Intel Alder Lake vs. Apple M1

[pshufd]

Sure but Apple is the OEM. Apple has generally always offered the balanced build. This is true for Apple’s direct competitors because tautologically they are direct competitors. So from the *perspective of Apple* it’s no different except they don’t have to pay Intel and AMD’s margins. Apple doesn’t sell its chips direct to customers or businesses so the lack of modularity isn’t an issue. It’s not the same business. So yes they can increase core counts and yes those will come with increased other things too. That’s also true for buying most Dell, HP, Lenovo, etc … computers. Can you find the machines that cater to a specialized market? Yes of course but that’s not a standard consumer build either. The PC market is massive. But Apple doesn’t compete in all of it. For the segments it competes in it's reckoned to be paying less than it was before (@EntropyQ3 's post above).

I think we may be talking past each other here. The original point of @BigPotatoLobbyist ‘s post was about architecture. What would Apple’s dot on that graph look like with an increased number of cores? Far better. Could they do it? Yes. Will they do it? Yes but probably not for laptops because Apple so far has prioritized noise and battery life.

The M1 is good enough for the vast majority of computing users but it's frustrating that you have to take the M1 Pro if you want a 16 inch laptop from Apple. The GB 5 MC on my M1 Pro is 12,115 and OpenCL is 38,359. I would be fine with 7K and 20K that you get with the base M1 if they offered it with 32 GB of RAM and a 16 inch display. But, given the choices, I got the 16 M1 Pro. It works but it's complete overkill for me. And I'd guess that there are lots of other customers that got the 16 for the screen, display, speakers; but really don't need the CPU and GPU.

I think that Apple will eventually produce something that better fits my needs but they're doing a great job at extracting consumer dollars given the choices in what they sell. The M1 was just such a big jump in performance from what we've been used to that it's screwed up the industry.

It is nice to have the options to pick the CPU and the GPU as needs vary so widely. I am not complaining as the 16 is a fantastic system. But it would be like me buying the Toyota Avalon because the Corolla doesn't have a few features that I really want.

I do like that Apple is making their own CPUs and skipping AMD and Intel. The AMD CPU shortages were not pleasant a while ago and the GPU shortages aren't pleasant right now. I'm fortunate to have bought GPUs before the pandemic but I wanted to upgrade during the pandemic, and, forget it. I think that the CPU situation is really good right now because Intel is pricing aggressively to win back marketshare and the premium is no longer in AMD CPUs. I don't plan to do another Windows build anytime soon but a 5900X would be more than enough for me. I am not going to do anything until I can see what the motherboard situation looks like. If Intel or AMD can put out a motherboard that will be good for five years, then I'll look. In the meantime, it's all Apple for me.

I'm selling my M1 mini as I replaced it with two old iMac 27s which cost less than the mini. I paid $1,099 for the mini (16/512) and $500 for a 2014 iMac 27 and $100 for a 2010 iMac 27 and $250 to get both of them to 32 GB RAM each. These two iMacs are far more usable than the mini for what I do. I don't need absolute CPU and GPU performance - what I want is great screens, speakers, a desktop with minimal cable clutter, and lots of RAM. Used equipment is really good for that.

 

[pshufd]

Absolutely, but then you have to consider adding a GPU's die area to ADL ... i.e. if we're going to consider the whole package, then we have to consider the whole package. Basically is Apple paying more for the similar solution that they would have to get if they went out and bought AMD GPUs and Intel CPUs? Probably not. That's where I was coming from in my response - also see @EntropyQ3 's post.

This is somewhat extreme with GPUs right now. I think that Apple bumped the prices of the 2020 iMac by a few hundred dollars in the past year or two as GPU prices shot up. It's still a better deal than buying a GPU on the open market but that couple of hundred dollars is pure market pressure pricing that Apple can squeeze out by making their own chips. I am hoping that the GPU situation improves with Intel getting into the market but I'm not completely sure that it will be anytime soon.

 

[crazy dave]

@pshufd I agree. One of the rumors was that Apple had thought about offering a 15” Air-like or base pro device in the M2 generation but decided against it for now. If that’s true, hopefully they reconsider for the M3 as I agree that’s a hole in their lineup and it should be filled sooner rather than later.

