Apple is an existential threat to the PC

[Entropy]

Hmm. The files you refer to have these entries:

{
"id": 15,
"name": "Processor Frequency",
"value": "2.40 GHz",
"ivalue": 2400000000,
"fvalue": 2400000000.0
}

and

{
"id": 15,
"name": "Processor Frequency",
"value": "2.49 GHz",
"ivalue": 2490000000,
"fvalue": 2489999872.0
}

respectively. I don’t have access yet to both so I can’t compare physically. (And when I do, I’d feel constrained by the NDA, however utterly pointless in this case.)

Regardless, Rosetta 2 seems to either do extremely well, or even extremer well.
Unless they are running measurement tools such as GB5, I don’t see that users will be able to tell from the seat of their pants whether the application code is native or not. Most apps spend considerable time in system calls (which are native), or constrained by some other bottleneck such as main memory latency or bandwidth, or I/O. GB5 differences are exaggerated compared to actual use. Which is actually quite remarkable.

 

[Arwin]

I think this long expected development - didn’t we have a discussion here about ARM based PCs, I think partly because of Microsoft’s attempts with Windows 10 S - very interesting for sure. I am however very curious about if Mac will finally start supporting touch on their laptops. The touch supporting Windows 10 laptop we bought - what, more than 7 years ago? - is still used by my son and he loves it and uses touch quite a lot.

Without it, being able to run iphone and iPad apps on your Mac really isn’t going to be nearly as cool or useful.

 

[Vitaly Vidmirov]

Hmm. The files you refer to have these entries:

These values are shown in system info and they are wrong. System info in GB is typically broken.
Many consider the whole GB as totally broken benchmark, but I'm optimistic.
The point is that CPUs works at exactly the same measured frequency.

I am however very curious about if Mac will finally start supporting touch on their laptops..

My Windows (on ARM) laptop has touch screen, but I never use it. I hate fingerprints all over the screen.

 

[Entropy]

These values are shown in system info and they are wrong. System info in GB is typically broken.
Many consider the whole GB as totally broken benchmark, but I'm optimistic.
The point is that CPUs works at exactly the same measured frequency.

Though I can’t confirm it myself yet, I think you’re right, and the chip in the dev kit actually works at the same frequency as in the iPad. Apart from the provided data, clocking it lower also doesn’t make much sense.

 

[Entropy]

The "Apple is an existential threat to the PC" is a bit of a misnomer of a thread title at this point. I'll adress it he
1. Apples iPhone already did the job. Mobile devices have taken over the personal computing landscape by over a factor 5:1. Personal Computers are now handheld by a vast margin.
Of course, what is commonly known as a PC is normally a product running Microsoft Windows on an x86 processor, and if so, Apples move will just remove a bit of revenue from one of the chip suppliers. That's not a threat, that's just a small budgetary adjustment for the bean counters.
So, what could "existential" mean, if not Apple acting as an example or catalyst to change the current business model of Windows PCs? Microsoft and Intel have been riding that gravy train for decades, and as their financials show, it still works splendidly for them. The problem is growth, for both companies, so ideally they want to preserve the current locked in situation, while opening up for future growth.
2. Microsoft could get its act together and create either an effective ARM version of Windows, or an ISA agnostic one. They have tried this both with .NET and with more recently windows for ARM. There is a question as to what extent and by what mechanic they would preserve backwards compatibility. This would be a blow to x86 profit margins and/or consumer space market share, but would give flexibility for Microsoft.
3. Microsoft could start selling their own devices. The traditional model has been Intel (AMD) supplying the CPU, Microsoft the OS, and then leaving the field open to third parties to provide supplementary capabilities and final assembly. Intel has gobbled up much of the supplementary chip functionality, andMicrosoft have shown themselves willing to change the paradigm with their Surface product line being the only one with broader commercial success, albeit very small volume compared to the overall market. As I mentioned earlier, they could probably take over 90% of the desktop market (roughly a third of the x86 PC market) by adding ports to the Xbox sx and (more significantly) the rumoured Lockhart devices. This would bypass the third parties, but would ensure full backwards compatibility and thus protect lock-in. They could do something similar in laptop space, but it would require a wider model line up. This is an approach where one regards a Windows PC as a computing whole, rather than an assemblage of building blocks. It would favour Microsoft, most customers, and possibly AMD. Intel would see its profits from client systems drop.

