Apple is an existential threat to the PC

Someone on Twitter claimed that Apple Silicon actually has a special "x86 memory model" mode for x86 emulation.


It's not unexpected, as x86's stronger memory model is a big problem when emulating x86 from an ISA with a weaker model (practically every RISC ISA out there has a weaker model).
Some leakers changed from A14X for the first Apple Silicon based Macs to A14X based Apple Silicon,

MacBook (July 30th):
"MacBook A14X"

between:
"sizes Apple Silicon A14X > A14X(for iPad)"

MacBook Pro (today):
"Apple Silicon (A14X based) (Same as MacBook) or Intel chips (optional)"

Mod edit (Pete): Formatting for readability.
 
Last edited by a moderator:
They have been planning this for years.

I doubt Apple would waste the opportunity to make their own Apple Silicon machines stand out compared to their Intel counterparts designwise.
 
It's not unexpected, as x86's stronger memory model is a big problem when emulating x86 from an ISA with a weaker model (practically every RISC ISA out there has a weaker model).


I would expect Apple to just leave their CPUs in strong memory ordering mode. The weak memory ordering model is a win in low end cores because you don't need deep store buffers to maintain visible order of stores.

Apple don't do low end cores.

High end cores with advanced memory reordering (speculative load under stores with unknown addresss) already have the apparatus in place to ensure store ordering. There is zero performance benefit from the weaker model.

Cheers
 
I would expect Apple to just leave their CPUs in strong memory ordering mode. The weak memory ordering model is a win in low end cores because you don't need deep store buffers to maintain visible order of stores.

Apple don't do low end cores.

High end cores with advanced memory reordering (speculative load under stores with unknown addresss) already have the apparatus in place to ensure store ordering. There is zero performance benefit from the weaker model.

Cheers

Yeah, I agree it'd be better if everyone just uses x86's strong memory ordering model.
However, it'd be another chicken-and-egg problem, as most ARM binaries do have those memory barrier instructions in place (e.g. LDAR). At least in ARM LDAR instructions have less flexible addressing modes than normal LDR instructions. However, if the CPU is able to treat all LDR as LDAR without incurring much cost, compilers will be able to produce binaries without those instructions. Furthermore, Apple do have the "bitcode" system allowing for some level of recompiling after release. Apple also does not have to worry about interoperability between their ARM CPU and other vendors' ARM CPU (Unless, of course, if Apple is considering the ability for ARM Macs to install Windows for ARM... but that's probably on a very low priority, and even with that it should be fine, as Windows for ARM binaries are compiled against a weaker memory model ARM, which should run fine on a strong memory model CPU).
 
Perception of speed depends on many factors beyond pure performance of CPU :)

For CPU performance, AnandTech runs SPECCPU 2006 on latest smartphones.

On that benchmark A13 is about 50% faster than S865 for integer, and about 30% faster on floating-point.

I thought to touch up on this abit late, but it's the end result that matters. In benchmarks the A13 iphone is much faster, but in the real world the iphone 11 loses it mostly against the Samsung S20 (S865). Guess that benchmarks don't say everything.
 
I thought to touch up on this abit late, but it's the end result that matters. In benchmarks the A13 iphone is much faster, but in the real world the iphone 11 loses it mostly against the Samsung S20 (S865). Guess that benchmarks don't say everything.
On what does it lose? S20 might feel snappier due to better software and a 120 Hz screen. I certainly agree with you that it's what matters in the end, but that doesn't prove the CPU of S20 is more powerful than the one in the A11 (and no it isn't, no matter whether I'd like an ARM CPU to be faster than an Apple CPU :D).
 
On what does it lose? S20 might feel snappier due to better software and a 120 Hz screen. I certainly agree with you that it's what matters in the end, but that doesn't prove the CPU of S20 is more powerful than the one in the A11 (and no it isn't, no matter whether I'd like an ARM CPU to be faster than an Apple CPU :D).

In general apps/games opening. I would say they perform about the same, I also think Apple has better software.
A13 being 50% faster in benches doesnt really reflect in real world performance. Oh and apple A apus are arm based?

https://www.androidauthority.com/snapdragon-865-beats-apple-a13-1066243/
 
In general apps/games opening. I would say they perform about the same, I also think Apple has better software.
A13 being 50% faster in benches doesnt really reflect in real world performance. Oh and apple A apus are arm based?

https://www.androidauthority.com/snapdragon-865-beats-apple-a13-1066243/
Great, so the guy doesn't like normal benchmarks results and cooks up his own, which probably tests the storage system more than the CPU (or the larger RAM capacity)

Apple is vertically integrated and they use this to maximum effect. Because of the very low number of different configuration of IOS devices, they can optimize the shit out of software components where it matters the most like DOM handling or the Nitro JS engine in WebView/Safari for example.

That's on top of having the best general purpose CPU performance (by a mile) and very good accelerator performance. Yeah sure you can find benchmarks where Android devices win thanks to more RAM, slightly better storage, faster refresh screen or faster special purpose accelerators, but on average (or geometric mean), Apple is faster.

Cheers
 
Last edited:
Apple uses their vertical integration to great effect. Be

Great, so the guy doesn't like normal benchmarks results and cooks up his own, which probably tests the storage system more than the CPU (or the larger RAM capacity)

Apple is vertically integrated and they use this to maximum effect. Because of the very low number of different configuration of IOS devices, they can optimize the shit out of software components where it matters the most like DOM handling or the Nitro JS engine in WebView/Safari for example.