I also wouldn’t mind some additional modularity in the next SOC generations. Maybe as multidie chips percolate downwards in M3/4 they’ll become more about Legos, combining what you want together, rather than just bigness. We won’t get, nor do we need, the overwhelming set of SKUs Intel produces for its CPUs alone but I can see such multidie chips as offering a potential balance depending on design. That’s awhile from now though and who knows what Apple has in mind that far down the pike.

I wasn’t just referring to the current silicon shortages - obviously that’s making life painful for everyone at all levels (although apparently it’s actually worst for “old” nodes producing things like automotive chips - like it’s bad for consumer cutting edge nodes, but cataclysmic for capacity in older nodes). Even without that Apple is probably spending less than they would be. That just exacerbates the issue (though OEMs feel the pinch less than DIYers).

 

[senttoschool]

But at the end of the day, we do have single-core results and we do know that Golden Cove is significantly less power efficient than zen3.

Where is the data for this claim?

 

[JouniS]

Sure but Apple is the OEM. Apple has generally always offered the balanced build. This is true for Apple’s direct competitors because tautologically they are direct competitors.

Desktop Macs have often had options for rather imbalanced builds, because you can choose the CPU and the GPU separately. For example, you can buy an iMac with a 10-core i9, 128 GB RAM, and Radeon Pro 5300. Apple's competitors extend the choice to laptops, because they have wider product ranges. You can often choose from a cheap laptop, an ultrabook, a premium laptop with integrated or discrete graphics, a gaming laptop, and a mobile workstation.

So from the *perspective of Apple* it’s no different except they don’t have to pay Intel and AMD’s margins. Apple doesn’t sell its chips direct to customers or businesses so the lack of modularity isn’t an issue. It’s not the same business. So yes they can increase core counts and yes those will come with increased other things too.

Apple is in the business of selling midrange mass market consumer devices and related services. From their perspective, five people buying a $1000 device each and then spending more money on apps and other services are better than a single high-end customer buying a $5000 device. The high-end customer won't spend as much money on services as the five others combined. Because Apple can only buy a limited number of chips from TSMC, the $5000 device won't have a giant chip that prevents them from selling eight $1000 devices. The giant chip will more likely be reserved for $10000 workstations.

Apple's competitors see the market differently, because they sell hardware, not services.

I think we may be talking past each other here. The original point of @BigPotatoLobbyist ‘s post was about architecture. What would Apple’s dot on that graph look like with an increased number of cores? Far better. Could they do it? Yes. Will they do it? Yes but probably not for laptops because Apple so far has prioritized noise and battery life.

Apple's power efficiency is largely a consequence of their choice to make a SoC. Their inability to make consumer chips with a high CPU core count is another consequence. You can't accept some consequences of their architectural choices while ignoring others.

 

[senttoschool]

Just forewarning, I've been trying to explain this to him ... you're basically retreading the last 3 hours of my conversation with him. Although you are far harsher on both AMD and Intel than I.?

Explaining to me that Zen2 and Zen3 are more efficient than ADL? Where's the data?

Is everyone just going to completely ignore real world results? We actually have them.

 

[pshufd]

@pshufd I agree. One of the rumors was that Apple had thought about offering a 15” Air-like or base pro device in the M2 generation but decided against it for now. If that’s true, hopefully they reconsider for the M3 as I agree that’s a hole in their lineup and it should be filled sooner rather than later.

I also wouldn’t mind some additional modularity in the next SOC generations. Maybe as multidie chips percolate downwards in M3/4 they’ll become more about Legos, combining what you want together, rather than just bigness. We won’t get, nor do we need, the overwhelming set of SKUs Intel produces for its CPUs alone but I can see such multidie chips as offering a potential balance depending on design. That’s awhile from now though and who knows what Apple has in mind that far down the pike.