Personally I don't believe either 2 or 3 will happen. Inertia and protecting current profit margins will be too strong for these companies to rock the boat, and only Microsoft can actually make it happen.

 

[pcchen]

I think for 2 to happen, Microsoft needs a big incentive for 3rd application vendors to port their products. Apple can do that because Apple can just say "hey we're going to stop making x86 based Macs, so if you still want to sell software on Mac, you'll need to port it over to ARM." Microsoft does not have this luxury, because they don't really control the hardware.
Microsoft did try to market ARM based Windows by trying to have some vendors selling cheaper ARM based hardware, but they really aren't much cheaper than x86 based ones. You can already buy cheap x86 based laptop from big brands (e.g. an ASUS Celeron laptop can be as cheap as US$250, an ARM based version is not going to be cheaper enough to make a difference).

 

[Entropy]

I think for 2 to happen, Microsoft needs a big incentive for 3rd application vendors to port their products. Apple can do that because Apple can just say "hey we're going to stop making x86 based Macs, so if you still want to sell software on Mac, you'll need to port it over to ARM." Microsoft does not have this luxury, because they don't really control the hardware.

They choose not to, which has benefits and drawbacks.

Microsoft did try to market ARM based Windows by trying to have some vendors selling cheaper ARM based hardware, but they really aren't much cheaper than x86 based ones. You can already buy cheap x86 based laptop from big brands (e.g. an ASUS Celeron laptop can be as cheap as US$250, an ARM based version is not going to be cheaper enough to make a difference).

Yep, the perceived customer benefit didn’t outweigh the limitations in terms of x86 backwards compatibility. Microsoft would have to work on both those aspects. They could be a lot more emphatic with both providing backwards compatibility and with demonstrating software benefits. The hardware angle is trickier, they could either commission their own silicon, or encourage Mediatek, Qualcomm et al to produce their own alternatives for third parties to choose from.
But if there aren’t meaningful benefits to be had for consumers to drive migration, it will still fall flat. Windows 7 was released in 2009, and it has taken a good long time to move customers even to a new OS version. Tangible benefits, or consumers won’t bite. And hardware wise, I don’t think that can be driven by the third party suppliers.

Also, at the end of the day, it has to pay off for Microsoft. That can concievably happen by making their own hardware and by providing subscription services for their platform. Making Windows toasters, basically.

We don’t even know if Apple will be successful in this transition. I think they can be, if they play their cards right and really take advantage of the hardware possibilities their move opens up. If so, it’s a given that Microsoft would like to emulate that success, but it is not a given that they can do so and still keep shareholders happy.

 

[MfA]

ARM was and is a distraction for Microsoft, one of many distractions Microsoft has saddled itself with.

 

[BRiT]

I wouldn't say distraction, I would say hedged bet.

 

[MfA]

3. Microsoft could start selling their own devices. The traditional model has been Intel (AMD) supplying the CPU, Microsoft the OS, and then leaving the field open to third parties to provide supplementary capabilities and final assembly.

Or just use the Chromebook method of versioned hardware platforms anyone can build (while still allowing custom PCs as well).

 

[Gubbi]

Bold move by Apple.

Intel has fallen so far behind in power efficiency that they no longer have a viable product for high end mobile devices of any sort (phones, tablets, laptops). It doesn't look much better for Intel for the next 12 months; So Apple go either go with AMD with twice the power efficiency and similar performance or their own with 2-3 times the power efficiency and similar performance.

They will lack a power be damned high core count SOC for highend iMacs and Mac Pros. Will they postpone the transition for these products, orphan them or develop high power, high core count SKU versions of their SOCs ?

Cheers

 

[Entropy]

Bold move by Apple.

Intel has fallen so far behind in power efficiency that they no longer have a viable product for high end mobile devices of any sort (phones, tablets, laptops). It doesn't look much better for Intel for the next 12 months; So Apple go either go with AMD with twice the power efficiency and similar performance or their own with 2-3 times the power efficiency and similar performance.

They will lack a power be damned high core count SOC for highend iMacs and Mac Pros. Will they postpone the transition for these products, orphan them or develop high power, high core count SKU versions of their SOCs ?

