Intel Alder Lake vs. Apple M1

[playtech1]

Alder Lake looks pretty impressive and I am pleased Intel is back in the game as the more competition the better in this marketplace.

But I don't believe I have seen anything that yet convinces me that Apple has chosen poorly by going the M1 route. Outside of it losing in some of the artificial drag races of CPU benchmarks, M1 is a very good performing general purpose CPU, that can also be amazingly quick when you leverage its specialised hardware. It also does it in a tight power budget leading to cool/quiet running, long battery life and high performance on battery.

So far I am not yet seeing anything on the Intel or AMD side in terms of real-world laptops that offers all of that in one package.

I think as well we need to bear in mind that there is a lag between Intel and AMD announcing a CPU and getting it released in the key chassis (Dell / HP / etc.), whereas Apple launches the laptop with the CPU inside rather than the CPU first. If Apple delivers new laptops in early spring, then M2 is probably the fairer comparator than the 15 month old M1.

Intel and AMD's biggest problem IMHO is that there is no Apple equivalent in the laptop space producing well-rounded high end hardware at scale. Windows is also a huge drag (as well as being the biggest positive!), since it lacks a lot of the refinements and deep integration that Mac OS has.

 

[JouniS]

You cannot say that if a computer cost more than $5000 then the CPU isn't a consumer build then turn around and say spending $12,000 on RAM is a consumer build because the CPU is an i-series. Your definitions are all over the place and incredibly inconsistent.

We are not talking about expensive hardware here. The iMac is a consumer device and the i9 is a consumer chip, and both support 128 GB RAM. That 128 GB costs ~$600, which is about the same as a high-end consumer CPU, or what a high-end consumer GPU would cost without shortages. Consumer chips don't support enough RAM that it would be a significant expense, because a higher RAM limit is one of the features that sell more expensive server/workstation chips.

Further it's important to note that the i-series vs xeon divide is not likely to translate to Apple silicon. So I'm just not really sure where you are going with your definitions. What is a consumer level chip to you? What is not a consumer level chip and how does that apply to the as of yet unreleased higher core count M1 chips?

Consumer chips are those that are typically used in consumer devices, such as Intel Core, AMD Ryzen, and Apple M1/Pro/Max. Their core counts are relatively low, they prioritize single-threaded performance, and multi-CPU configurations are usually not supported. They also tend to have low RAM limits and relatively little external bandwidth, and most don't support ECC memory. The exact numbers change from year to year.

If a high-end chip doesn't have most of these properties, it's probably a server chip or a workstation chip.

What are you including in this $6K machine? You can get a 10-core MacBook Pro for $2200. It's extremely likely that you'll be able to get a 10-core Mac Mini for less.

As I said, I expect that the entry-level Mac Pro comes with an M1 Max or equivalent.

The current $6k entry-level model has roughly the same performance as a $3500 iMac, and the iMac includes a display. You pay the price for reliability, expandability, and upgradeability, and for a case designed for $10k or $20k workstations rather than for performance.

Yeah ... that's my argument to you which nullifies your argument that they can choose to sell 4 M1s for every M1 Max and therefore any SOC with more than that will be priced to the moon. You're repeating my argument that Apple plans its production to meet the demand that it thinks it is going to get all the way up the product stack. Yes they could choose to *make* 4 M1 dies for every M1 Max die, but that's not what they would be able to *sell*. Apple has a lot of data about what its customers want and how much of each product tier to make and what to price things at in order to move them.

Demand is a curve, not a fixed quantity. Apple chooses the demand by setting the prices. If they increase the prices of M1 Max devices, the demand for them goes down and the demand for M1 devices goes up, and the other way around. They have to commit to the total chip volume long in advance (and hope for the best). Within that commitment, they choose the ratio of M1 chips to M1 Max chips, and they set the prices as high as they can without lowering the demand below production numbers. The ultimate aim is to make as much profit as possible from that chip volume, which involves choosing how many low-end and high-end devices they should sell.

Yes and Apple has a 10 core consumer level chip, I think even by your definition. Why would a higher core count Apple made chip that is more than 10 cores not be consumer grade?

The exact core count is a moving target. If Apple chooses to make a bit bigger SoC with 16 CPU cores, Intel and AMD can respond by making a bit bigger CPUs with 24 cores. The economic argument stays the same: you can always scale the single-purpose chip a bit further.

