
The z1 can be modded to android 10, but thats not what most would bother with, and most likely wont help all that much anyway in speed.

Guess il have to compare more apps :) Btw, night sky runs rather well on the 5s, almost as good as the 11 pro max, aside from the AR camera function.

Overall obviously im very impressed of how well old iphones hold up (7 years). Means i could have the 11 pro up to 2026 and beyond i guess:)
 
Given Apple volumes and lack of need for very fine grained mix and match differentiation I'm not sure what a discrete GPU would get them. They can just make a SoC for every performance level they want.

They would find it hard to compete with PCs with multiple reticule buster CPUs plus multiple reticule buster GPUs ... but only the Mac Pro has to do that, I still think a NUMA approach makes the most sense there. You're already kinda breaking the unified memory architecture the moment you split the CPU and GPU, in fact I would say that would be much farther removed from "standard" Mac than a NUMA Mac Pro made from integrated SoCs.

Not that easy (or cheap) to make large SoCs on leading edge nodes though. And with the iMac and Macbook Pros they offer multiple dGPU options. Given that these also sell in limited volume, I don't see the business case for multiple dedicated SoCs.

I don't think they intend to compete with the highest end PC GPUs, and I suspect they will continue to offer the Mac Pros with Intel/AMD parts even beyond the 2 years.

They can get around the unified memory issue by using their own dGPUs over a cache coherent link, and this is an easier approach than making a larger SoC with a bigger integrated GPU. There have been rumours of an Apple dGPU as well.
 
Not that easy (or cheap) to make large SoCs on leading edge nodes though. And with the iMac and Macbook Pros they offer multiple dGPU options. Given that these also sell in limited volume, I don't see the business case for multiple dedicated SoCs.

I don't think they intend to compete with the highest end PC GPUs, and I suspect they will continue to offer the Mac Pros with Intel/AMD parts even beyond the 2 years.

They can get around the unified memory issue by using their own dGPUs over a cache coherent link, and this is an easier approach than making a larger SoC with a bigger integrated GPU. There have been rumours of an Apple dGPU as well.

I see no reason why they can’t scale up from 120mm2 to 300-350mm2, like the PS5 and Xbox One Box X (I think I got that one right).
 
I think either bad habit or for more software compatibility with mobile.

I doubt that if you've got a 13" high brightness screen pumping out photons it makes much sense to worry about the power efficiency of the big processing cores at the extreme low end. Intel Athena Ultrabooks can do 24+ hour battery life under light load too.
 
I think either bad habit or for more software compatibility with mobile.

I doubt that if you've got a 13" high brightness screen pumping out photons it makes much sense to worry about the power efficiency of the big processing cores at the extreme low end. Intel Athena Ultrabooks can do 24+ hour battery life under light load too.
No, reducing the baseline power consumption has absolute paramount for battery life, even on a laptop. And the efficiency cores are still plenty powerful at roughly 1/4th the speed of the big cores.
 
Back
Top