AppleInsider has estimated 2% from its surveys of repair centers.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Intel Delays 7-Nanometer Chips Until Late 2022 or Early 2023
- Thread starter MacRumors
- Start date
- Sort by reaction score
They are very significant. You keep saying that x86 is "legacy", but the only thing legacy is the ISA. The actual CPU architecture (that implements the ISA) of modern x86 CPUs is not "legacy" at all.
The first Itanium version had basically an x86-32 CPU on die. Very few of them were sold. Itanium itself was an entirely different architecture (using EPIC principles) and ISA, and not backwards compatible, although Intel had a software emulation layer (not unlike what Apple is doing with Rosetta).
That must be why x86 has dominated the market for decades.
It's a specialized Atom CPU with network acceleration functions. Not sure how that is a negative. It's an example of Intel selling into diverse markets that e.g. AMD does not cover.
A couple more thoughts:
Apple have many desktop computers that haven’t been updated in many years - and that has been the case with Apple desktops for a very long time. So two years wouldn’t matter much and certainly on an Apple desktop it wouldn’t.
I love what AMD is doing and have a 3900x. But if you put an intel $329 10700 in a Mac Pro sized case with a $30 fan (much less extensive thermals than the g5 Mac pro needed to be extremely competitive with AMD even at 14nm ) you really have an incredible amount the performance for any non scientific computing or non 3d computing, and it’s faster than AMD’s 3700x offering, at basically the same price. (For workstation computing it’s a complete unknown what Apple will be able to do or if they care and just because a ARM servers exist now, it does not mean that Apple will have an Apple silicon workstation anytime soon)
CPU processing is cheap and powerful in 2020. However when you have a battery powered device in a small case then the heat generated and power use by x86-64 to break down the instructions into RISC like processing down the line is an inefficiency that in an ideal world would be eliminated. (Assuming your software is updated which is not a trivial concern)
Since clearly notebook computers continue to be a larger share of the market and considering the trajectory Apple created with it's chips this move is about creating laptops that are as fast as desktops. WWDC literally said as much. That is the reason for the switch and that is the reason to be excited.
(I’m sorry to those of you who were apparently traumatized when Apple had told you how special you were for having a PowerPC but then moved you to unwashed Intel)
Apple have many desktop computers that haven’t been updated in many years - and that has been the case with Apple desktops for a very long time. So two years wouldn’t matter much and certainly on an Apple desktop it wouldn’t.
I love what AMD is doing and have a 3900x. But if you put an intel $329 10700 in a Mac Pro sized case with a $30 fan (much less extensive thermals than the g5 Mac pro needed to be extremely competitive with AMD even at 14nm ) you really have an incredible amount the performance for any non scientific computing or non 3d computing, and it’s faster than AMD’s 3700x offering, at basically the same price. (For workstation computing it’s a complete unknown what Apple will be able to do or if they care and just because a ARM servers exist now, it does not mean that Apple will have an Apple silicon workstation anytime soon)
CPU processing is cheap and powerful in 2020. However when you have a battery powered device in a small case then the heat generated and power use by x86-64 to break down the instructions into RISC like processing down the line is an inefficiency that in an ideal world would be eliminated. (Assuming your software is updated which is not a trivial concern)
Since clearly notebook computers continue to be a larger share of the market and considering the trajectory Apple created with it's chips this move is about creating laptops that are as fast as desktops. WWDC literally said as much. That is the reason for the switch and that is the reason to be excited.
(I’m sorry to those of you who were apparently traumatized when Apple had told you how special you were for having a PowerPC but then moved you to unwashed Intel)
Last edited:
Do you have the link? I don’t recall him being that adamant, but we could have heard/read him in different interviews. Or most possibly memory doesn’t serve me at all 😂
But you CAN blame Intel when Apple says, “Hey, this is what I’m coming out with over the next couple years, can your processor handle that” and Intel says, “OH SURE!” but then goes, “Ummmm, so we didn’t quite hit the mark, we overshot power consumption by like 20 watts”. Which is not far from what happened.
