Intel Alder Lake vs. Apple M1

[senttoschool]

Intel's current strategy is not the "classical" big.little though. Its parallel thoughtput. They are not focusing on combination of high performance cores for burst and high efficiency cores for efficiency, they are focusing on MANY mid-efficiency cores for higher performance in multi-core workloads. Alder Lake already comes with 16 cores (24 threads!), and so requires software that can take really good advantage of multiple cores to use the chip well. Raptor Lake is supposed to up the number throughput cores to 16, that will essentially result in a 32-thread CPU. Great for benchmarks, a bit awkward for real-world performance.

This is very different from, say, Apple's big.little (where high-performance cores use as much power as Intel's efficiency cores, and the efficiency cores are really there for efficiency). And it is also very different from what AMD is rumoured to do (their performance cores are also more efficient than Intel's). And of course, there is the modern Android approach... but the main purpose of their single "very fast core" is essentially benchmarking shenanigans.

big.Little simply means you have big, hungry, power cores and smaller, slower efficiency cores that exist in the same CPU. Intel's implementation isn't any different than Apple's. It's just that Intel is pushing both the P cores and E cores harder than Apple.

If Intel wants, they can simply dial it down and increase efficiency.

Alder Lake is incredibly efficient (for x86) when you don't push it past its limits.

We know nothing about AMD's big.Little plan. That's not happening until Zen5 which might be 2.5-3 years away.

People like to crap on Intel. But they're clearly ahead of AMD right now in non-server market and based on roadmaps, they're likely to stay ahead. Of course Apple is a different story.

 

[leman]

big.Little simply means you have big, hungry, power cores and smaller, slower efficiency cores that exist in the same CPU. Intel's implementation isn't any different than Apple's. It's just that Intel is pushing both the P cores and E cores harder than Apple.

What matters is the purpose. And Intel's purpose behind heterogenous CPU design is very different than Apples. Intel simply does not know how to make fast cores that are energy efficient. This is why they are compensating for their lack of computational throughout with many smaller slower cores that are more efficient. That's it.

We know nothing about AMD's big.Little plan. That's not happening until Zen5 which might be 2.5-3 years away.

AMD Readying 16-core "Zen 4" CCDs Exclusively for the Client Segment with an Answer to Intel E-cores?

AMD already declared the CPU core counts of its EPYC "Genoa" and "Bergamo" processors to top out at 96 and 128, respectively, a core-count believed to have been facilitated by the larger fiberglass substrate of the next-gen SP5 CPU socket, letting AMD add more 8-core "Zen 4" chiplets, dubbed CPU...

www.techpowerup.com

People like to crap on Intel. But they're clearly ahead of AMD right now in non-server market and based on roadmaps, they're likely to stay ahead. Of course Apple is a different story.

Personally, I don't think that either Intel or AMD lives up to the hype. AMD CPUs have become more "fashionable" in the recent years, but the simply truth is that AMD has (mostly) caught up with Intel in terms of performance. As to Intel... I am not convinced with their approach. Golden Cove is IMO disappointing and as to the rest, Intel is simply throwing more cores at the problem. That is hardly sustainable.

The interesting stuff will start happening when the Meteor Lake and multi-chip module tech arrives. Until then they are just plugging the leaks on a rusty bucket.

 

[senttoschool]

What matters is the purpose. And Intel's purpose behind heterogenous CPU design is very different than Apples. Intel simply does not know how to make fast cores that are energy efficient. This is why they are compensating for their lack of computational throughout with many smaller slower cores that are more efficient. That's it.

The only company ahead of Intel in "fast cores" at a reasonable power rating is Apple. Intel currently has the fastest ST if you discount power.

Everyone, including AMD, Qualcomm, and ARM is now behind Intel in "fast cores". Let's see what Qualcomm's Nuvia team comes up with later this year.

