Intel says costly 10nm ramp will counter PC slowdown Intel promises that a costly ramp in production of 10nm processors, which includes the forthcoming Sapphire Rapids server chips, will pay off in the second half of the year to counter a slowdown in the PC business. In the x86 titan's first-quarter earnings call Thursday [PDF], the semiconductor maker stuck with its previous …
You can have it when you stop faffing around in the 10nm range & catch up to the competition that's been merrily kicking sand in your face from the 7, 5, and even 2nm playground.
You go on about how much better your 10nm is compared to everyone else's 7nm stuff, but all it takes is a side-by-side comparison of how energy hungry your chips are compared to everyone else's & all the public is left with is the taste of the whiney kid shouting about how much better their hand-me-downs are versus everyone else's brand new clothes.
You want respect, figure out what's keeping you from the advanced table with all your competition & get your ass in gear. As long as you're still in the double digit nm range, all the single digit players are gleefully eating your lunch.
Look at the very first image and learn https://en.wikichip.org/wiki/7nm
"Intel 10nm process" is just a commercial name for a generation of a certain size and its technology, and does not represent any geometry of the transistor on the chip. Intel "10nm process" has more transistors per square millimetre than the "TSMC 7nm process" or the "Samsung 7nm process".
The real problem is that TSMC was using their "TSMC 7nm process" back in 2018, Samsung 2019 and Intel has only caught up in 2021 with their "Intel 10nm process". All three are similar enough to be roughly considered the same, even though Intel's technically has about 10% more transistors per unit area than TSMC and 5% more than Samsung.
You could think of the process name like a brand name, it has no connection whatsoever to any measurement on the silicon.
If a "<insert company brand name here> 1 nm process" actually used structures in a silicon chip that were 1 nm (they will not):
The Van der Waals radius of a silicon atom is about 210 pico-metre, so 1 nanometre in the real world would be like 2 silicon atoms wide. And that is not even thinking about the effects of natural impurities or doping agents.
At that scale you would need to use only one Isotope of Silicon ( Silicon-28 ) because trace amounts of normal silicon is naturally radioactive. How would a 2 atom wide signal path behave when one atom undergoes radioactive decay, probably emits an electron and then changes into a phosphorus atom (which is a doping agent for silicon)! So you need to purify the silicon using similar techniques to the ones used to purify the elements used in nuclear bombs, but to make isotopically enriched silicon tetrachlorides instead which can then be used to grow single crystal 99.9999% pure silicon-28.
Oh and the doping of silicon never actually stops, it just slows down a lot at room temperature, so the lifetime of chips made with such small structures is very limited, at some future date the dopants in bleeding edge chips made today will be homogeneous. And how long that will be is a low number of years.
Intel shareholders' current reaction to the plans to regain the process advantage is probably a big reason why Intel lost the lead in the first place.
From the linked article:
"Stacy Rasgon, a senior analyst at Bernstein Research, gave Intel an “underperform” rating and set its target price to $40 per share in a note to clients after the presentation. He told CRN that the big concern for investors is that “there’s no quick fix” for the chipmaker’s problems that began before Gelsinger became CEO, which means lower returns for shareholders for the next few years.
“They basically said there’s no free cash flow until 2025. And so even if you believe them, and you say, ‘Well, yeah, 2025 is going to be beautiful,’ like, what’s your rush? It‘s only February of 2022, and I got three years of pain before I can even hope to get anything,” he said. "
Geez, there is no quick fix to problems that developed during the last decade (when the shareholders were all too happy)? I wonder why...
Too much looking at the stock market and the numbers for the next quarter or year, not enough prestige for the multiyear R&D efforts required to keep the cash cow alive?
And when R&D does not lead a company like Intel, when technically successful projects don't lead to promotion, it's all too easy to cut a few corners and save a few bucks, assign the wrong people to lead the projects, and/or fail the favourite people upward all the way to senior management. Takes a long time for the damage to show up.
I have no inside information, except about the way things went at another (former) giant.
If he thinks that producing 10nm chips will revitalise the PC world then he is sadly mistaken.
This might be a message that the stockholders can swallow but out in the real world things are a tad different.
People and companies all over the world (apart from the oil/gas giants) are fighting rising costs. Food, energy and raw materials are all increasing in price rapidly. After over decade of low, low inflation people and companies are hurting.
Just because you are colluding with Microsoft to make/force people upgrade their PC's to run the monstrosity that is Windows 11... won't return Intel to their past glories.
Most people with W10 on their laptop/desktop already have more than enough computing power to do what they want (gamers apart). With everything under the sun, going into the cloud, then it is not your leccy that is powering it (directly that is) but someone else's. You will foot the bill but indirectly.
What USP will make people move to your 10nm chips Mr Intel?
A lot of people are seeing what the future looks like with ARM powered devices. After all, they already have a super powered device in the shape of their phones. What more do they need eh?
Sorry Mr Intel, you are as Jethro Tull said, 'Living in the Past'.
I am done buying X86 based CPU's in any shape or form. I suspect that I am not alone there.
This makes for interesting reading and to me, is a sign of the times ahead.
https://hardware.slashdot.org/story/22/04/30/1728232/mac-studios-m1-ultra-chip-outperforms-on-computational-fluid-dynamics-benchmarks
I'm sure SR will be a good product, but I do wonder about quick adoption - it's got a lot of expensive new tech. DDR5 will have a significant premium over "legacy" DDR4 for example, (not just because the die size has to be bigger).
As much as they've abused their position, I hope they are allowed to succeed. The "West" does need a technology leader in manufacturing as well as design.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