I wasn’t just referring to the current silicon shortages - obviously that’s making life painful for everyone at all levels (although apparently it’s actually worst for “old” nodes producing things like automotive chips - like it’s bad for consumer cutting edge nodes, but cataclysmic for capacity in older nodes). Even without that Apple is probably spending less than they would be. That just exacerbates the issue (though OEMs feel the pinch less than DIYers).

One other approach that I wouldn't mind is if Apple produced Limited Edition products. They brought back the iPhone mini for generations 12 and 13 and I have a 13 and love it but it looks like it is gone next year. I'm okay with that but would like them to bring it back a couple of years down the road. I would not mind if they made a 15/16 Air for a year or two and then discontinued it for a while. I don't think that Apple thinks like this though. The M2 Air is going to be interesting as they could come up with improvements that cannibalizes MBP 14 sales.

 

[Fomalhaut]

I'm selling my M1 mini as I replaced it with two old iMac 27s which cost less than the mini. I paid $1,099 for the mini (16/512) and $500 for a 2014 iMac 27 and $100 for a 2010 iMac 27 and $250 to get both of them to 32 GB RAM each. These two iMacs are far more usable than the mini for what I do. I don't need absolute CPU and GPU performance - what I want is great screens, speakers, a desktop with minimal cable clutter, and lots of RAM. Used equipment is really good for that.

True. We are often tempted by the "latest and greatest" without really researching whether our needs could be met by older hardware, which can often be had at knockdown prices. I've bought a couple of Xeon workstations in the past that were about 2-3 years old, which companies were offloading and selling through "used server" vendors. It was surprising that something that cost over $5000 when new was being sold 3 years later for less than $1000, but it was a great deal and perfectly adequate for what I wanted, and a great platform to upgrade with faster SSDs and additional PCIe cards and storage.

Having just sold a 2-year old MBP16 and recouping about 43% of my original cost, I've seen the other side of computer depreciation. It was a bit painful to realise that my ownership cost was about $1000 / year. In a year or two, used M1/M1 Pro/Max Macs will probably represent a pretty good deal for those who don't require top performance, but still want a capable computer.

Let's face it, unless we upgrade within 1-2 years, most of use will already be using "legacy hardware" due to the simple passage of time. If you can live with that 2-3 years in the future, maybe you can live with it now? It's not as though computers wear out the same way as cars, so used machines are often in just as good condition as those bought new a few years ago.

 

[Gerdi]

What I do find from this is that Intel is quite a bit ahead of AMD chips, despite, in my opinion, being on an inferior Node. Intel 7 is still worse than TSMC 7nm. This basically confirms that ADL does have better performance/watt than Ryzen when it's configured more optimally in its power curve.

That is the conclusion you cannot possibly draw, and i did show precisely this in one of my previous post. We already concluded that you can say that the particular product 12700HX vs. 5900HX is more efficient at iso-power. This is the only thing you can instantly see.

However you cannot generalize this for the architecture e.g. Alder Lake vs. Zen 3 because you would have to compare at iso-voltage. If you would do this, you would see that the Alder Lake architecture is still far behind Zen 3 if the metric is power efficiency.
At this point I can only suggest that you go back to the graphs i did draw and try to understand what the consequences are - you should start asking yourself why the blue graph (Alder Lake) looks like beating the orange graph (Zen3) in power efficiency, while the grey graph, which is again Zen 3 handily beats Alder Lake.

 

[Andropov]

PS. Performance/watt is not all that matters. What actually matters is "desired performance by user / watt". If you try to push Zen2 to have equivalent performance to ADL, I'm 100% sure that you'd need far more power and probably LN2 to achieve that goal.

If you go by performance/watt as the only metric, then some little tiny CPU would beat the pants off the M1, A15, Intel, AMD, etc. Performance does not scale linearly with power.

How is this, er, desired performance by user / watt measured and which units does it have? Who is this 'user'? Is he made of 90% platinum and 10% iridium and kept in the basement of a fancy research center?

In all seriousness, it's convenient for your argument that this user desires a performance level that is in a much better spot for Intel's performance/power curve than it is on AMD's. This hypothetical user may as well be sufficiently happy with less performance if that means better battery life. That is, on the other extreme of the power/performance curve, can ADL cores be brought down to Zen3 power consumption levels?