AMD Readying 16-core "Zen 4" CCDs Exclusively for the Client Segment with an Answer to Intel E-cores?​

AMD already declared the CPU core counts of its EPYC "Genoa" and "Bergamo" processors to top out at 96 and 128, respectively, a core-count believed to have been facilitated by the larger fiberglass substrate of the next-gen SP5 CPU socket, letting AMD add more 8-core "Zen 4" chiplets, dubbed CPU...

www.techpowerup.com

This isn't big.Little. These are two different CPUs. Genoa is the general purpose Zen4, a direct step up from Zen3. Bergamo is a "cloud-optimized" CPU. Zen4 won't be big.Little. AMD has never indicated that Zen4 will be big.Little. This is just some fan-theory in response to Alder Lake.

Golden Cove is IMO disappointing and as to the rest, Intel is simply throwing more cores at the problem. That is hardly sustainable.

Golden Cove is very impressive in the x86 world. But you're comparing everything to Apple Silicon nowadays so yea, it's not that impressive to you.

 

[senttoschool]

Intel simply does not know how to make fast cores that are energy efficient. This is why they are compensating for their lack of computational throughout with many smaller slower cores that are more efficient. That's it.

FYI, the wall that Intel is hitting might be near Apple Silicon too. A15 increased its efficiency cores much more so than its high-performance cores and we're no longer seeing the drastic increases in raw CPU performance that we previously saw with the A-series. I think Apple will likely have to increase the number of smaller cores to compensate for the slow down in high-performance cores in the near future.

Even still, Intel's Alder Lake is delivering 15-20% faster ST over last-gen. That's nothing to scoff at and they're the king of the consumer x86 world again.

 

[Andropov]

big.Little simply means you have big, hungry, power cores and smaller, slower efficiency cores that exist in the same CPU. Intel's implementation isn't any different than Apple's. It's just that Intel is pushing both the P cores and E cores harder than Apple.

The difference lies in what they're using the extra E cores for. Apple has the E cores as a mean to reduce idle or light-workload power consumption. That's why the M1 Pro / Max only has two of them. On the other hand, Intel is using its "E" cores to boost multicore performance, since they're smaller than their "P" cores and use less power (they're lower clocked too), so they can afford to have a few super power-hungry "P" cores for their single-core and low thread count workloads. That avoids the (in my opinion, huge) problem that high core count Intel chips had in the past, where the single core performance was horrendous, but they're hardly efficient cores. In fact, they're just ~50% slower than the P cores!

There's also scheduling difficulties to be sorted with such a model. Apple pushes tasks to the fastest available cores first, so if a (non-low priority) task is scheduled (spills) to a E core, that means the CPU as a whole is already at ~94% of max multicore throughput. Until you reach that point, all tasks are run into equally fast cores.

Intel, on the other hand, has E cores contributing to a much greater extent to that multicore performance, so for the i9 12900HK (6 Perf, 8 Efficiency), if everything is scheduled to the P cores first, tasks spill to the E cores at ~52% of max multicore throughput.

That means there's a much wider range of CPU utilization where proper thread scheduling is critical to the responsiveness of the system (so the tasks you care the most about are not unknowingly put into the slower cores, with the CPU still only at ~half utilization). Intel's solution is Thread Director, which uses user focus (really) to heuristically determine this, but that's a very naive approach. Apple has a proper API to label task priorities (GCD's QoS), although in all fairness, Intel couldn't do that since they have no control over the OS.

If Intel wants, they can simply dial it down and increase efficiency.

Alder Lake is incredibly efficient (for x86) when you don't push it past its limits.

They *could*, but the performance drop would be abysmal. Just look at the benchmark scores for unplugged Alder Lake systems. They're literally cut in half. A battery-powered Alder Lake laptop barely competes with a M1 (non Pro/Max) MacBook. And still manages to consume several times more power.

We know nothing about AMD's big.Little plan. That's not happening until Zen5 which might be 2.5-3 years away.

People like to crap on Intel. But they're clearly ahead of AMD right now in non-server market and based on roadmaps, they're likely to stay ahead. Of course Apple is a different story.

I don't know. I agree Intel currently offers a better range of CPUs now, but AMD's offerings are already a year old, and it seems to me that Intel pushed absolutely everything they had, ready or not (like the weird AVX-512 instruction disparity between core types), into this generation. So I doubt they have anything meaningful next year. Surely their roadmap says they will, but Intel's roadmaps don't exactly have a great track for reliability right now.