Again, it should be noted that @EntropyQ3 confirmed that industry watchers have concluded that Apple is paying significantly less now that they are manufacturing their own chips, not more as you seem to be insistent that they must be.

Apple has already committed to the total chip volume. It doesn't matter whether they got those chips for free or paid ridiculous amounts of money for them. If they want to maximize the profits, they have to make the same choices in both cases.

 

[MayaUser]

But I don't believe I have seen anything that yet convinces me that Apple has chosen poorly by going the M1 route

They just lost some app support, thats all.

 

[Jorbanead]

Alder Lake looks pretty impressive and I am pleased Intel is back in the game as the more competition the better in this marketplace.

But I don't believe I have seen anything that yet convinces me that Apple has chosen poorly by going the M1 route. Outside of it losing in some of the artificial drag races of CPU benchmarks, M1 is a very good performing general purpose CPU, that can also be amazingly quick when you leverage its specialised hardware. It also does it in a tight power budget leading to cool/quiet running, long battery life and high performance on battery.

So far I am not yet seeing anything on the Intel or AMD side in terms of real-world laptops that offers all of that in one package.

I think as well we need to bear in mind that there is a lag between Intel and AMD announcing a CPU and getting it released in the key chassis (Dell / HP / etc.), whereas Apple launches the laptop with the CPU inside rather than the CPU first. If Apple delivers new laptops in early spring, then M2 is probably the fairer comparator than the 15 month old M1.

Intel and AMD's biggest problem IMHO is that there is no Apple equivalent in the laptop space producing well-rounded high end hardware at scale. Windows is also a huge drag (as well as being the biggest positive!), since it lacks a lot of the refinements and deep integration that Mac OS has.

They made the right choice because more than anything, Apple cares about performance per watt, and their chips seem to be the most efficient in this area. With my limited knowledge, it seems this is an important consideration when looking into the future of chip design and product design.

The other important aspect for me is how much Apple is prioritizing non-CPU/GPU tasks too. As a video/audio professional, having dedicated encoders/decoders and accelerators is another huge bonus. Plus I think features like their NPU are going to become more and more vital going forward. With both Logic and FC I’ve seen some impressive utilizations of the NPU.

 

[mi7chy]

AMD is idle power consumption champ. Explains why it rarely needs to be charged. Alder Lake may have taken over AMD desktop below 5900x and mobile Ryzen non-U but upcoming 6000U may still be efficiency champ. CPU is just a coprocessor to dGPU anyway so U-series CPU with 50-60W 5nm RTX4060 8GB would be ideal for progressives. Heck, if Nvidia acquisition of ARM goes through just embed an ARM SoC inside dGPU plus some ports then you'll have a complete system for GPU compute workloads.

AMD 4650U

 

[deconstruct60]

As I said, I expect that the entry-level Mac Pro comes with an M1 Max or equivalent.

The current $6k entry-level model has roughly the same performance as a $3500 iMac, and the iMac includes a display. You pay the price for reliability, expandability, and upgradeability, and for a case designed for $10k or $20k workstations rather than for performance.

The M1 Max only provisions x4 PCI-e v4 lanes. Either you are grossly changing the metrics of what a "Mac Pro" is or the M1 Max is woefully inadequate to fit the bill. Your are setting up metrics for the Mac Pro that are almost solely dependent upon CPU core count and performance. Apple is free to make up their own rules to assigned to the Mac Pro . But saying that x4 PCI-e v4 lanes was a "highest bandwidth and performance" solution would make them hypocrites from their position back in 2017. Even more so if they don't enable any 3rd party drivers.

The M1 Max could go into a Mini Pro ( along with a M1 Pro). Apple could use the current enclosure as a baseline and make some modifications.


It certainly will be that there will be a large set of people who 10 years would have bought a "Mac Pro" that can by a MBP 16" , mini pro , or iMac Pro ( with a M1 Max) and do just fine with their workloads into the mid-range future.

 

[crazy dave]

We are not talking about expensive hardware here. The iMac is a consumer device and the i9 is a consumer chip, and both support 128 GB RAM. That 128 GB costs ~$600, which is about the same as a high-end consumer CPU, or what a high-end consumer GPU would cost without shortages. Consumer chips don't support enough RAM that it would be a significant expense, because a higher RAM limit is one of the features that sell more expensive server/workstation chips.