Because ADL has more cores. You are comparing a 6+8 design to a 8-core design on tests that are pretty much optimal for multiple cores. You can always improve efficiency by throwing more lower clocked cores at the problem. But that does not make your cores more efficient. Basically, this is exactly the same situation thst we had in AMD vs. Intel but reversed. Folks were praising AMD in 15 and 30W bracket because it did so well in throughput benchmarks compared to Intel, but that was 8 cores vs. 4. Now Intel did exactly the same thing.

But at the end of the day, we do have single-core results and we do know that Golden Cove is significantly less power efficient than zen3. At the same time though, ADL is faster. That aspect of Intel vs. AMD has not changed in years. AMD Zen was always slightly slower, but 30% or some more efficient than Intel cores, and AMD had the tech and yields to ship 8 cores even at the low end. Now Intel put more low-power cores to close the gap, and that’s exactly what we see.

There is no innovation here. No progress. No interesting technical solutions. Just “oh yeah? then I bring more dudes and we’ll mess you up”. Frankly, Apples refusal to play on this level is one of the main reasons why I like their architecture so much. They focus on things that actually matter for real users.

Basically this (and what @crazy dave said). Adding more cores will always improve the power efficiency of a chip (unless there are so many that their static power consumption becomes bigger than the dynamic power of a system with less cores, but I believe you'd need an absurd amount of cores to reach that point). That means that:

• You'll have a more power efficient product on the market (which is what @senttoschool keeps pointing out).
• That's it.

It doesn't say anything about the μarch of the cores. If we're comparing Golden Cove cores to AMD's Zen3, it doesn't make sense to say: hey look! This particular product, with a different number of cores, is more power efficient! That's must mean that it has a better μarch!

It doesn't. Products with the exact same core design can have wildly different power efficiencies if the number of cores is not fixed. There's nothing particularly interesting about what configurations they're offering. It's interesting if you're buying a computer this generation, sure, but it's hardly predictive about what future generations will bring.

If all of Intel's advantage is because they created a product with more cores, then AMD won't have a hard time catching up. What's Intel going to do next gen anyway? Add even more cores? They already have almost twice as many cores. And we don't even know if they have the die area headroom to do that.

PS: Also note, all of these power consumption figures are on benchmarks running at full tilt. Most user-activity doesn't follow that pattern, computers are hardly ever run at max capacity. How do this CPUs fare in this scenario? I bet having a million CPU cores lying around doing nothing isn't helpful for power consumption.

 

[leman]

Where is the data for this claim?

Anandtech reviews for Zen3 and Golden Cove. ADL uses 25-30W in single-core SPEC. Zen3 maxes out at 20W per core.

The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity

www.anandtech.com

AMD Zen 3 Ryzen Deep Dive Review: 5950X, 5900X, 5800X and 5600X Tested

www.anandtech.com

 

[Gerdi]

Anandtech reviews for Zen3 and Golden Cove. ADL uses 25-30W in single-core SPEC. Zen3 maxes out at 20W per core.

The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity

www.anandtech.com

AMD Zen 3 Ryzen Deep Dive Review: 5950X, 5900X, 5800X and 5600X Tested

www.anandtech.com

In theory we can directly derive efficiency from here. If we would for SpecFp2017 Rate-1. Zen 3 is close to E-cores and much better than P-cores. I said in theory, because practically what i said earlier still holds, we need to compare iso-voltage. We do know that Zen 3 runs at 1.42V but we do not know voltage of the Intel parts. If the voltage is lower for Intel (and i have the suspicion that it is especially for the E-cores) such a calculation would over-estimate the efficiency for the Intel cores.

ps: Likewise, if we want to throw M1 into the mix, we would need the performance and power of an M1 core running with 1.42V (and according higher frequency) ideally having all critical paths using ULVT gates. Apparently ULVT gates help you with efficiency under high load.