 

[BigPotatoLobbyist]

Gamers drive the opinion in the consumer computing space unfortunately. It’s a really bleak landscape anyway. Most purchases are driven by soulless IT departments who’s only concern is to buy the cheapest crap they can get away with and “knowledge” is driven by gamers who have no clue what they are talking about.

Yes. Latter in particular are vehemently opposed to ARM in general I've found, be it custom cores like Apples or reference from ARM ltd. Weird conservative impulses all around with that lot.

 

[diamond.g]

@cmaier could Intel see a performance increase going from 4/6-wide to 8-wide? What would be the downsides of doing so?

 

[BigPotatoLobbyist]

The difference lies in what they're using the extra E cores for. Apple has the E cores as a mean to reduce idle or light-workload power consumption. That's why the M1 Pro / Max only has two of them. On the other hand, Intel is using its "E" cores to boost multicore performance, since they're smaller than their "P" cores and use less power (they're lower clocked too), so they can afford to have a few super power-hungry "P" cores for their single-core and low thread count workloads. That avoids the (in my opinion, huge) problem that high core count Intel chips had in the past, where the single core performance was horrendous, but they're hardly efficient cores. In fact, they're just ~50% slower than the P cores!

There's also scheduling difficulties to be sorted with such a model. Apple pushes tasks to the fastest available cores first, so if a (non-low priority) task is scheduled (spills) to a E core, that means the CPU as a whole is already at ~94% of max multicore throughput. Until you reach that point, all tasks are run into equally fast cores.

Intel, on the other hand, has E cores contributing to a much greater extent to that multicore performance, so for the i9 12900HK (6 Perf, 8 Efficiency), if everything is scheduled to the P cores first, tasks spill to the E cores at ~52% of max multicore throughput.

That means there's a much wider range of CPU utilization where proper thread scheduling is critical to the responsiveness of the system (so the tasks you care the most about are not unknowingly put into the slower cores, with the CPU still only at ~half utilization). Intel's solution is Thread Director, which uses user focus (really) to heuristically determine this, but that's a very naive approach. Apple has a proper API to label task priorities (GCD's QoS), although in all fairness, Intel couldn't do that since they have no control over the OS.


They *could*, but the performance drop would be abysmal. Just look at the benchmark scores for unplugged Alder Lake systems. They're literally cut in half. A battery-powered Alder Lake laptop barely competes with a M1 (non Pro/Max) MacBook. And still manages to consume several times more power.


I don't know. I agree Intel currently offers a better range of CPUs now, but AMD's offerings are already a year old, and it seems to me that Intel pushed absolutely everything they had, ready or not (like the weird AVX-512 instruction disparity between core types), into this generation. So I doubt they have anything meaningful next year. Surely their roadmap says they will, but Intel's roadmaps don't exactly have a great track for reliability right now.

Intel E cores are basically their profit-sparing means of competing in throughput, the actual idle efficiency gains are fairly minimal though there may be some QoS benefits. Still, the fact that they're pushing them so hard or that even in modest clock ranges their ST performance (the E cores) is so horrible is telling - it's basically a bandage on the real issue of sorts, which is that they do not have the kind of IPC Apple do.

If you have a strong base microarchitecture for your primary performance cores, one benefit is you don't have to choose high-performance cell libraries prone to massively increased leakage albeit able to boost to high voltages & clock rates — because performance is good enough at more modest clocks. This, in addition to the logic density afforded by TSMC processes (though even on N7, N10 Apple's cores were already quite efficient and powerful) is really the key - and one I trust Nuvia will have in mind and realize in contrast to the power-hungry strategies employed by AMD and Intel.

 

[senttoschool]

The difference lies in what they're using the extra E cores for. Apple has the E cores as a mean to reduce idle or light-workload power consumption. That's why the M1 Pro / Max only has two of them. On the other hand, Intel is using its "E" cores to boost multicore performance, since they're smaller than their "P" cores and use less power (they're lower clocked too), so they can afford to have a few super power-hungry "P" cores for their single-core and low thread count workloads. That avoids the (in my opinion, huge) problem that high core count Intel chips had in the past, where the single core performance was horrendous, but they're hardly efficient cores. In fact, they're just ~50% slower than the P cores!

