A few bold Apple Silicon Mac Predictions

[adib]

More and more companies are moving to the cloud. Even developers are moving to cloud-based environments because of low cost, minimal maintenance. and higher performance especially in fields like Machine Learning.

Low cost, yes. Minimal maintenance, yes.

Higher performance? Not by far. Even a typical 2015 Intel i7 MacBook Pro is faster than a Heroku L. This is the development use case, where there is only one user (i.e. the developer), not about serving many concurrent users.

Machine Learning for development purposes would still be faster with a local machine. Quicker turnarounds and easier job submissions. Smaller companies can buy $10K desktop-server to share with 10 data scientists just fine at half the prices that AWS and Azure pays NVidia for their servers, since these smaller companies can buy gaming GPUs (instead of server ones that Cloud Companies would need to pay for). Larger companies are fine with sharing $100K NVidia servers with 16 GPUs for a development department.

 

[Falhófnir]

Consoles only represent a portion of gaming market. PC is still very popular for gaming. And you can't just develop your game assuming baseline hardware that only 10% of PC gamers own. Current generation of consoles represent high-end gaming (4K, 60/120fps, high quality settings) but games will target mid-range as well. And while I doubt that Macs will be a high-end gaming machines, every Apple Silicon Mac has the potential of being a mid-range one — which cannot be claimed for a regular PC laptop.

I think Apple can make Arcade a player in the gaming space, but there's still a lot of work for them to do. Presenting the Mac and iPad market as one unified platform would be a big shot in the arm, and getting the A14 with its much better graphics performance into the mini and base iPad ASAP would create a workable baseline of performance (at present only the iPad Pro and Air 4 have 'console level' graphics, and that's last gen consoles at that). The A12 falls way too far short, so it's going to require M1 macs and A14 iPads to proliferate and build a competent base over the next couple of years before Apple will be able to present an appealing market for studios to target with recent releases. So while M1 Macs absolutely present a new opportunity, the larger installed base of Intel Macs will hold Apple back until the Apple Silicon Macs become a dominant slice of the pie.

 

[JouniS]

Consoles only represent a portion of gaming market. PC is still very popular for gaming. And you can't just develop your game assuming baseline hardware that only 10% of PC gamers own. Current generation of consoles represent high-end gaming (4K, 60/120fps, high quality settings) but games will target mid-range as well.

PS4 and Xbox One were released in late 2013. Most 2014 games still supported the earlier console generation, but the support was generally dropped in ambitious 2015 games. Games designed for almost decade-old hardware looked rather underwhelming next to ones designed for current hardware, and they often didn't sell particularly well. And when the games were designed for new consoles, they were quite demanding for PCs.

 

[senttoschool]

Consoles only represent a portion of gaming market. PC is still very popular for gaming. And you can't just develop your game assuming baseline hardware that only 10% of PC gamers own. Current generation of consoles represent high-end gaming (4K, 60/120fps, high quality settings) but games will target mid-range as well. And while I doubt that Macs will be a high-end gaming machines, every Apple Silicon Mac has the potential of being a mid-range one — which cannot be claimed for a regular PC laptop.

The issue here is that @JouniS thinks that the majority of PC games will soon upgrade to match a XSX/PS5, which means going out to buy a $500 GPU and a $300 CPU.

This simply won't happen. Yes, many PC gamers will upgrade soon but it's a very slow process that will take many years. And as I wrote previously, Cyberpunk still targets a nearly 8-year-old GPU and an 8-year-old CPU.

PC games do not target the high end. They target the low-end and then scale up for high-end GPUs.

I fully expect the median Mac to be a more powerful gaming machine than the median gaming PC in 3 years.

I have read it.

I find the information contained within irrelevant to the long-term goals of Apple, which do not involve gaming.

Again, this is simply spec-racing and has no foundation in reality.

Apple is the largest gaming company in the world but Apple doesn't have long-term goals for gaming. Ok.

 

[LongWayHome]

Apple is no longer a brand just for the wealthy.