 

[Gerdi]

I would quickly like to motivate why iso-voltage is so important. Efficiency is the quotient of a performance metric divided by a power metric. For performance we know it is proportional to IPC*f. For (dynamic) power we know it is P=C*f*V^2 for semiconductor circuits. C is the switching capacitance per cycle of your circuit, which is dependent on manufacturing process and the circuit itself (and for CPUs the application/task you are running). Lets assume the process and application is constant, so it is just a circuit property (e.g. an architectural property, which is proportional to the energy consumed per cycle). The other two factor V and f are just the operating point dependent, which we want want to get rid of when reasoning about architecture efficiency.

This yields:
Effective Efficiency = Performance/power = IPC*f/C*f*U^2 = IPC/C*U^2
Architectural Efficiency (which is independent of the operating point) = IPC/C

So if we are comparing two circuits (e.g. 2 CPU architectures) with respect to architectural efficiency we need to divide by V^2 or just compare the two circuits iso-voltage.

 

[crazy dave]

Explaining to me that Zen2 and Zen3 are more efficient than ADL? Where's the data?

Is everyone just going to completely ignore real world results? We actually have them.

View attachment 1955071

Everyone else responded so I’ll let their answers stand.

 

[JeepGuy]

I'm selling my M1 mini as I replaced it with two old iMac 27s which cost less than the mini. I paid $1,099 for the mini (16/512) and $500 for a 2014 iMac 27 and $100 for a 2010 iMac 27 and $250 to get both of them to 32 GB RAM each. These two iMacs are far more usable than the mini for what I do. I don't need absolute CPU and GPU performance - what I want is great screens, speakers, a desktop with minimal cable clutter, and lots of RAM. Used equipment is really good for that.

Just did this myself sold my 2020 mini, and picked up a 2014 Imac 5k for like $400, I really needed a better VPN solution than the mini + screen was providing. The real horse power is in my Workstation, that still need to run under Windows for work (Revit Architecture).

 

[Krevnik]

Is everyone just going to completely ignore real world results? We actually have them.

Anandtech reviews for Zen3 and Golden Cove. ADL uses 25-30W in single-core SPEC. Zen3 maxes out at 20W per core.

I think it’s worth emphasizing leman’s point here. Just because two CPUs are capped to not take more than X watts doesn’t mean their consumption are the same in a given scenario.

So to apply that to the chart, a number of the scenarios listed are single threaded scenarios. Excel being one of them. So is that because of higher efficiency, or higher power consumption (i.e. higher boosting that stays under the 45W limit on the one core under load)? 25-30W is 20-50% more power, which would explain the ST benchmarks in the graph. Without power measurements to go along with the benchmarks, you can’t say anything with certainty one way or another.

 

[Xiao_Xi]

Could Intel Core i7-1280P (base TDP of 28W) be the closest Intel's CPU to M1 in terms of performance and perf/watt?

LENOVO 4810RD0100 - Geekbench

Benchmark results for a LENOVO 4810RD0100 with an Intel Core i7-1280P processor.

browser.geekbench.com

MacBook Air (Late 2020) - Geekbench

Benchmark results for a MacBook Air (Late 2020) with an Apple M1 processor.

browser.geekbench.com

 

[mi7chy]

Could Intel Core i7-1280P (base TDP of 28W) be the closest Intel's CPU to M1 in terms of performance and perf/watt?

LENOVO 4810RD0100 - Geekbench

Benchmark results for a LENOVO 4810RD0100 with an Intel Core i7-1280P processor.

browser.geekbench.com

MacBook Air (Late 2020) - Geekbench

Benchmark results for a MacBook Air (Late 2020) with an Apple M1 processor.

browser.geekbench.com

More interested to see AMD 6000U. One thing that hasn't been covered with Alder Lake mobile is idle power consumption where 5800H does well at ~2W.

 

[Rigby]

This yields:
Effective Efficiency = Performance/power = IPC*f/C*f*U^2 = IPC/C*U^2
Architectural Efficiency (which is independent of the operating point) = IPC/C

So if we are comparing two circuits (e.g. 2 CPU architectures) with respect to architectural efficiency we need to divide by V^2 or just compare the two circuits iso-voltage.