But Intel's E cores ARE being used for background tasks, hence, more efficient cores are being used for low priority threads. That is big.Little.

Now, the reason why Intel added 8+8 configuration is because they wanted to beat AMD's 5950x, which they succeeded. If they went with 8+2, would that make them "more big.Little"?

They *could*, but the performance drop would be abysmal. Just look at the benchmark scores for unplugged Alder Lake systems. They're literally cut in half. A battery-powered Alder Lake laptop barely competes with a M1 (non Pro/Max) MacBook. And still manages to consume several times more power.

We're talking about AMD vs Intel here in the last several posts.

I've posted a detailed analysis before on ADL vs M1 Pro/Max power usage. I'm well aware of M1's efficiency advantage.

I don't know. I agree Intel currently offers a better range of CPUs now, but AMD's offerings are already a year old, and it seems to me that Intel pushed absolutely everything they had, ready or not (like the weird AVX-512 instruction disparity between core types), into this generation. So I doubt they have anything meaningful next year. Surely their roadmap says they will, but Intel's roadmaps don't exactly have a great track for reliability right now.

Chips get designed years in advance. They can't just decide last minute to put everything into ADL like you suggest. Raptor Lake will be Intel's second crack at big.Little and rumors are that it will be a fairly substantial boost. The bigger boost will be Meteor Lake which will use Intel 4 process and will come only 2 quarters after Raptor Lake. By the time Zen5 comes, Intel will be on their 3rd or 4th big.Little iteration.

Again, there is zero indication that Zen4 will be big.Little. What @leman posted is an unsubstantiated rumor, theory, wishlist, whatever it is. Genoa and Bergamo are different variations of the same design, optimized for different server workloads. They're even optimized to use different TSMC nodes. They're not to be combined into a single CPU.

Intel, on the other hand, has E cores contributing to a much greater extent to that multicore performance, so for the i9 12900HK (6 Perf, 8 Efficiency), if everything is scheduled to the P cores first, tasks spill to the E cores at ~52% of max multicore throughput.

I don't see anything wrong with this.

In most big.Little designs, there are more efficiency cores than power cores. Even for Alder Lake, it varies by a lot: 8/8, 6/8, 8/4 to 6/0, 2/0.

Intel is configuring the core counts to compete with AMD.

 

[BigPotatoLobbyist]

@cmaier could Intel see a performance increase going from 4/6-wide to 8-wide? What would be the downsides of doing so?

They already went 6-Wide with Golden Cove and massively expanded the reorder buffer & improved the branch predictor.
They also did so while retaining high clocks and a (relatively) less dense implementation of the P cores, in part in order to allow for high clockspeeds unlike denser implementations.


They certainly proved it possible to go wider with X86, though it remains to be seen if it is possible to implement an 8-wide uarch as well on X86. uOP caches probably help — but Cliff may be a better person to ask on just how much they can mitigate the variable instruction length penalties.

 

[diamond.g]

Yes. Latter in particular are vehemently opposed to ARM in general I've found, be it custom cores like Apples or reference from ARM ltd. Weird conservative impulses all around with that lot.

The vocal lot are all about big numbers, customization etc. The vocal lot is also the smallest lot in terms of PC Gaming. PCMR. From that lens, they see Arm as too rigid. It doesn't offer the ability to Overclock, or use phase change cooling for bragging rights, etc.
PCMR for gaming is like Blue bubblers for messaging.

 

[BigPotatoLobbyist]

The vocal lot are all about big numbers, customization etc. The vocal lot is also the smallest lot in terms of PC Gaming. PCMR. From that lens, they see Arm as too rigid. It doesn't offer the ability to Overclock, or use phase change cooling for bragging rights, etc.
PCMR for gaming is like Blue bubblers for messaging.

