techconc

cloudguy said:

OK. Say Intel goes belly up in 5 years.

1. Apple had 8% market share last quarter. Presume that they maintain it - it has dropped to as little as 5% in the past - or even increases to 10%. Good for Apple. Now, what operating system will the other 90% use?
2. Apple's Mx chips are only going to be used in Apple products. What CPUs will the other 90% use?

I don't think anyone is suggesting any majority marketshare will shift from PC OEMs to Apple.  The concern is that Apple has raised the bar that PC OEMs will likely not be able to match or remain competitive with based on Intel's chips.  The threat is if Microsoft / Qualcomm / ARM actually address the PC market with a chip similar to what Apple has started doing with the M1.  The cloud servers are itching to switch to ARM but they basically have to follow the PC market.  Software is developed on PCs and then hosted in the cloud.  However, there is a significant threat that this dynamic can change.  Honestly, they could even make a go of it by producing a SoC based on ARM's existing X1 cores that the upcoming SD888 will use.  Make a higher wattage variant with more cores and you have a legitimate competitor rather than the anemic ARM attempts to date for PC.

cloudguy said:
Intel's inability to get past 10nm has nothing to do with the limitations of their ISA.

You're right.  These are two distinct problems that Intel is facing.  One has nothing to do with the other, however, both are holding Intel back and responsible for Intel's current predicament. 

Apple doesn't manufacture their own chips either. So it isn't Apple that is ahead of Intel but TSMC. Were TSMC's foundries to start having the same issues that Intel is, Apple would be equally impacted. 

Your comment assumes the only reason Intel is behind is because of their lag in manufacturing process. That's not the case. Intel's chip designs were never the best. They were mediocre at best. They remained competitive by always being on the cutting edge manufacturing process. That is, their manufacturing process ability masked their deficiencies with their chip designs. They had a few notable exceptions such as their Core design, etc. but this is largely true throughout their history. Today Intel is faced with a chip design that is held back because of their legacy CISC ISA coupled with continued failures to improve their manufacturing process. Similarly, AMD can indeed benefit from TSMC's manufacturing process, but they also cannot escape the handicaps of the x64 ISA.

Designing a chip for any given process node isn't the challenge.  Anyone can do that.  Intel's problem is that even if they were to achieve process node parity from a manufacturing perspective, they still won't catch up on performance or efficiency.  Intel's process node failures have simply magnified their processor design deficiencies.  Their only hope of even achieving some level of parity is to somehow not only catchup, but massively leapfrog the competition in manufacturing.  Either that or dump their legacy ISA which they can't do.  Neither of those things are going to happen. 

canukstorm said:

They’ve had a true ARM version of Windows for two years now but so far it’s hardly gained any traction with users or developers. I can assure you that Apple has sold more M1 Macs in the last two months than the combined total of ARM-based Windows devices sold in the last two years. 

The difference is this... previously, if you wanted performance you have to stay with x86.   ARM chips available for Windows have been anemic and Microsoft's emulation layer was neither good at emulation nor did it perform well.  Likewise, ARM wasn't a real choice so far. 

What the M1 chip did is wake people up to the fact that you can in fact have the best of both worlds.   That being highest performance and great efficiency.  The bar has been raised.   Microsoft will have to respond if they wish to remain competitive with Apple's devices after their transition is complete.  Intel doing big / little cores and even moving to a better process node will help a bit, but it will not solve the underlying problem.  Microsoft will either need to develop their own custom ARM cores or they will have to commission someone like Qualcomm to step up and do it for them in order to remain competitive with Apple.

rcfa said:

So intel’s 10nm is like TSMC’s 7nm

Yes, that's partially true.  Intel's 10nm process has a transistor density of 100.76 (MTr/mm2) which is roughly equivalent to TSMC's 1st generation 7NFF process (96.5) but significantly less than TSMC's 2nd generation 7nm N7FF+ process (113.9).  Even still, Intel is having major yield problems with their 10nm process and they are struggling to get where TSMC was in 2018.  I think we can all agree Intel is a long way away from TSMC's 5nm process (173) in terms of transistor density.  Intel's attempt at 7nm has been delayed and is at least 2 years way.  TSMC will have moved on considerably by that time as well.

Eventually they may just have to do a new RISC architecture, but one designed to run a Rosetta-like software layer for x86 emulation particularly efficiently, i.e. quasi exposing microcode as RISC instruction set...

Intel already has a RISC core that executes RISC like micro-ops. The have since the P6 (Pentium Pro).  One of their fundamental problems is their legacy ISA with CISC instructions.  This is problematic in that it limits the optimizations possible for their decoder due to the unpredictable instruction sizes.  They'll never solve that problem without abandoning their current ISA.

cloudguy said:

According to what I have read, Samsung is 1 quarter behind TSMC, not 6 months. But that is academic anyway. Apple always comes out with devices featuring their new chips in September. Devices with the new Qualcomm and Samsung chips debut a few weeks later in November (except when midrange Samsung Exynos devices at times launch a bit earlier than the new iPhones). So the 3nm A16 would be first sold in iPhones and iPads that debut September 2022. While some obscure midrange 3nm Android devices may launch at about the same time or even slightly earlier, the flagship 3nm Android phones will launch starting with the Chinese brands in December 2022 and the rest in 2023.