This is all nice and well on a theoretical level, but there is a lot more to the power efficiency of a CPU design than IPC and switching capacitance, such as e.g. the design of the power delivery system and the node power management.

Why is simply taking the benchmark result divided by measured package power not sufficient to judge the power efficiency of a CPU again?

 

[leman]

Could Intel Core i7-1280P (base TDP of 28W) be the closest Intel's CPU to M1 in terms of performance and perf/watt?

LENOVO 4810RD0100 - Geekbench

Benchmark results for a LENOVO 4810RD0100 with an Intel Core i7-1280P processor.

browser.geekbench.com

MacBook Air (Late 2020) - Geekbench

Benchmark results for a MacBook Air (Late 2020) with an Apple M1 processor.

browser.geekbench.com

Doubtful. I would like to see the real power consumption, not just TDP. ADL TDP is the power dissipation expected under prolonged sustained operation. But it can use much more power for short periods of time. Since Geekbench has such a short duration, the CPU will probably operate at PL2 or even higher, and that is around 70W.

As to M1, we know for sure that the 4+4 config does not go beyond 20W in multi-core operation.

 

[crazy dave]

Desktop Macs have often had options for rather imbalanced builds, because you can choose the CPU and the GPU separately. For example, you can buy an iMac with a 10-core i9, 128 GB RAM, and Radeon Pro 5300. Apple's competitors extend the choice to laptops, because they have wider product ranges. You can often choose from a cheap laptop, an ultrabook, a premium laptop with integrated or discrete graphics, a gaming laptop, and a mobile workstation.

That's basically the M1 Pro 6+2/14 config minus the 128GB (and the 128 GB isn't a "consumer level" build). That Apple's broad range of competitors offer a broad range of products that Apple doesn't was *my point* - it's why I brought it up! You're taking a position that simply doesn't apply and hasn't to Apple.

Apple is in the business of selling midrange mass market consumer devices and related services. From their perspective, five people buying a $1000 device each and then spending more money on apps and other services are better than a single high-end customer buying a $5000 device. The high-end customer won't spend as much money on services as the five others combined. Because Apple can only buy a limited number of chips from TSMC, the $5000 device won't have a giant chip that prevents them from selling eight $1000 devices. The giant chip will more likely be reserved for $10000 workstations.

Maybe you need to be specific about what yo think is going to be a $10,000+ machine. A 40 core Mac? Sure. Most PC workstations with 40 cores would go for that too. A 14 core Apple machine for $10,000 as discussed in @Gerdi 's post? That's ... silly. Sorry, but it is. Unless again it comes with 128 GB of RAM.

Other than that your analysis is based on flawed assumptions. Apple most certainly doesn't view its hardware as a way to sell software. But even *if they did*, your argument still doesn't work. Apple making multiple levels of chips doesn't mean they don't sell 8 $1000 devices for every $5000 device they sell. Apple tailors their production to meet the various consumer levels they think they are going to sell to. There is no evidence that Apple's M1 product line is supply constrained because they are producing M1 Pros or Maxes. In fact all the M1 models have a delivery date of *tomorrow* if I were to order one right now at 7pm. The M1 Pros and Maxes are constrained which according to rumors is because of the mini-led screens, not SOCs. So no they don't have to double the price of a workstation chip to make it so unattractive that no one will buy it to funnel them to purchase the entry level machines.

Apple's power efficiency is largely a consequence of their choice to make a SoC. Their inability to make consumer chips with a high CPU core count is another consequence. You can't accept some consequences of their architectural choices while ignoring others.

Again, you need to define a high CPU core count. 10 cores is a high core count consumer level machine - especially given the relative performance envelope of those 10 cores. The CPU in the Pro/Max is basically a desktop Alder Lake i5 (also i7s depending on clocks and specific config - i.e. DDR4 vs 5). Yes, SOCs have certain limitations with regards to modularity. Obviously, but you've taken that to an extreme.

 

[Gerdi]

This is all nice and well on a theoretical level, but there is a lot more to the power efficiency of a CPU design than IPC and switching capacitance, such as e.g. the design of the power delivery system and the node power management.