Makes sense. I find it all weird and annoying, reminds me of gearheads and pimped out vehicles with a kind of aesthetic appeal reminiscent of Jersey Shore or some ****. But whatever, I just wish they wouldn't lie about things.

 

[diamond.g]

Makes sense. I find it all weird and annoying, reminds me of gearheads and pimped out vehicles with the kind of aesthetic appeal reminiscent of Jersey Shore or some ****. But whatever, I just wish they wouldn't lie about things.

The interesting part in this is that Intel (specifically) caters to that market (the K skus are all unlocked). I would just chalk it up to people want their things to be their things and in this case Arm (by extension Apple I guess) doesn't really cater to that idea.

To be honest tinkering with that stuff somethings can be fun.

 

[TSE]

Intel: Fastest CPUs, most heat and energy cost
Apple: Most efficient CPUs
AMD: Somewhere in the middle

I'd only get an AMD laptop if I had to get a PC. In the real world where laptops aren't the 7 pound sample units chosen to hand over to Youtubers to benchmark specifically by Intel because of their massive cooling, the story is not going to change. These Alder Lake CPUs are going to be cooled terribly and will actually probably perform worse.

 

[senttoschool]

Intel: Fastest CPUs, most heat and energy cost
Apple: Most efficient CPUs
AMD: Somewhere in the middle

I'd only get an AMD laptop if I had to get a PC. In the real world where laptops aren't the 7 pound sample units chosen to hand over to Youtubers to benchmark specifically by Intel because of their massive cooling, the story is not going to change. These Alder Lake CPUs are going to be cooled terribly and will actually probably perform worse.

You should wait until reviews of low-powered Alder Lake. The first reviews out the gate are for the most power-hungry Alder Lake laptop chip: 12900HK. The number one goal 12900HK tried to accomplish is to lead all mobile chips (including Apple Silicon) in performance. It succeeded. But lower-powered Alder Lake chips are more interesting to me anyways.

Alder Lake is expected to beat AMD in laptops in both speed and battery life.

By the way, even Tiger Lake is already highly competitive with AMD in laptops.

 

[TSE]

You should wait until reviews of low-powered Alder Lake. The first reviews out the gate are for the most power-hungry Alder Lake laptop chip: 12900HK. The number one goal 12900HK tried to accomplish is to lead all mobile chips (including Apple Silicon) in performance. It succeeded. But lower-powered Alder Lake chips are more interesting to me anyways.

Alder Lake is expected to beat AMD in laptops in both speed and battery life.

By the way, even Tiger Lake is already highly competitive with AMD in laptops.

Good note - I'll wait for low-power Alder Lake to make any sweeping judgments. But I'm not impressed with the Alder Lake CPUs that have come out so far.

 

[senttoschool]

Good note - I'll wait for low-power Alder Lake to make any sweeping judgments. But I'm not impressed with the Alder Lake CPUs that have come out so far.

That's interesting.

I'm actually very impressed with Alder Lake. Sure, it's still nowhere close to being as efficient as M1 right now. But it's extremely impressive compared to what AMD has at the moment and what Intel has churned out for the last 6 years. I'm even more impressed by the roadmap Intel has. Intel has become extremely aggressive. I like it.

If Apple isn't aggressive such as releasing a new M series every 18-24 months instead of yearly, Intel might eventually catch up. I'm actually having a hard time imagining Apple surpassing Intel in raw single core performance in the next few years like they did with the original M1 back in 2020.

 

[Andropov]

Intel E cores are basically their profit-sparing means of competing in throughput, the actual idle efficiency gains are fairly minimal though there may be some QoS benefits. Still, the fact that they're pushing them so hard or that even in modest clock ranges their ST performance (the E cores) is so horrible is telling - it's basically a bandage on the real issue of sorts, which is that they do not have the kind of IPC Apple do.

I don't think there's anything inherently wrong with that. If they can make a workload faster and more efficient by throwing more (smaller) cores into it, great! That's what matters in the end. There's the question of whether they will hit a ceiling with this approach, though. Too many CPU cores will have other problems (cache synchronization between cores, tasks not be parallelizable enough...). And if a task truly is parallelizable enough, there's a good chance it's better on the GPU.