AirPods Max would like a word with you.

 

[leman]

PS4 and Xbox One were released in late 2013. Most 2014 games still supported the earlier console generation, but the support was generally dropped in ambitious 2015 games. Games designed for almost decade-old hardware looked rather underwhelming next to ones designed for current hardware, and they often didn't sell particularly well. And when the games were designed for new consoles, they were quite demanding for PCs.

The question is: did ported console games run bad on the PC because the consoles are just that fast or is it just because they were bad ports? This part really depends on the developer. Consoles are very different to program for — unified, high-latency memory, direct hardware access, known exact specs... if a developer is targeting a console — any only the console — from the start, then yes, they will probably have a hard time porting it to the PC. Funnily enough, Apple Silicon Macs are more similar to gaming consoles in their performance characteristics than Windows PCs are.

To give you a concrete example: I am developing a 2D side-scroller game as a hobby project that uses complex polygonal maps for the game world. On Apple Silicon Macs I can generate triangles on the runtime using the CPU and immediately render them on the GPU. The performance is amazing because of low-latency CPU/GPU communication. I can do some fairly cool stuff with it — fully destructible worlds, arbitrary zoom levels, complex physics — all while being buttery smooth. I just can't do the same thing on a regular gaming PC because it takes ages to transfer the data from the CPU to the GPU... so in the end I have a game that runs amazingly on an iPad but where I have to either tone things down or thing of some really complicated workarounds to make it run well on a much faster gaming laptop.

 

[senttoschool]

AirPods Max would like a word with you.

I didn't say Apple doesn't market to the rich. I just said they're not just marketing to the rich.

Apple products are now in the low-mid-end. iPhone SE, Watch SE, iPad, Homepod Mini.

And you can probably bet good money that Apple will release a cheaper Airpods Max in the future.

 

[JouniS]

The question is: did ported console games run bad on the PC because the consoles are just that fast or is it just because they were bad ports? This part really depends on the developer.

They were not ported console games but games developed simultaneously for all three major platforms.

The overall effect was a sudden jump in hardware requirements to match the capabilities of the new consoles. Memory requirements increased from 4 GB to 6-8 GB, and GPU requirements jumped 2-3 generations. While games used to run on almost anything, the minimum GPU requirement was suddenly a midrange card from 3 years ago.

 

[neinjohn]

I think you're missing the point.

The point isn't about pure power. The point is: how many AAA capable computers are available for developers to develop for? How many computers have a 1050Ti/1060 or better GPU are there? These two are the most common GPUs on Steam which means AAA developers have to develop for them.

Kuo predicts that Apple will sell 35 million Macs/year (20% market share) within 3 years. That's 35 million computers capable of playing AAA games at low to high settings. According to IDC, the number of gaming laptops and gaming desktops sold in 2020 was 37.1 million. That means in 3 years, 50% of all gaming computers sold will be Macs assuming they don't overlap.

Do you think AAA game developers will ignore 50% of the available market?

And if Macs take 40% of market share in 5 years, then the number of AAA capable Macs will be 70m/year sold versus 37m for PCs. That means 65% of all AAA capable gaming computers will Macs. But the reality is probably even worse for PCs because many people who buy gaming PCs will have switched over to Macs if Macs continue to increase market share.

Thus, Macs are poised to become the #1 platform for AAA games.

On my ML-tinted glasses they can even go a little further. I hope Apple has a good plataform to support image upscaling on the Neural Engine as it may give another boost for a very capable base hardware to render games at 720p, or lower, on medium details, and output native resolution. No need to go super high on detail/resolution that is hardly possible to see on a 13'' so just focus on the best compromise for stable 60 FPS.

Similar to the high base plataform on the Air, they could provide similar position with a higher end Mini, a non-4K iMac and the 4K iMac if nVidia RTX series don't lower their base price.

Again with my ML-tinted glasses they could focus on every creativity trickery deep learning trained and untrained model they can find to work with the Neural Engine to do more with less on M1 on further software updates until a point it has done the job of bringing more people on their services.

