TSMC are working towards 3nm. Their 5nm processors are ready for this year’s iPhones.
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
nah, AMD are still making significant strides.
lts intel’s process tech. within 5 years they will abandon building their own and farm out manufacturing to TSMC. or become irrelevant.
AMD went through this a decade ago.
You have to remember that process node sizes are now meaningless and something marketing made up. Intel's processes are 1/2 to 1 whole step ahead.
While Intel will be late, Intel's 7 nm wil likely be competitive with TSMC's 5 nm.
that’s what intel were claiming with their 10nm being equal to or better than TSMCs 7nm. since like 2016.
it isn’t. it’s trash. it doesn’t beat 14nm on power consumption or clock speed. and they still can’t manufacture even 10nm in volume.
intels leadership does not exist past 14nm. it just doesn’t. it’s vapour.
It isn't "trash" when you look at actual engineering-based numbers like SRAM bitcell sizes and voltages.
14 nm lithography process - WikiChip
The 14 nanometer (14 nm) lithography process is a semiconductor manufacturing process node serving as shrink from the 22 nm process. The term '14 nm' is simply a commercial name for a generation of a certain size and its technology, as opposed to gate length or half pitch.
en.wikichip.org
10 nm lithography process - WikiChip
The 10 nanometer (10 nm) lithography process is a semiconductor manufacturing process node serving as shrink from the 14 nm process. The term '10 nm' is simply a commercial name for a generation of a certain size and its technology, as opposed to gate length or half pitch. The 10 nm node is...
en.wikichip.org
Well, it takes more than waving a wand to make Moore's Law come true. People act like there's some unlimited printing press of computer engineers for microarchitecture where salary necessarily equals progress. It's more often that money that inhibits the process.
It isn't "trash" when you look at actual engineering-based numbers like SRAM bitcell sizes and voltages.
14 nm lithography process - WikiChip
The 14 nanometer (14 nm) lithography process is a semiconductor manufacturing process node serving as shrink from the 22 nm process. The term '14 nm' is simply a commercial name for a generation of a certain size and its technology, as opposed to gate length or half pitch.
it is “trash” when you look objectively at performance. end users don’t care about spec. they care about end result, and the end result is trash. on paper spec means nothing if they can’t get it to yield properly. and they can’t.
they are using 14nm on their ultra low power stuff because it is better. they are using 14nm on the high performance because it is better. they have the only stuff they can barely produce on 10nm to be able to claim to shareholders they shipped 10nm in volume.
ive got a 10nm machine. the process is garbage. it’s no better than 14nm and my ipad pro from 3 years ago performs better with no fan.
intels 14nm clocks higher and consumes less power.
ice lake has some decent architecture improvements but that is not process related. if it did not have significant architecture improvements, the parts would be worse than skylake. as it is, it barely creeps over the line, crippled by intel’s 10nm.
even intel admit that 10nm is a failure by continuing with 14 and attempting to skip to 7. which is now having issues already. same old same old.
Last edited:
Why assume? Tim Cook explicitly stated this is Apple's plan
[automerge]1595569490[/automerge]
The difference this time is that Apple's in-house silicon design team is every bit as good as Intel's and has little to no trouble attracting top talent in this field.
Amazing. Obviously, Intel has been capable of producing 7 nm and smaller chips for ages (remember that for decades they reduced sizes by 30% or so every 1.5 to 2 years or so; really revealing!), but they do not do it and delay forever because once they reach 1 or 2 nm or so, that is the end of silicon for chips. But there are competitors that are doing it. It is shocking how a company like Intel is killing itself.
Just look at Boeing for example. MBAs running everything with total lack of insight.
For now. But they're already behind the curve and it sounds like they're at least 30 months away from bringing up their 7nm node...
They have in the past. If TSMC remains reasonably priced and maintains their technology edge, then there's no need to go anywhere else. If TSMC falls behind, that implies someone else stepped ahead, and Apple can move their business there.