But Intel's E cores ARE being used for background tasks, hence, more efficient cores are being used for low priority threads. That is big.Little.

Now, the reason why Intel added 8+8 configuration is because they wanted to beat AMD's 5950x, which they succeeded. If they went with 8+2, would that make them "more big.Little"?

I don't see much point in discussing the semantics of whether calling it a big.LITTLE design is appropriate or not, at long as we all understand how the cores are configured in both systems. Which I believe we do. My point was that while both systems use heterogenous cores, Intel's are more of a 'medium' core than a small and efficient core like Apple's.

Chips get designed years in advance. They can't just decide last minute to put everything into ADL like you suggest. Raptor Lake will be Intel's second crack at big.Little and rumors are that it will be a fairly substantial boost. The bigger boost will be Meteor Lake which will use Intel 4 process and will come only 2 quarters after Raptor Lake. By the time Zen5 comes, Intel will be on their 3rd or 4th big.Little iteration.

When I said I believe they rushed Alder Lake, I didn't mean in a span of months. The Apple Silicon transition started a year and a half ago, and it was an open secret for a year before that. They didn't had access to the M1 itself until late 2020, but the A13X had been around since 2018 (which is basically what apple ended up using in their Developement Transition Kits), so it's not like they had no idea of what Apple was going to release. Some forum members here made spot-on estimates of what Apple was going to come up with for their fist Mac CPU a year in advance, so it stands to reason that a company with the resources Intel has was able to come to similar conclusions. Same thing happened before the release of the M1 Pro/Max.

The 12900HK is clearly tuned to beat the M1 Pro/Max (it's not a coincidence that they are only +4% faster than the M1 Pro/Max in multicore (Geekbench)). Releasing it later risked Apple coming up with the M2 and not being able to say they have the fastest laptop CPU in the world as planned. The design is clearly half baked in some aspects (it's still not known if AVX-512 instructions will be available on the P cores or not, and while it increasingly looks like it won't be possible in the future, the units are not fused off, which points to a very late minute change in direction).

About the roadmap, Meteor Lake, etc. that's just fantasy.

 

[TSE]

That's interesting.

I'm actually very impressed with Alder Lake. Sure, it's still nowhere close to being as efficient as M1 right now. But it's extremely impressive compared to what AMD has at the moment and what Intel has churned out for the last 6 years. I'm even more impressed by the roadmap Intel has. Intel has become extremely aggressive. I like it.

If Apple isn't aggressive such as releasing a new M series every 18-24 months instead of yearly, Intel might eventually catch up. I'm actually having a hard time imagining Apple surpassing Intel in raw single core performance in the next few years like they did with the original M1 back in 2020.

The reason I personally find Alder Lake so far disappointing is because the ones that have been out have zero nuance to their design. It's an engineer's wet dream. Unfortunately for engineers, experience matters. When your CPUs require a 7 pound chassis and a 2 pound charger and your competition gets 4x the battery life in a chassis that's 40% lighter, it just isn't impressive to me.

BUT, you are right - let's wait for Alder Lake low energy models and compare.

 

[senttoschool]

I don't see much point in discussing the semantics of whether calling it a big.LITTLE design is appropriate or not, at long as we all understand how the cores are configured in both systems. Which I believe we do. My point was that while both systems use heterogenous cores, Intel's are more of a 'medium' core than a small and efficient core like Apple's.

big.Little or big.Medium. Does it matter? It still accomplishes the goal of using smaller, more efficient cores to do low priority jobs and high performance cores to do high priority jobs.

When I said I believe they rushed Alder Lake, I didn't mean in a span of months. The Apple Silicon transition started a year and a half ago, and it was an open secret for a year before that. They didn't had access to the M1 itself until late 2020, but the A13X had been around since 2018 (which is basically what apple ended up using in their Developement Transition Kits), so it's not like they had no idea of what Apple was going to release. Some forum members here made spot-on estimates of what Apple was going to come up with for their fist Mac CPU a year in advance, so it stands to reason that a company with the resources Intel has was able to come to similar conclusions. Same thing happened before the release of the M1 Pro/Max.