Consumer chips are those that are typically used in consumer devices, such as Intel Core, AMD Ryzen, and Apple M1/Pro/Max. Their core counts are relatively low, they prioritize single-threaded performance, and multi-CPU configurations are usually not supported. They also tend to have low RAM limits and relatively little external bandwidth, and most don't support ECC memory. The exact numbers change from year to year.

If a high-end chip doesn't have most of these properties, it's probably a server chip or a workstation chip.


As I said, I expect that the entry-level Mac Pro comes with an M1 Max or equivalent.

The current $6k entry-level model has roughly the same performance as a $3500 iMac, and the iMac includes a display. You pay the price for reliability, expandability, and upgradeability, and for a case designed for $10k or $20k workstations rather than for performance.


Demand is a curve, not a fixed quantity. Apple chooses the demand by setting the prices. If they increase the prices of M1 Max devices, the demand for them goes down and the demand for M1 devices goes up, and the other way around. They have to commit to the total chip volume long in advance (and hope for the best). Within that commitment, they choose the ratio of M1 chips to M1 Max chips, and they set the prices as high as they can without lowering the demand below production numbers. The ultimate aim is to make as much profit as possible from that chip volume, which involves choosing how many low-end and high-end devices they should sell.


The exact core count is a moving target. If Apple chooses to make a bit bigger SoC with 16 CPU cores, Intel and AMD can respond by making a bit bigger CPUs with 24 cores. The economic argument stays the same: you can always scale the single-purpose chip a bit further.


Apple has already committed to the total chip volume. It doesn't matter whether they got those chips for free or paid ridiculous amounts of money for them. If they want to maximize the profits, they have to make the same choices in both cases.

It wasn’t $600 for 128Gb when it launched and isn’t today if you buy through Apple. It’s the cost of a new computer.

But regardless your distinctions are either arbitrary or will no longer apply with apple silicon. There’s no evidence of a Xeon-like split. Some of your arguments apply to xeons vs i-series and the original Mac Pro not apple silicon and the new Mac Pro mini which we haven’t seen. The M1 Max doesn’t become a magically new chip when if it were to go into a Mac Pro chassis and is probably destined for the mini. It just doesn’t have the ancillary features of a Xeon and no one would spend $6000 to get the exact same specced machine as a $2000 machine without the cavalcade of features provided by Xeons and Mac Pro that the Max chip (or its derivatives) and the Mac Pro mini don’t. This brings up another point: Apple knows exactly the characteristics of the machines they are putting these chips into. The reason there are so many variants, the reason Intel/AMD have to have so many product tiers is that they have no idea. Apple is only building and paying for the exact set of features it wants to put in. They don’t have to over-provision PCIe lanes or in general wasting silicon on features that they don’t plan to use.

Further I keep trying to tell you that your arguments about supply chain and pricing back up my position and not your original assertion that Apple has to price things according to not selling 8 base products for every 1 high end product. Also, of course the price Apple are paying now for their chips vs what they were paying affects prices and/or margins. That’s how capitalism works. If the cost to Apple to build a line of computers goes up or down regardless of who they paid their money to or when or how that either affects their margins or their prices.

Truthfully, feel free to respond but I think I’ll drop this on my end. We’ll find out when it launches.

 

[mi7chy]

Setting aside synthetic benchmarks, AMD may be king of ~27W AAA gaming.

Forward to 23:36

 

[JouniS]

The M1 Max only provisions x4 PCI-e v4 lanes. Either you are grossly changing the metrics of what a "Mac Pro" is or the M1 Max is woefully inadequate to fit the bill.

That's why I said "or equivalent". 4x M1 Max would also be inadequate for a Mac Pro for the same reason, as 16 PCIe lanes is a bit low even for a consumer chip. It seems reasonable to expect that the Pro chip will be based on the M1 Max, with some changes to modules such as PCIe controller and memory controller.

It wasn’t $600 for 128Gb when it launched and isn’t today if you buy through Apple. It’s the cost of a new computer.

I was $600 when it lauched (I bought one). If you care about $2000 price differences, you obviously don't buy RAM from Apple when there is an option to install it yourself.

Also, of course the price Apple are paying now for their chips vs what they were paying affects prices and/or margins. That’s how capitalism works.

It affects their margins, not their prices. The customer does not care about the prices Apple pays for components, and it does not affect the price they are willing to pay for Apple's products. Because Apple is not a charity, it always tries to extract maximal value from the customer.