 

[ChrisA]

You're referring to Amazon's Graviton/Marvell's ThunderX2 right? Those are server chips.

Let me just be clear then, if you want a non-server, Apple will be the fastest.

And no, AWS Graviton2 ARM or Marvell's ThunderX2 chips aren't faster for daily tasks because its single-core speed is weak and no consumer software will utilize that many cores. Single-core is what's responsible for the "snappiness" that people feel when using their computers. ARM server chips would get destroyed by the M1 when it comes to "snappiness".

PS. This is a bit like saying Apple computers will never be the fastest because you can always buy a supercomputer if you have enough money.

I pointed to those server chips to show that Apple's M1 is not magic. Others can make fast chips and there is a huge market out there.

I doubt Apple will even try to make the fastest singe core Arm processor. They are more likely to go for best performance per Watt than for best absolute performance HP and Dell and others will offer a faster CPU but with 1.5-hour battery life using cheap 18650 cells.

As for consumer software not being able to take advantage of many cores. That will change. Today I think very hard about threads when I design new software. If there is some way to do the job in parallel I'll try. Eventually, it might get to the point where compilers do this for us at a low level

Be sure an watch Nvidia. They are buying Arm. I'm sure their goal is to make integrated CPU/GPU systems and maybe hundreds or even thousands of Arm cores in one computer. Maybe the main use of this is not for PCs but for self-drive cars and data-center hosted AI? But some of this will go into PCs.

One other thing is that the PC might be a dead-end with most computation moving into data centers. Already we see so many web-based apps. I do agree that for 99% of the people a Mac Air is about right, maybe a larger screen but if computation moves to the cloud, all you need in the PC is a big GPU mostly.

The others, HP, Del, and so on, will catch up and otter products on both sides of what Apple offers. Just like today. They make some pretty nice $600 notebook PCs and some workstations that are faster than any Mac Pro.

In any case, the next ten years should be far more interesting than the last ten years where we saw only very incremental innovation.

I expect that in 10 years the PC market will be very diverse with many products. Today they are al just closed of the 1980's IBM PC with small improvements each year. I'm hoping this 30-year trend blows up.

My next computer will maybe be this $700 Nvidia computer. It has 32GB RAM, 8 Arm cores, and 512 GPU cores.
Note that it is selling on Amazon and so is on the fringe of almost being almost "mainstream". https://www.amazon.com/NVIDIA-Jetson-Xavier-Developer-32GB/dp/B083ZL3X5B Most people would have zero use for this but this is what I meant by diverse market, many specialist products available and a more fragmented market. Not just IBM PC clones.

 

[jerryk]

Low cost, yes. Minimal maintenance, yes.

Higher performance? Not by far. Even a typical 2015 Intel i7 MacBook Pro is faster than a Heroku L. This is the development use case, where there is only one user (i.e. the developer), not about serving many concurrent users.

Machine Learning for development purposes would still be faster with a local machine. Quicker turnarounds and easier job submissions. Smaller companies can buy $10K desktop-server to share with 10 data scientists just fine at half the prices that AWS and Azure pays NVidia for their servers, since these smaller companies can buy gaming GPUs (instead of server ones that Cloud Companies would need to pay for). Larger companies are fine with sharing $100K NVidia servers with 16 GPUs for a development department.

I disagree that training would faster on a local machine. I have a desktop with i9900K CPU, 64 GB of memory, 4 TB of SSD, and 2 RTX 2070 GPUs. But when I want to train fast I deploy to a Google Cloud instance with 256 GB of memory and 16 TPUs. Once the model is trained I bring it down to my local system to try it.

 

[JMacHack]

Be sure an watch Nvidia. They are buying Arm. I'm sure their goal is to make integrated CPU/GPU systems and maybe hundreds or even thousands of Arm cores in one computer. Maybe the main use of this is not for PCs but for self-drive cars and data-center hosted AI? But some of this will go into PCs.