Cheers

On the previous page, I linked to video timestamps where both Johny Srouji and Sri Santhanam stated that Apple will produce dedicated SoCs for their different enclosure thermal dissipation capabilities. The rumoured 8+4 SoC looks like a pretty straightforward extension of the 2018 iPad SoC on 5nm. Keeping the 128-bit bus and swapping memory to LPDDR5, this SoC should be able to stay within tablet power draw levels and be suitable for the most compact laptops. So what would higher performance SoCs look like? This is where ones tech geek wet dreams can be given free rein.
However, I don’t believe we will see any power be damned SoCs. Even if it was just a general presentation slide, the graphic where they showed where they wanted to be indicated much higher performane, but lower than desktop power power draw which today is around 200W for optimum performance for the CPU alone + roughly similar for the GPU. I’d assume, and it’s only an assumption, that they will want to stay below 100W for SoC + memory even for their desktop systems, for comfortable and quiet cooling.
Looking at the rumoured 8+4 SoC as the 10W baseline, that still gives a lot of room. (I’d say it necessitates better than 256-bit LPDDR5 main memory systems, even though that would be a huge improvement in itself compared to current desktops.) It makes for tantalising speculation fodder indeed, because why go to the expense of fabbing dedicated SoCs, if not to provide capabilities that would otherwise be unattainable? They wording of the people in charge indicate at the very least three tiers of Mac SoCs. Speculate away.

 

[glow]

They really only have to exceed Intel's 28 core Skylake-based CPUs on their Mac Pros, if they were going for that. It's not an easy task, but it's also not a 64core Ryzen chip!

 

[Entropy]

While I prefer to speculate within tighter constraints, Apple does make and sell for instance the AfterBurner for their Mac Pros targeting high end film production. God knows what niches they might target.

 

[Gubbi]

They really only have to exceed Intel's 28 core Skylake-based CPUs on their Mac Pros

Apple do have some big clients for their iMac Pro series. I know Cisco lease a bunch of the 18 core models which are automatically replaced every two years. While the number of units sold/leased this way is a lot less than their laptop units, the margins on these units are high.

Cheers

 

[MfA]

If they don't give a shit about backwards compatibility, NUMA doesn't need much processor support and will give them all the power they want even without a reticle buster processor. Media processing is relatively parallel with relatively low communication bandwidth required.

It's legacy trash code which needs SMP ... in general legacy trash code is a giant anchor holding processor innovation back.

 

[Entropy]

If they don't give a shit about backwards compatibility, NUMA doesn't need much processor support and will give them all the power they want even without a reticle buster processor. Media processing is relatively parallel with relatively low communication bandwidth required.

It's legacy trash code which needs SMP ... in general legacy trash code is a giant anchor holding processor innovation back.

NUMA isn't likely to be in the cards, given how they stressed "unified memory" as a feature. Which it isn't, obviously, if it means feeding all your processing elements, GPU included, through the same itty bitty straw the CPU cores are already suckling from. There is space for major improvements there compared to what their supplier provided, without being very exotic.

Day late edit: And of course they care a lot about backwards compatibility. Everyone outside a computer science department and other select parts of academia need to, because there is value in work spent, and cost in work needing to be done. The balance to be struck concerns how much the old constrains your forward movement.
Apple chooses to (try to, but it’s bound to have effect) migrate developers, and development effort spent, on iOS. Arguably that is a larger and more current value than the unsupported cruft that will run under Rosetta 2.

 

[Arnold Beckenbauer]

https://www.trendforce.com/presscenter/news/20200707-10377.html

Apple to Start Mass Producing Self-Designed Mac SoC, Projected to Cost under US$100, in 1H21, Says TrendForce

"Although Apple still needs TSMC to manufacture its self-designed processors, the production cost of a Mac processor made with TSMC’s 5nm node is currently estimated under US$100, which is considerably more cost-effective compared to the 10nm Intel Core i3 processors, priced around US$200 to US$300 on the market"

 

[Malo]

They're comparing the production cost of an internal design compared to the market value of an Intel CPU? I'm sure Apple buys all their CPUs from Newegg...

 

[MfA]

Day late edit: And of course they care a lot about backwards compatibility. Everyone outside a computer science department and other select parts of academia need to, because there is value in work spent, and cost in work needing to be done. The balance to be struck concerns how much the old constrains your forward movement.

For most legacy applications just running on a single processor will be fine. Only the stuff which would actually need to be run simultaneously on more cores than in a single processor would need to be re-engineered if they went for multiple processors to scale for the highest end.