If the price of components increases, Apple's profit margins suffer and it may cancel plans for some low-end products that would no longer be sufficiently profitable. If the price of components decreases, profit margins soar, because the customer is still willing to pay the same price.

Prices are based on supply and demand. Costs mostly just determine whether a product is commercially viable.

 

[Sydde]

Heck, if Nvidia acquisition of ARM goes through just embed an ARM SoC inside dGPU plus some ports then you'll have a complete system for GPU compute workloads.

Why does that matter? nVidia already has an architecture license. They could do that right now today and only have to pay whatever the royalty is. They only want to acquire ARM in order to screw vicegrips down on everyone else whilst leeching royalties off of them.

 

[mi7chy]

Take the baton from Arm Ltd/SoftBank to evolve ARM especially on the GPU side for modern GPU compute workloads instead of focusing too much on the past with CPU. And, make it the baseline vs one off vendor. FTC can allow the acquisition but with stipulations so they can't pull an Apple anti-consumer like, for example, removing bootcamp with AS.

 

[cmaier]

Take the baton from Arm Ltd/SoftBank to evolve ARM especially on the GPU side for modern GPU compute workloads instead of focusing too much on the past with CPU. And, make it the baseline vs one off vendor. FTC can allow the acquisition but with stipulations so they can't pull an Apple anti-consumer like, for example, removing bootcamp with AS.

That ship has sailed. nVidia isn’t buying Arm.

It’s pretty hilarious that you think Apple “pulled” something by “removing bootcamp with AS.”. What OS would you like running on bootcamp on apple silicon? Because Microsoft isn’t licensing one, and they’re the ones that get to decide.

 

[mi7chy]

https://distrowatch.com/

For Raspberry Pi ARM
https://www.fossmint.com/operating-systems-for-raspberry-pi/

 

[januarydrive7]

https://distrowatch.com/

For Raspberry Pi ARM
https://www.fossmint.com/operating-systems-for-raspberry-pi/

Apple has not blocked running unsigned kernels on Apple Silicon. What they did do, is remove direct support for loading Windows through Bootcamp, which never supported linux distros to begin with.

As @cmaier said, your suggestions are hilarious.

 

[mi7chy]

He's not Jim Keller though.

 

[cmaier]

He's not Jim Keller though.

If you mean i don’t change jobs every couple of years, you are correct.

 

[mi7chy]

Can even bare metal install Windows on Raspberry Pi. I guess one can stick his head in the ground and pretend it's not anti-consumer.

https://www.tomshardware.com/how-to/install-windows-11-raspberry-pi

 

[cmaier]

You can even bare metal install Windows on Raspberry Pi. I guess you can stick your head in the ground and pretend it's not anti-consumer.

https://www.tomshardware.com/how-to/install-windows-11-raspberry-pi

You can install it on Mac, too, but you’re violating Microsoft’s licensing terms. What’s your point?

I ran Windows 11 on an M1 Mac: Here’s my experience

We put Parallels’ Windows 11 virtual machine to the test on an M1 Mac Mini to see if it’s worth your money.

www.laptopmag.com

 

[mi7chy]

Jim Keller might say the difference is one is bare metal while the other requires Parallels and a subscription. Is subscription a positive now?

 

[cmaier]

Jim Keller might say the difference is one is bare metal while the other requires Parallels and a subscription. Is subscription a positive now?

Windows 11 requires a TPM - it’s Microsoft being consumer unfriendly. You expect apple to build in a piece of hardware that is useful only for running windows, an operating system that Microsoft will not license for running on macs? Hence parallels to get around it. There are other methods of course. Parallels is just the easiest.

What’s with your weird Keller fetish?

 

[mi7chy]

What's wrong with Jim Keller? Some like Batman and some like Robin.

 

[exoticSpice]

AMD is idle power consumption champ.

Nope. M1 Max uses 0.2 watts in idle and M1 Max only has 2 E cores. M1 and A15 should have better idle usage due to having 4 E cores.

https://www.anandtech.com/show/17024/apple-m1-max-performance-review/3 - source

https://distrowatch.com/

For Raspberry Pi ARM
https://www.fossmint.com/operating-systems-for-raspberry-pi/

You can load your own Kernels like the Pi on the M1 Macs. GPU drivers are being worked on so far.
This is running bare metal.

https://twitter.com/i/web/status/1458473546225577987

 

[exoticSpice]

Can even bare metal install Windows on Raspberry Pi. I guess one can stick his head in the ground and pretend it's not anti-consumer.

https://www.tomshardware.com/how-to/install-windows-11-raspberry-pi

You can RUN Linux kernels bare metal on M1 Macs. Its not blocked by Apple. The reason you cannot run Windows 11 ARM because Microsoft does not license it yet other than Qualcomm.