Your comment assumes TSMC and Samsung's process nodes are the same based on similar marketing naming conventions such as 5nm.  They are not.  TSMC's 5nm process (5N) has a density of 173 million transistors per square millimeter.  Samsung's 5nm process (5LPE) has a density of only 127 million.  The point being, even at the same marketing name, TSMC has the superior process.  Honestly, Samsung's 5nm process is only marginally better than TSMC's 7nm process. 

Also, who would be the other buyers for TSMC's 5nm process anyway? MediaTek avoids the latest process nodes in order to save money. Their best 2021 chips will use a 6nm process and their 2023 ones a 4nm. While Qualcomm prefers TSMC, going back to Samsung is fine for them, as it would be for Huawei - presuming they are allowed to buy chips again - also. AMD's 5nm chips won't launch until 4Q 2021 meaning their 3nm versions won't until 4Q 2023. As for Intel, as they are considering having Samsung make their first batch of 7nm chips in 2021 (Intel would prefer TSMC but TSMC's 7nm nodes are fully occupied with chips for Qualcomm, MediaTek and AMD right now) they may be ready for 5nm by 2023. 

Who would buyer's of TSMC's 5nm process be?  Anyone that has the chance.  This is currently the world's leading process.  TSMC has customers lined up to buy what's left over.  Apple gets first dibs because they are laying out huge amounts of capital upfront that other companies can't afford to do.  This in turn gives Apple a perpetual advantage. 

So even were Apple to have exclusive access to TSMC's 3nm process for all of 2022, that statement isn't very meaningful anyway: Apple, Qualcomm and Samsung are going to be the only swimmers in that pool, and Samsung will make the 3nm chips for Qualcomm and themselves. That will be the same situation as with the 5nm this year. The first 5nm chip was actually supposed to be the Samsung Exynos 995, which was going to be in certain international Galaxy Note 20, Galaxy Fold 2 and Galaxy Flip 2 phones as well as all Galaxy S20 Fan Edition phones. Samsung suffered a setback at the last stage resulting in bad yields and the 995 was cancelled. However, Samsung did release the 5nm Exynos 1080 midrange chip in a Vivo phone 3 weeks after the iPhone 12's release.

Again, your assumption that the marketing name of 3nm actually means the process nodes are truly equivalent.  They are not.  There is a reason Apple switched from Samsung to TSMC.  Yes, Samsung will put something out called 3nm, but for the past few generations, TSMC's clearly had the superior process node with the same marketing name.  I highly doubt Samsung is magically going to leapfrog anyone with their 3nm process being somehow better than TSMC's.

cloudguy said:

bageljoey said:

I know people like to make fun of folding mobile devices, I think the problem is that it hasn’t been done well yet. I assume a well thought out and a well done folding device that is durable could be awesome. 

I would be glad Apple is working on it—I want to see what they come up with when the technology is there to make it to Apple’s standards.

"People" don't make fun of folding mobile devices. Apple fans do. They make fun of every idea and feature that Android OEMs develop first ... right up until Apple copies it.

People make fun of bad products.  Period.  As bageljoey said, "I think the problem is that it hasn’t been done well yet.". I'm interested in the concept of a folding phone, but I'm not interested in any of the existing implementations of that concept. Also, before accusing Apple of "copying" anything, you should to a little research and you'll see that Apple has had patents on this as well for YEARS before Samsung, etc. every brought one to market. The difference is, Apple usually doesn't bring half-baked ideas to market as most of the Android OEMs do.

cloudguy said:

Look, the idea that "it takes Apple to come along and get things right" is just something that Apple fans come up with to justify getting features years late. Samsung actually has a better mobile payments solution than Apple does because it includes both MST and NFC, making it work at every credit card terminal that stores don't software-block. 

Actually, NFC is a funny example. Google just put the technology in the phone but literally didn't know what to do with it. Their "Wallet" was a joke. Apple had to show them how to properly implement a secure e-commerce system with ApplePay. That's the difference between iOS and Android. On Android, lots of raw technologies are thrown out in a rather haphazard fashion. With Apple, they implement features and use the appropriate technology for those features. Apple didn't just add NFC, they build a secure enclave in their chip, they built a proper payment system and they created ApplePay. Google had no choice but to abandon their garbage and copy Apple with Android Pay. The same with Touch ID. Sure, a really bad fingerprint scanner existed on the Motorola Attrix. It was garbage and it wasn't even secure. Of course no other Android OEM copied that feature. Not until Apple showed the industry how to get it right. Not only was their scanner far better, but was also secure. Want another example? How about FaceID. Sure, Android had versions that could easily be fooled by photographs. They were first right? (Despite this technology already existing in iPhoto). It wasn't until FaceID that used real depth mapping before we had an implementation that was secure enough for e-commerce. etc, etc.. The list goes on.

cloudguy said:

1. Make the Galaxy Note essentially the Galaxy S with a bigger screen and stylus.

2. Enable stylus support on all their premium phones. 

What makes you think Apple would ever be dumb enough to put a stylus on a phone? Clearly, Apple has the technology and they've done a much better job of it with their iPad / pencil combination for years now. It's fine for a tablet but rather stupid for a phone. I don't see Apple changing their position on that, nor should they.