The 12900HK is clearly tuned to beat the M1 Pro/Max (it's not a coincidence that they are only +4% faster than the M1 Pro/Max in multicore (Geekbench)). Releasing it later risked Apple coming up with the M2 and not being able to say they have the fastest laptop CPU in the world as planned. The design is clearly half baked in some aspects (it's still not known if AVX-512 instructions will be available on the P cores or not, and while it increasingly looks like it won't be possible in the future, the units are not fused off, which points to a very late minute change in direction).

About the roadmap, Meteor Lake, etc. that's just fantasy.

Chips get designed and specced out multiple years in advance. I doubt the M1 influenced ADL design much except for how much wattage Intel was willing to push.

So your basis for ADL being half-baked is that some reviewers found a way to enable AVX-512?

Not sure why you think it's a fantasy. It's Intel's official roadmap.

I don't want to get too ahead of ourselves here. All my posts today focused mostly on Intel vs AMD, not Intel vs M1. I'm trying to dispel the myth here that Intel is still behind AMD in desktop and upcoming ADL laptops.

 

[leman]

big.Little or big.Medium. Does it matter? It still accomplishes the goal of using smaller, more efficient cores to do low priority jobs and high performance cores to do high priority jobs.

How many low priority jobs you think there are? You don’t need eight or sixteen(!!) cores to take care of them. Again: Intels E-cores are not there to take care of low priority tasks (that’s just the side effect), they are there to improve the performance on sustained multi core workloads in a thermally constrained context.

 

[senttoschool]

How many low priority jobs you think there are? You don’t need eight or sixteen(!!) cores to take care of them. Again: Intels E-cores are not there to take care of low priority tasks (that’s just the side effect), they are there to improve the performance on sustained multi core workloads in a thermally constrained context.

That's only on the highest end to beat 5950x.

The mass consumer ADL aren't 8+8.

Generally, in big.Little designs, there are more efficiency cores than high powered ones. This is exactly how it's configured on mobile ADL.

 

[Andropov]

big.Little or big.Medium. Does it matter? It still accomplishes the goal of using smaller, more efficient cores to do low priority jobs and high performance cores to do high priority jobs.

I agree that it gets the job done. But it's not for low priority jobs only. It's almost half the power of the chip. It does the low priority jobs AND the high priority ones. As @leman said, there aren't that many low priority jobs. Given how powerful their E cores are compared to Apple's, if Apple managed to use only two E cores for the low-priority jobs, Intel would have more than enough with just one. They have EIGHT in most Alder Lake mobile configurations, never less than four.

Not sure why you think it's a fantasy. It's Intel's official roadmap.

Well that's why I think it's fantasy.

 

[senttoschool]

I agree that it gets the job done. But it's not for low priority jobs only. It's almost half the power of the chip. It does the low priority jobs AND the high priority ones. As @leman said, there aren't that many low priority jobs. Given how powerful their E cores are compared to Apple's, if Apple managed to use only two E cores for the low-priority jobs, Intel would have more than enough with just one. They have EIGHT in most Alder Lake mobile configurations, never less than four.

I'm not sure why it's hard for you and @leman to understand.

On desktop, yes, ADL's efficiency cores are mainly used to boost MT performance. And they accomplish that goal well because they're matching and beating AMD's best but at a cheaper price.

And then on mainstream laptops, where efficiency matters more than having the most multithreaded performance, Intel configures ADL to have more efficiency cores than power cores. This is exactly how all big.Little designs are (the exception being M1 Pro/Max). All Qualcomm and Apple A series big.Little chips have more efficiency cores than high power cores. In fact, most Android SoCs have just 1 high power core, a few mid-tier cores and more low-power cores.

 

[senttoschool]

Well that's why I think it's fantasy.

Sure. Intel's roadmap got screwed by 10nm delays over and over again in the last 5 years. Nevertheless, the roadmaps are the best available dates we have. And the roadmaps suggests that Intel probably won't relinquish the lead over AMD in the next few years.