Installing Windows on ARM is not officially supported by MS on Pi.

 

[exoticSpice]

Jim Keller might say the difference is one is bare metal while the other requires Parallels and a subscription. Is subscription a positive now?

You can use VMware Fusion or UTMand that does not require a subscription and both are free.

 

[deconstruct60]

That's why I said "or equivalent".

An equivalent would be a SoC of about the same size. A M1 is not equivalent of a M1 Max.
A M2 Max would likely be equivalent of a M1 Max. In the M1 generation there isn't going to be something equivalent of a M1 Max. If it has a different name then it is different ( i.e., not equivalent).

Most likely the M1 Max would be modified to be used in a two (Duo) or four (Quattro ) set up. [ I've seen handwaving that that is it is all already there but none of it is highly creditable. Parts yes, all no. ). But those additional suffixes would be indicative that not really equivalent.

4x M1 Max would also be inadequate for a Mac Pro for the same reason, as 16 PCIe lanes is a bit low even for a consumer chip.

First, several of those are going to be out to do single/duo Ethernet , Wifi/BT , supplemental USB , and SATA that the current Mac Pro 2019 does.

Second, would be an aggregate of x16 lanes worth of bandwidth but there would be no single card worth of x16 to get to a card that was PCI-e v3 (or v2 ). So lots of mismatches.

Third, once get past 4 Thunderbolt ports it doesn't make much sense to allocate the silicon to that versus more general PCI-e v4 in a workstation context. Don't need 4 SSD controllers if going to just provision one and only one internal disk. Secure element.. 4 ... probably overkill.

Similar with the dual ProRes / Video en/decode complexes. Don't really need two per die if going to multiple dies. ( if one per die then two dies gets back to two and four dies would be 4. the latter is likely getting into diminishing returns zone. ). Apple could easily chuck those Max subcomponents at the bottom for interdie communication logic.

Once start to rack up changes to the logic implemented on the die the equivalency starts to fade.

It seems reasonable to expect that the Pro chip will be based on the M1 Max, with some changes to modules such as PCIe controller and memory controller.

Apple already has a "Pro" chip. The M1 Pro ( Jade-chop) is a chopped down version of the M1 Max (Jade). Perhaps the "Mac Pro" will be shifted over the "Mac Max" because Apple has superseded Pro as being top of the line suffix.

The M1 Pro is based on the M1 Max. ( it is a designed for alternative mask change ), but the M1 Pro is not an "equivalent" to the M1 Max.

But yes, Apple will try to engage in high amount of design reuse between M1 Max and these alternatives to keep the costs under control. That is a dual edge sword because will be grounding it to many of the same constraints of the laptop design parameters that went into the M1 Max. Doubtful the memory controllers would change much. Perhaps positioned differently on the die edges differently to promote more uniform multiple chip packaging layouts.

Die size though. Those are probably close. I don't think it is likely Apple wants to make an individual die that is much bigger than a Max.

It affects their margins, not their prices. The customer does not care about the prices Apple pays for components, and it does not affect the price they are willing to pay for Apple's products. Because Apple is not a charity, it always tries to extract maximal value from the customer.

If the price of components increases, Apple's profit margins suffer and it may cancel plans for some low-end products that would no longer be sufficiently profitable. If the price of components decreases, profit margins soar, because the customer is still willing to pay the same price.

Prices are based on supply and demand. Costs mostly just determine whether a product is commercially viable.

When Apple cranked the entry price of a Mac Pro up 100% ( MP 2013 entry -> MP 2019 entry) Apple paid a price in unit volume. You claim that the all of the Apple customers in a product segment are perfectly elastic in ability to pay doesn't match reality. That is why Apple has a global market share sub 10%. and a substantive number "Mac Pro" class user base still sitting on Mac Pro 2009-2010 systems. And why the MP 2019 has a "low volume " tax on it. Apple is keeping the margin, but they are loosing users where they go too far into the "the cult followers will just pay anything" mode.