Of course there is, but thats not my point at all. I am explaining something fundamental. (see below)

Why is simply taking the benchmark result divided by measured package power not sufficient to judge the power efficiency of a CPU again?

It is totally sufficient. However I did give answer to the question which points on the two graphs should be compared when reasoning about architectural efficiency. These are precisely the iso-voltage points as my calculation above did show.
In addition I did illustrate this with 2 different Zen 3 CPUs, 8 core and 16 core. It is very obvious that the 16 core CPU is vastly more efficient when comparing iso-power - but they are the same efficiency when comparing iso-voltage, which is the supposed outcome because both CPU's architecture are based on Zen 3 - which gives additional evidence that iso-voltage is the right way to compare.

 

[JouniS]

That's basically the M1 Pro 6+2/14 config minus the 128GB (and the 128 GB isn't a "consumer level" build).

That configuration had one of the highest-end Intel consumer CPUs of its generation with the lowest-end discrete AMD GPU of its generation. I would not call it very balanced. And the 128 GB is definitely a consumer level build, because it's in a consumer device powered by a consumer CPU.

Maybe you need to be specific about what yo think is going to be a $10,000+ machine. A 40 core Mac? Sure. Most PC workstations with 40 cores would go for that too. A 14 core Apple machine for $10,000 as discussed in @Gerdi 's post? That's ... silly. Sorry, but it is. Unless again it comes with 128 GB of RAM.

I expect something like $6k for a 10-core Mac Pro, $10k for a 20-core model, and $15k for a 40-core model, assuming reasonable RAM/SSD options and including a decent monitor and other peripherals. The entry-level model will probably have an M1 Max or equivalent, just like the current entry-level model is comparable to a much cheaper iMac.

There is no evidence that Apple's M1 product line is supply constrained because they are producing M1 Pros or Maxes.

It's more about planning than constraints. Apple has reserved a certain amount of capacity from TSMC and paid for it, and they expect to sell everything TSMC is able to produce. They wouldn't like it if they end up with large amounts of unused capacity or huge piles of devices nobody wants to buy. Because the M1 Max is 4x bigger than the M1, Apple has to choose whether it would rather sell four M1 devices or one M1 Max device. They choose the ratio between the devices and their prices in a way that they expect to maximize the profits.

The reality is obviously more complex, but that's the basic idea. Big businesses are centrally planned economies that would make the USSR jealous. Spare capacity was seen as a waste of money that was optimized away, which is why the supply chain crisis hit so hard.

Again, you need to define a high CPU core count. 10 cores is a high core count consumer level machine - especially given the relative performance envelope of those 10 cores.

In this context, a high core count is a relative concept. It's about adding more cores than people are used to, in an attempt to gain competitive advantage. 10-core consumer chips have been common for a few years, and both Intel and AMD have been toying with higher core counts for a while.

 

[senttoschool]

Anandtech reviews for Zen3 and Golden Cove. ADL uses 25-30W in single-core SPEC. Zen3 maxes out at 20W per core.

The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity

www.anandtech.com

AMD Zen 3 Ryzen Deep Dive Review: 5950X, 5900X, 5800X and 5600X Tested

www.anandtech.com

But the ADL scored higher....

And that's the 12900K, which is pushed far above its optimal range. We know all along that mobile ADL is where it will shine more because of it's "big.Medium" (according to you) architecture and the fact that Intel won't push it as hard where efficiency matters more.

 

[senttoschool]

Everyone else responded so I’ll let their answers stand.

So on desktop, where efficiency matters less, and where people generally buy Intel K processors for gaming, ADL is more efficient than Zen 3 for its main purpose.

And on laptops, where efficiency matters the most, ADL is more efficient than Zen3.

Why does anything else matter? @leman @Andropov @Krevnik

When are we going to stop making excuses for AMD and actually give ADL its due as the king of x86 consumer? Look, I get it. Apple Silicon has conditioned everyone here to hate on Intel. And the fact that Intel was stuck on 14nm for so long made it worse. But ADL is clearly superior to Zen3 in just about everywhere. Let's move on.