For their mobile systems, the ones that make up 80% of what they sell, there’s been a few times where Apple introduces a product that’s using an unspecified Intel processor that shows up in Intel’s database AFTER the release.
Bootcamp was a thing primarily because at the hardware level, the systems were VERY similar at as they were both using Intel compatible chipsets/motherboards. That’s not the case for ARM. If the architecture put together by Microsoft is not structurally the same as the one put together by Apple (which it’s probably not) then Bootcamp becomes much less of a certainty.
ARM is currently owned by SoftBank which is a Japanese company
[automerge]1595606723[/automerge]
Not necessarily. As long as Apple Silicon Macs get the developer mindshare and support, eventually they will do what your x86 Mac will do. That's what it comes down to
[automerge]1595607135[/automerge]
Apple is one of the few customers that had deep access to Intel's roadmap. And it's based on that roadmap and Intel's promises upon which Apple designs their laptops. It's not Apple's fault that Intel has missed their roadmap targets for the last 4 to 5 years.
Last edited:
I am the former Product Manager for the G3 upgrade cards from Newer Technology. I "negotiated" chip prices in Burlington with the Boss Lady. I do not remember her name. We would sit around a table and ask why the auto industry go the best prices and Apple next, then us. They seemed to not understand that lower chip prices to us would provide a more robust business model for them. Instead it was, "we have a schedule we go by and you don't buy enough". Foolish then... foolish now it seems.
However, that 2% includes a lot of people who don't actually regularly use Boot Camp. I still think it's likely less than 1% (or at least closer to 1%) that use it regularly. For example, I had Boot Camp installed too on a couple of machines, but stopped using it because Parallels was sufficient and more convenient. It took me a while to finally remove Boot Camp because I couldn't be bothered, and I liked having Boot Camp there "just in case".
So yes, there are certainly people out there who regularly use Boot Camp, but at least with anecdotal evidence, most of the people I knew who had Boot Camp installed didn't actually use it regularly, myself included.
Hmm... That seems perfectly reasonable to me. I don't know why you would think you should get the same pricing as Apple. Obviously, companies like Sonnet and Newer Technology were small fry in comparison.
Last edited:
Just understand that Apple is hostage to TSMC and their process.
TSMC 7nm = Intel 10nm
Apple doesn't own a FAB.
[automerge]1595608612[/automerge]
Could you explain what this means?
Are you saying that diverse candidates aren't qualified candidates?
Diversity isn't hiring unqualified candidates. Diversity is making sure you hire *ANY* qualified candidate and that you also pursue qualified people of color. Nowhere does it say hire someone for the job that is not qualified.
Well, as you know, TSMC isn't the only fab out there. There is also Samsung and Global Foundries for example. However, TSMC is currently the best (and it's not Samsung) so Apple goes with them.
That's the general rule of thumb. 10 nm from companies like Samsung and TSMC is much less dense than 10 nm from Intel. Take a look at this Wiki page on 7 nm for example:
7 nm process - Wikipedia
Transistor density
Intel 10 nm - 100.8 MTr/mm2
Samsung 7LPP - 95.3 MTr/mm2
TSMC N7FF - 96.5 MTr/mm2
In contrast, look at the numbers for 10 nm:
10 nm process - Wikipedia
Intel 10 nm - 100.8 MTr/mm2
TSMC 10 nm - 52.5 MTr/mm2
Samsung 10 nm - 51.8 MTr/mm2
As you can see, Samsung and TSMC are much more liberal in what they call 10 nm or 7 nm, and in fact, TSMC's and Samsung's 10 nm aren't even in the same league as Intel's 10 nm in terms of transistor density. It is about one process generation less dense than Intel's.
However, as far as Apple is concerned, they could call it Ralphie and Apple couldn't care less. TSMC is already on 5 nm, and Intel isn't on 7 nm yet. In fact, Intel is still having issues with 10 nm, whereas TSMC's 7 nm is relatively mature.
Last edited:
But they make other compromises. Also, fans blowing don’t indicate throttling or overheating in themselves.
In any case, all PC makers would like more efficient, lower heat-generating processors as it gives them more flexibility in their own designs.