That's on top of having the best general purpose CPU performance (by a mile) and very good accelerator performance. Yeah sure you can find benchmarks where Android devices win thanks to more RAM, slightly better storage, faster refresh screen or faster special purpose accelerators, but on average (or geometric mean), Apple is faster.

Cheers

I mean in day to day usage, using both the S20 Ultra and 11 pro max, i don't notice much of that 50% faster CPU/GPU. Apps and games start about as fast, web pages load about as fast. I think deciding between those devices, one should look at what OS one prefers. Because in performance you won't notice the difference (despite benchmarks where apple wins by huge amounts).

Even gaming performance their as close as it can be. I like both, 11 Pro will probably last much longer thanks to 5 year OS updates versus 2 or 3. If that's a fair comparison.

Btw, shouldn't Apples storage be much faster, considering it's NVME vs samsungs UFS? Whatever i test, Apple always wins in loading/transferring, even iphone 5s vs Z1 compact (competitors of the time).

Im kinda new to all this mobile devices tech, just newly started comparing them (half a year). It's fun in-between high end pcs and consoles ;)
 
https://www.digitimes.com.tw/tech/dt/n/shwnws.asp?CnlID=1&cat=40&id=0000593298_RMP1WNG0LQ496S7M250T4

According to this digitimes report, initial production capacity for Macbook/ipad chips will be 5-6k WPM in Q4. Unsure how high (or low) that is for the market.
Well, it depends on where Apple chooses to deploy these chips. Their iPad business is quite substantial compared to their mac business.
That said, lets do some back-of-the-envelope calculations. Assume that, in shifting to 5nm, they go wider. A12x is roughly 130mm2 on the original 7nm process. Further, lets speculate that they are ambitious and let the die size increase a bit on 5nm to 12x12mm. We can then use a DiesPerWafer (300mm) calculator which tells us that, ignoring yield, we should get some 400 dies per wafer, or some two million+ per month. Now yield is not going to be 100% obviously, but according to TSMC the defect rate of 5nm is already quite good, and we don’t know to which extent Apple will design in redundancy, and use binning. Lets just go for 75% useable dies per wafer and we end up in the ball park of just over 1.5 million dies per month or roughly 20 million yearly. Apple sold 50 million iPads during 2019 (according to some, another number from other analysts is 40 million), with an unknown but substantial portion being iPad Pros. So yes, the numbers pan out for a chip aimed at iPad Pros and relatively high volume small laptops. They can make small adjustments depending on market success.

Just under 150mm2 on a low power 5nm process is a lot of gates.

Another small thing I noted from the TSMC tech symposium recently that may or may not be related to Apple was this slide.
Advanced%20Packaging%20Technology%20Leadership.mkv_snapshot_17.38_%5B2020.08.25_14.14.30%5D.jpg


What caught my eye was the small reticles with HBM2E, that has a strangely well defined future production date, and the HBM couples to what TSMC chooses to label a "SoC" as opposed to an "ASIC". Of course, this may have nothing to do with Apple. But if it does, we might see some pretty performant MacBook Pros and/or iMacs some months later.
 
Apple has announced a Special Event for Tuesday, November 10, 2020, which is universally speculated to introduce the first ARM Macs (not counting the devkit).

It seems like we could see an ARM 13" MacBook Pro and an ARM 13" MacBook Air, given a rumor from Ming-Chi Kuo from the middle of this year and a tweet from @l0vetodream earlier today.
Ming-Chi Kuo said:
We predict that Apple will launch new MacBook models including the new 13.3-inch ‌MacBook Pro‌ equipped with the ‌Apple Silicon‌ in 4Q20, the new ‌MacBook Air‌ equipped with the ‌Apple Silicon‌ in 4Q20 or 1Q21, and new 14- and 16-inch ‌MacBook Pro‌ models equipped with the ‌Apple Silicon‌ and all-new form factor design in late 2Q21 or 3Q21.
l0vetodream said:
13 inch X 2

Update: According to Mark Gurman and Debby Wu (Bloomberg), the first three ARM Macs will be a 13" MacBook Pro, a 13" MacBook Air, and a 16" MacBook Pro.
Mark Gurman and Debby Wu said:
The smaller models are further ahead in production and at least those two laptops will be shown at next week’s event. Beyond the processor switch, the devices won’t have significant design changes.
Mark Gurman and Debby Wu said:
The first Mac processors from Apple will be based on the A14 chip found in the latest iPhones and iPad Air, and tests inside Apple indicate improved power efficiency over the Intel parts they are replacing.
This rumor also mentions a few other upcoming ARM Macs including a smaller Mac Pro which was (AFAIK) not previously rumored.
Mark Gurman and Debby Wu said:
The company is already at work on a redesigned iMac […]

Apple engineers are currently developing a new Mac Pro that looks like the current design at about half the size. It’s unclear if that Mac will replace the current Mac Pro or if it’s an additional model. Apple’s chip designs could help the company reduce the size of its computers due to increased power efficiency, but the current Mac Pro is large, in part, to fit components like additional storage drives and graphics chips.
 
Last edited:
Back
Top