The point I was answering is that Apple doesn't need to scale a new process, they can shop for one.
Yep, my mistake. I meant x86-64
I'm playing a little fast and loose with the distinctions between instruction set and CPU architecture because I don't think the distinctions are material to the conversation.
Cleaned up, sure. But still necessarily backwards compatible. x86-64 still supports memory segmentation for gods' sake.
Other CPU architectures can compete on performance, what they can't do is run most of the commercially available software. DEC Alpha was bought by Intel and killed. PA-RISC was a casualty to Itanium. Power is still viable, but niche. As we're seeing, ARM is stepping into the ring.
Yes, it was. It performed about as well as a 486, but they put almost a third of the die into supporting x86.
The failure of Itanium isn't proof that x86 is "top notch"-- it's proof that Intel's development process is an embarrassment.
Fair enough, then "in support of" might have been a better choice of words. If the storage group is only 8% of the revenue in total (and almost certainly less than 8% of the profits), then the non Optanium part is not what's going to carry Intel into the future if the CPU business craters.
Memory and storage are commodity devices versus the insanely high margins on their CPUs. They run memory through their fabs to defray the fixed costs, not because it's a cash cow.
In 2019 the storage business reported a $1.2B operating loss.
Correct. Modems and basestations are different things. My point is that other than a few hobbies, Intel isn't very interested in or successful at anything other than cranking out x86 chips. The failure of the modem business is evidence of that.
In the Intel press release, they highlighted their growth in 5G base stations by pointing to the Atom P5900 which is, of course, an x86 processor.
Last edited:
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.
What are you talking about ? ARM is RISC PowerPC... it never went away, maybe the other one did.. but PPC never died, it just REINCARNATED into ARM.. PPC won, intel lost.. And so proud of that fact.
AMD
is already on 7 and will be on 5.... ARM will be a joke vs AMD Processors. sorry Apple made a big mistake this time.ARM is great for ULV.. and fixed data streams.
[automerge]1595572240[/automerge]
Nah, they haven't.
We're all running into the laws of physics, and TSMC is simply doing better at bending them than intel is these days.
There used to be a lot more foundries than there are today. Things are getting more difficult and they're slowly going out of business and the more competent ones are surviving.
Intel even identified this happening a decade ago. There are leaked internal intel slides on the net about it. They however assumed that they would be the one to succeed, based on their historical process lead. They stumbled with 10nm, got leapfrogged by TSMC and Samsung and are now playing catchup. Poorly.
[automerge]1595572409[/automerge]
This is exactly what they're planning to do, and its even in the linked article IIRC. They're planning to farm out some of the tiles on their multi-die arch to other foundries.
Yeah, but they will be 2 years late, 2 years TSMC can use to drive their process forward. In volume.
[automerge]1595572490[/automerge]
Bad wording on my side, I meant that it will take them the whole two years, no more, no less.
When you talk about something that it isnt even released....from developers ARM is the future...one chip for all platforms...who still think that x86 is still the future...its clear it doesnt work in the chip department or app dev, at the end the best fastest chip/computer is an Arm in present, the system is called Fugaku and is installed at the Riken Center for Computational Science in JapanAMD
ARM is great for ULV.. and fixed data streams.
[automerge]1595572240[/automerge]
Now show me your AMD, or you are just a fanatic amd chip that doesnt look around him!?
The difference is that power pc was tiny volume, and was wiped out by the immense volume of cheap clone pcs bring made at the time. Intel could afford to win that battle because of their volume.
This time, Apple is moving to an architecture that's not only immense, but that they themselves already make. 100 million a year of them.
I'm surprised that they didn't announce this earlier, to be honest. Perhaps it took many years to be ready. My son uses the computer I bought in 2012 with its Intel i7, and it's embarrassing how little the processors have improved since then. Intel has been failing to provide regular significant upgrades for Apple for years, and I bet they've been planning this for a very long time.