[automerge]1595609660[/automerge]
The enterprise ship has sailed, at least for those customers who use Windows. Buying a Windows PC is better than using Boot Camp, since Apple doesn’t spend the resources to optimize Boot Camp drivers (they have better things to do). It’s also not 2005 anymore. Mac is an important product, but not the only product or even the top product anymore. Having tighter integration with the rest of its ecosystem is more important for Apple than having it be fully compatible with the Windows world.
The classic idea or RISC/CISC is no longer applicable.
The ARM 32/64Bit architectures are most definitely not even closely related to the RISC architectures of Hennessy and Patterson of 30 years ago.
ARM is heavily pipelined and although it can dispatch an instruction in a single cycle it takes more than one cycle to execute an instruction. ARM 32/64 are CISC-y: variable length instructions, instructions that read/write multiple registers (push/pop),they also have floating point (NEON), speculative execute, branch prediction, multiple threads, etc....
This is not classic RISC where it takes one clock cycle to execute an instruction. When you look at ARM server class processors from companies like AMPRE; they don't look like cell phones anymore.
There are no true CISC or RISC processors anymore.
X86 is a CISC with a RISC core architecture.
ARM is a RISC-y with a lot of CISC features.
TMSC is spending 20 billion dollars for their 3nm fab. I assume Intel would have to do the same amount of expenditure to keep up. The things is there is a much higher turnover for smart phones than there are for PCs. That is why I think Intel will fall farther behind has time goes on.
www.fierceelectronics.com
TMSC just manufactures chips. Intel does much more than manufacture cpus and so they cannot focus on the process node like TMSC can.
TSMC starts building 3nm plant in Taiwan worth $20B
Apple iPhone runs new A13 processors fabricated by TSMC
www.fierceelectronics.com
TMSC just manufactures chips. Intel does much more than manufacture cpus and so they cannot focus on the process node like TMSC can.
I used TSMC as the comparison because Apple uses TSMC.
I know they have also used Samsung.
Global Foundries stopped 7nm development; they have 12nm.
Granted, Intel has struggled to get 10nm going and delayed 7nm, again.
Apple to date has been fortunate working with TSMC and Samsung.
The A series processors are not as large as desktop CPUs.
TSMC is moving to 5nm and worked on the process with AMD.
Apple will no doubt benefit.
But Apple is moving into territory that only AMD and Intel have been successful playing.
Sun/Oracle couldn't compete and SPARC is for the most part dead. Fujitsu is one of the last hold outs and they recently said their massively parallel machine will move to ARM.
Apple will need to keep performance and TDP in check as they now compete with the likes of AMD and Intel.
Make no mistake; Apple will face the same challenges that Intel and AMD face in terms of power for PCIe Gen4 and other differential and power hungry peripheral interface, Yield on L2/l3 caches, etc.
This isn't the cell phone processor in an iPhone 11.
If Apple has a misstep, it will be immediate and they can't blame anyone else.
If TSMC has a misstep, Apple will be affected.
Apple has to make sure they don't mess up in their designs and pray that TSMC does't screw up with their process.
[automerge]1595610970[/automerge]
TMSC is spending 20 billion dollars for their 3nm fab. I assume Intel would have to do the same amount of expenditure to keep up. The things is there is a much higher turnover for smart phones than there are for PCs. That is why I think Intel will fall farther behind has time goes on.
TSMC starts building 3nm plant in Taiwan worth $20B
Apple iPhone runs new A13 processors fabricated by TSMC
TMSC just manufactures chips. Intel does much more than manufacture cpus and so they cannot focus on the process node like TMSC can.
I disagree.
The same group running the fabs and doing process development is not the same group doing other stuff.
Different business units.
Using that logic, Apple is at a big disadvantage for chip design, because they also have to make the computers and iDevices and the OSes to run them.
But we already know that is not the case here. Companies can do more than one thing at a time.
I'm guessing you have a 4k monitor?
I'm running my 16" on a TB2 LG 21:9 @ ultra wide 3840 x 1440 (I think). The fans are fine, but sometimes connection/ disconnection causes a reboot.