Literally stated in their announcment on buying ARM holdings: Uniting NVIDIA’s AI computing capabilities with the vast ecosystem of Arm’s CPU, we can advance computing from the cloud, smartphones, PCs, self-driving cars and robotics, to edge IoT, and expand AI computing to every corner of the globe.
(emphasis mine)
source:https://nvidianews.nvidia.com/news/...s-premier-computing-company-for-the-age-of-ai

One other thing is that the PC might be a dead-end with most computation moving into data centers.

Been hearing this for decades, still not true.

Already we see so many web-based apps.

Which are garbage, stuff like Google Docs only gained traction because they're even cheaper than buying MS Office for schools. Electron and web technology based apps also run like hot garbo.

I do agree that for 99% of the people a Mac Air is about right, maybe a larger screen but if computation moves to the cloud, all you need in the PC is a big GPU mostly.

Any big GPU will need to be fed by a powerful CPU. You can't get the full power of a high-end graphics card with a low-end processor.

My next computer will maybe be this $700 Nvidia computer. It has 32GB RAM, 8 Arm cores, and 512 GPU cores.
Note that it is selling on Amazon and so is on the fringe of almost being almost "mainstream". https://www.amazon.com/NVIDIA-Jetson-Xavier-Developer-32GB/dp/B083ZL3X5B Most people would have zero use for this but this is what I meant by diverse market, many specialist products available and a more fragmented market. Not just IBM PC clones.

I'm gonna assume you know that's an AI development SBC, not a traditional "PC".

 

[leman]

I doubt Apple will even try to make the fastest singe core Arm processor. They are more likely to go for best performance per Watt than for best absolute performance HP and Dell and others will offer a faster CPU but with 1.5-hour battery life using cheap 18650 cells.

Apple already has the fastest single core ARM processor. By a large margin. Those Server CPUs you keep mentioning are completely different devices, designed for parallel computation. Their single-core performance is abysmal. Apples single core performance on the other hand is up there with the fastest currently shipping CPUs in the world.

 

[leman]

My next computer will maybe be this $700 Nvidia computer. It has 32GB RAM, 8 Arm cores, and 512 GPU cores.
Note that it is selling on Amazon and so is on the fringe of almost being almost "mainstream". https://www.amazon.com/NVIDIA-Jetson-Xavier-Developer-32GB/dp/B083ZL3X5B Most people would have zero use for this but this is what I meant by diverse market, many specialist products available and a more fragmented market. Not just IBM PC clones.

Your next computer is going to be a specialized chip for developing autonomous robots? What are you using now, LEGO NXT?

 

[dmccloud]

My next computer will maybe be this $700 Nvidia computer. It has 32GB RAM, 8 Arm cores, and 512 GPU cores.
Note that it is selling on Amazon and so is on the fringe of almost being almost "mainstream". https://www.amazon.com/NVIDIA-Jetson-Xavier-Developer-32GB/dp/B083ZL3X5B Most people would have zero use for this but this is what I meant by diverse market, many specialist products available and a more fragmented market. Not just IBM PC clones.

Selling a product on Amazon does not make it almost mainstream. Furthermore, that system wouldn't run any commonly used software, so you'd essentially be buying an expensive paperweight with extremely limited functionality.

 

[Yebubbleman]

• Macs will take 50% of the laptop/desktop market within 6 years

Not super likely. You're forgetting to factor in Windows PCs in the enterprise and Chromebooks in education.

• Soon, if you want the fastest computer you have to buy a Mac

Apple has a decent lead over Intel, AMD, and other ARM vendors. That won't last forever. They'll have to make less efficient SoC's to compete (as is done in the smartphone space), but they'll eventually get there. We won't, however, see any company doing the kind of vertical integration Apple is doing for a LONG time, if ever.

• If you want a laptop that has the best battery life, you have to buy a Mac

That will be true for a while, but not forever and certainly not for longer than 4 years.

