By: --- (---.delete@this.redheron.com), September 9, 2022 8:23 am
Room: Moderated Discussions
Mark Roulo (nothanks.delete@this.xxx.com) on September 9, 2022 7:41 am wrote:
> anonymou5 (no.delete@this.spam.com) on September 8, 2022 8:08 pm wrote:
> > > So ~6% faster.
> > >
> > > We probably get a new micro-architecture with the A17?
> >
> > Who cares.
> >
> > It's a phone.
>
> Some of the posters or the Real World Tech forums find CPU performance interesting
> whether faster CPUs make our text editors run faster or not.
>
> You may not be one of these posters, but they do exist.
If you are interested in the technology, then saying "the A16 is the same speed as the A15", while technically true is very misleading.
It's the same speed (actually ~10% faster) at 20% less power...
So there are two takeaways from this:
The first is that Apple are optimizing the phones for energy rather than performance. This is probably a good idea given that most people care about the battery life of their phone more than the performance. This is a second year in a row that they have essentially optimized much more for energy than performance.
The second is, how do we interpret this in light of the mac?
Well, you can say that what has been the most impressive CPU-design machine in the world for the past 10 years or so has inexplicably fallen apart. (This claim, of course, has as a background assumption that optimizing for energy is something so trivial any monkey can do it; it's just some strange unimportant weirdness that no-one else has managed to do it as successfully as Apple, either ten years ago or in the past two years...)
OR you can think about the problem more carefully.
Apple currently design three cores
- Chinook (very small)
- E
- P
Why not add a fourth class to this, call it the U core or whatever? This allows devices on wall power to make more aggressive choices regarding performance, while retaining the energy/performance tradeoffs that are appropriate to phones, iPads, and lightweight laptops.
Obviously the CPU teams will share IP (as they do today). But going forward Apple has two directions of interest:
One is wearables [glasses], where energy will continue to dominate.
The other is extreme compute [desktop, but also AI and also cars], where performance will matter.
They can try to straddle both (like Intel did) and perhaps achieve what Intel did, being not especially satisfactory to either the high-end or the low-end. Or they can accept this reality and bifurcate the cores.
Like I have said before, I think it's significant that they did not update the mac mini with the M2 (even as they continue to sell an Intel mac mini, for gods sake). I expect the plan is to ship new cores in new desktop/high-end laptop SoCs "soon", but Apple aren't Intel, they don't telegraph the plans in advance. IF this plan has been delayed (eg N3 delay) we might see M2 Max's and Pro's; but my guess is (modulo branding, which is less interesting) the next round of Mac announcements (maybe October?) will be based on a new core.
> anonymou5 (no.delete@this.spam.com) on September 8, 2022 8:08 pm wrote:
> > > So ~6% faster.
> > >
> > > We probably get a new micro-architecture with the A17?
> >
> > Who cares.
> >
> > It's a phone.
>
> Some of the posters or the Real World Tech forums find CPU performance interesting
> whether faster CPUs make our text editors run faster or not.
>
> You may not be one of these posters, but they do exist.
If you are interested in the technology, then saying "the A16 is the same speed as the A15", while technically true is very misleading.
It's the same speed (actually ~10% faster) at 20% less power...
So there are two takeaways from this:
The first is that Apple are optimizing the phones for energy rather than performance. This is probably a good idea given that most people care about the battery life of their phone more than the performance. This is a second year in a row that they have essentially optimized much more for energy than performance.
The second is, how do we interpret this in light of the mac?
Well, you can say that what has been the most impressive CPU-design machine in the world for the past 10 years or so has inexplicably fallen apart. (This claim, of course, has as a background assumption that optimizing for energy is something so trivial any monkey can do it; it's just some strange unimportant weirdness that no-one else has managed to do it as successfully as Apple, either ten years ago or in the past two years...)
OR you can think about the problem more carefully.
Apple currently design three cores
- Chinook (very small)
- E
- P
Why not add a fourth class to this, call it the U core or whatever? This allows devices on wall power to make more aggressive choices regarding performance, while retaining the energy/performance tradeoffs that are appropriate to phones, iPads, and lightweight laptops.
Obviously the CPU teams will share IP (as they do today). But going forward Apple has two directions of interest:
One is wearables [glasses], where energy will continue to dominate.
The other is extreme compute [desktop, but also AI and also cars], where performance will matter.
They can try to straddle both (like Intel did) and perhaps achieve what Intel did, being not especially satisfactory to either the high-end or the low-end. Or they can accept this reality and bifurcate the cores.
Like I have said before, I think it's significant that they did not update the mac mini with the M2 (even as they continue to sell an Intel mac mini, for gods sake). I expect the plan is to ship new cores in new desktop/high-end laptop SoCs "soon", but Apple aren't Intel, they don't telegraph the plans in advance. IF this plan has been delayed (eg N3 delay) we might see M2 Max's and Pro's; but my guess is (modulo branding, which is less interesting) the next round of Mac announcements (maybe October?) will be based on a new core.