I'm running my 16" on a TB2 LG 21:9 @ ultra wide 3840 x 1440 (I think). The fans are fine, but sometimes connection/ disconnection causes a reboot.
Move over, Intel. You're the 2020 version of the PowerPC. Can't wait for ARM. Typing this on a 16" MBP with blaring fans just because an external monitor is plugged in.
Right because Zen 3 is about to be out on the most mature 7nm TSMC node and Zen 4 is slated for 3nm with X3D stacking that Apple won't have allowing for AMD to really distance itself from anything Intel can dream of doing; and Apple won't touch, but hey! We're doing our own Silicon! It's not ready and won't compete with AMD but it will won't be far behind Intel first go around and who knows where it will end up!
Apple should have switched to AMD and continued refining Silicon to compete close to AMD, if possible, but that's never going to happen.
You have a CEO named Lisa Su who is one of the most accomplished minds in all of CPU/Memory design whose team is world class just as much as Apple's team. Just ask Jim Keller who knows both of them.
I get the feeling you're arguing in circles. What I said is:
Which I think agrees with your statement that the only thing legacy is the ISA.
Modern x86 CPUs are very sophisticated Rube Goldberg devices meant to perpetuate a legacy ISA because the only reason Intel exists is for backwards compatibility. I think you're confusing "top notch architecture" with "best you can do with a sub optimal instruction set". This is why I say the distinction between ISA and CPU architecture isn't significant. Putting alloy rims on your Edsel doesn't give you all the benefits of a ground up rethink.
No, what the first Itanium did was infinitely more stupid. They tried to create a hardware version of Rosetta. It wasn't just an x86 coprocessor, it was a JIT translator that fed the EPIC pipeline. If you read the link I gave you, it clearly states that Itanium had to be backwards compatible to MS-DOS version 1.
If you look carefully, you might be able to spot the albatross...
HP pushed this same objective: "It is a gateway into the 64-bit future but it still remains completely 32-bit compatible."
They later moved to software translation, but by that point the writing was on the wall for Itanium and AMD provided a much needed parachute.
This isn't the only thing that sank Itanic. It was also badly conceived and poorly executed. EPIC was just a punchier way to say VLIW, and VLIW has proven quite good for dedicated applications like DSPs but it's not a great way of building a future proof processor family. The whole concept depends on the compiler scheduling the execution units and that breaks down if you want to run old code on an updated set of execution units.
It was late, hot, big, expensive and never performed as promised. It's competitive advantage was supposed to be that supporting x86 would bring along Intel's existing installed base, and it failed at that just as badly.
x86 has dominated the market because it has dominated the market. It was selected for the IBM PC and Intel has been trying to repair the spacecraft midflight ever since. It has stayed dominant because people want to know they can run last year's code on this year's machine. Compilers have been optimized for it, coders have grown up on it, libraries have been built for it. This backwards compatibility commands higher margins which, in turn, provides Intel a huge pool of capital with which to buy lipstick for their pig in the forms of super complex pipeline architectures, microcode optimizations and very importantly, cutting edge process technologies to mitigate the speed and power penalties of backwards compatibility.
The installed base has drawn in innovation from outside, most notably from AMD which managed to bring it into the modern 64bit world while Intel was busy rearranging deck chairs on the Itanic.
x86 hasn't dominated because it is better, it has dominated because it's familiar. Oh, and it hasn't hurt that Intel has the money and clout to kill off rivals in very non-technical ways.
Nobody said it was a negative. I said this:
and you replied with what sounded like it was meant to be a counter argument:
I'm merely pointing out that their storage business lost $1.2B last year and the chips they're selling into basestations are x86 processors. I don't think that undermines my argument in any way, except I should have probably phrased it as "can't make money on anything" rather than "can't ship anything". They do seem to be able to ship other things at a loss.
My point through all of this has been that Intel has never been able to really branch out beyond x86 and has relied heavily on their process technology to keep them alive. If they can't get it together, they're in for some real pain. Apple showing a path out of the caverns isn't going to help.