• Apple will release a Macbook SE priced at $700-$750. They couldn't before because cheap Intel chips were too slow.

Doubtful, unless by "Macbook SE" you mean "iPad Air".

• Many iPhone users (50% market share in U.S.), iPad users (65%), and Apple Watch (55%) users bought cheap Costco laptops. Apple can win these customers back with an affordable Macbook.

Apple would much rather get those customers on iPads. Frankly, given the markets being targeted, Apple's strategy isn't bad there.

• AAA gaming will come to Macs. Even the slowest Apple Silicon GPU is as fast as a 1050Ti. Soon, the median GPU in a Mac will be faster than the median gaming PC. Combined with a projected 50% market share, AAA game developers can't ignore Macs. Apple will take a cut of every AAA game sale because they will have to go through the App Store just like how they have to go through Steam on Windows.

Just because the M1's graphics capability is impressive doesn't mean that you're going to get AAA titles ported over. It's HARDER to port things over to Apple's GPUs than it is to merely port a Windows game to the Intel version of macOS (as at least GPUs are designed the same way on Intel Macs as they are on PCs). The SoCs in both tvOS-based versions of the Apple TV are more than capable of delivering "console class graphics", but tvOS hasn't, at all, taken off as a living room video gaming platform, despite it being open to developers to develop AAA games for it. You're going to see similar difficulties with Apple Silicon Macs as well.

• By increasing their laptop/desktop share, Apple will sell more Apple One+ subscriptions because they're adding one more major device to the customer's ecosystem.

Eh...Apple One is mostly beneficial on the other platforms. I don't see the Mac bringing Apple anywhere near as many new Apple One subscribers as I do the Apple TV, iPhones, and iPads.

Apple is no longer a brand just for the wealthy. Apple is destroying the low mid-end market with SE products that are simply better than the competition. A $400 iPhone SE is faster than any Android phone. It's quite likely that a $700 Macbook SE will be faster than any Windows laptop. Apple is gunning for marketshare. And then they want to sell subscriptions to their customers.

Apple won't sell a Mac notebook for $700. At best, they may keep an M1 model on sale after every other Mac is already past it. Like, I could totally see the M1 Air, M1 mini, or M1 2-port 13" MacBook Pro lingering around to cater to the low-end market after a better successor version is available, but that's the best you're going to get in terms of a MacBook SE.

Bonus bold predictions: Apple will enter the cloud hardware business and enterprise markets within 10 years. Bloomberg reports that Apple is making a 36 core SoC. Soon, Apple will realize that they can make a better server chip than AMD/Intel. Apple knows that most computing will be done in the cloud in the future. And because Macs will become a dominant laptop/desktop player, Apple will make a huge push into selling more Macs to enterprises and start competing directly with Microsoft there.

Disclosure: I own Apple shares and have bought more since M1 Macs came out.

You have ARM SoCs that are designed and optimized for the enterprise with server hardware, and, on the other end of the spectrum you have ARM SoCs that are designed and optimized for personal computing. Apple's SoCs are on the latter side of that spectrum. Apple could design their 36-core SoC to be somewhere in the middle of that spectrum, but they likely won't. They tried it with Xserve and failed because they didn't know what they were doing (and certainly, once Xserve went from PowerPC to Intel, where the number of rack mounted server options were far more plentiful, there was way less of a need for it).

I do, however believe that Apple SHOULD try to build the next Mac Pro (and Rack-Mounted variants) to be datacenter-capable such that one Mac Pro tower or rack-mounted box can host multiple Mac VDIs for lower-end client systems like MacBook Airs and Mac minis. This is something the Windows and Linux worlds have been able to enjoy for a while now. Certainly Apple could sell thin client variants of the Mac mini and MacBook Air (reducing the cost of both by at least half) that tie into Mac Pro systems used in this fashion (and/or hosted in the cloud). But, I'm not holding my breath. Apple tends to only care about end-user computing and not back-end infrastructure-y things like that.