Intel's Future Now Depends On Making Everyone Else's Chips (arstechnica.com) 61
An anonymous reader quotes a report from Ars Technica, written by Tim De Chant: Over the last year and a half, as the pandemic has everyone turned to their screens, demand has surged for devices (phones and laptops) and cloud services (Netflix and Zoom), all powered by a range of advanced semiconductors. Manufacturers have raced to squeeze more chips out of their fabs, but many were running near their limits before the pandemic. Still, Intel and its competitors didn't rush to build new fabs -- fabs are startlingly expensive, and without continued demand, semiconductor firms are loath to build more. But now, as the global pandemic continues to disrupt supply chains, chipmakers have decided that the current spike in demand isn't going away. Intel's $20 billion investment [to build two new chip factories in Chandler, Arizona] is only one example. Samsung announced in May that it would spend $151 billion over the next decade to boost its semiconductor capacity. TSMC made a similar announcement in April, pledging to invest $100 billion in the next three years alone.
The investments required to stay at the leading edge -- where the most advanced chips are made -- has whittled down the number of semiconductor competitors from more than 20 in 2001 to just two today. "There's really only so much room at the leading edge, just because of the huge capital costs involved," said Will Hunt, a research analyst at Georgetown University's Center for Security and Emerging Technology. That cost is driven by the price of the equipment that's required to etch ever-smaller features onto chips. A few years ago, the industry began to use extreme ultraviolet lithography (EUV) to shrink transistor sizes. EUV machines are marvels of physics and engineering, and one tool costs upwards of $120 million. To stay relevant, companies will need to buy a dozen or more annually for the next several years. For those sorts of investments to make sense, semiconductor manufacturers must produce and sell an enormous volume of chips. "When you have volume orders, then you can do yield experiments, you can improve your yield, and yield is everything because that's how you cover your costs," said Willy Shih, a professor of management at Harvard Business School. Which is why Intel, under [Intel CEO Pat Gelsinger], is doing something now that it historically has shunned. "We are now a foundry," Gelsinger said at the Arizona groundbreaking. In the coming years, he said, Intel will "open the doors of our fab wide for the community at large to serve the foundry needs of our customers -- many of them US companies that are dependent on solely having foreign supply sources today."
But becoming a leading-edge foundry isn't just about building fabs and telling customers you've got space to make their chips. Gelsinger will have to change Intel's culture and, to some extent, its technology, both of which are deep-seated. "He has to turn a huge ship around," said Robert Maire, president of Semiconductor Advisors. In the coming years, Intel has several challenges to master at once. As the company rolls out a new business model, it also needs to redouble its R&D efforts while still being careful with cash flow. (Intel has fallen so far behind that it now plans to outsource production of its most advanced chips -- and a portion of the profits that accompany them -- to TSMC.) The transition will demand intense focus. "The foundry business could be a distraction," Shih said. At the same time, he added, Apple, Google, Amazon, and other companies are moving away from Intel's standardized chips toward their own customized designs. If Intel doesn't change with the times, it risks being left behind. "There will be many challenges, and there will be tests that will face them," Shih said. "It's going to be hard."
The investments required to stay at the leading edge -- where the most advanced chips are made -- has whittled down the number of semiconductor competitors from more than 20 in 2001 to just two today. "There's really only so much room at the leading edge, just because of the huge capital costs involved," said Will Hunt, a research analyst at Georgetown University's Center for Security and Emerging Technology. That cost is driven by the price of the equipment that's required to etch ever-smaller features onto chips. A few years ago, the industry began to use extreme ultraviolet lithography (EUV) to shrink transistor sizes. EUV machines are marvels of physics and engineering, and one tool costs upwards of $120 million. To stay relevant, companies will need to buy a dozen or more annually for the next several years. For those sorts of investments to make sense, semiconductor manufacturers must produce and sell an enormous volume of chips. "When you have volume orders, then you can do yield experiments, you can improve your yield, and yield is everything because that's how you cover your costs," said Willy Shih, a professor of management at Harvard Business School. Which is why Intel, under [Intel CEO Pat Gelsinger], is doing something now that it historically has shunned. "We are now a foundry," Gelsinger said at the Arizona groundbreaking. In the coming years, he said, Intel will "open the doors of our fab wide for the community at large to serve the foundry needs of our customers -- many of them US companies that are dependent on solely having foreign supply sources today."
But becoming a leading-edge foundry isn't just about building fabs and telling customers you've got space to make their chips. Gelsinger will have to change Intel's culture and, to some extent, its technology, both of which are deep-seated. "He has to turn a huge ship around," said Robert Maire, president of Semiconductor Advisors. In the coming years, Intel has several challenges to master at once. As the company rolls out a new business model, it also needs to redouble its R&D efforts while still being careful with cash flow. (Intel has fallen so far behind that it now plans to outsource production of its most advanced chips -- and a portion of the profits that accompany them -- to TSMC.) The transition will demand intense focus. "The foundry business could be a distraction," Shih said. At the same time, he added, Apple, Google, Amazon, and other companies are moving away from Intel's standardized chips toward their own customized designs. If Intel doesn't change with the times, it risks being left behind. "There will be many challenges, and there will be tests that will face them," Shih said. "It's going to be hard."
Good luck, Intel (Score:4, Insightful)
They can't get enough EUV machines to produce more than 20 wkpm on Intel 7nm/Intel 4 by 2023:
https://twitter.com/chiakokhua... [twitter.com]
The majority of Intel's own silicon will be fabbed on 10nm variants: 10nm+ will likely be retired (IceLake-SP is their final product on that node), and production will eventually shift to 10SF or 10ESF/Intel 7. Without more EUV gear, volume on 7nm/Intel 4 and 5nm/Intel 3 will be seriously low, no matter how many more facilities they build.
So maybe Intel can expand their 10ESF/Intel 7 capacity, but anything else . . . ? Maybe they will have the problem solved by 2025 when they shift to 20A with help from IBM. Maybe! In any case, 10ESF/Intel 7 is already inferior to TSMC's N5 and N5P. Nobody can claim that Intel will be a leading-edge foundry.
Bean counters (Score:2)
Re:Bean counters (Score:4, Insightful)
. . . maybe. Their 10nm struggles ought to show you that throwing money at a problem isn't necessarily a solution. However, they could have significantly improved their prospects by ordering more equipment from ASML.
Re: (Score:2)
Intel is lucky. They have the market share, better branding, and hardly anyone one can tell if their chips are inferior or not.
Re:Bean counters (Score:5, Interesting)
That is one of the 'benefits' of investing money into advertising; people will prefer the product they've heard about already over some unknown product. If they don't know that your product is inferior in reality, they can't even begin to care. It's a real shame that situation. But that's how it is.
Before this pandemic situation I've built a lot of PCs for customers, and I did it primarily with AMD Ryzen CPUs. The thing is that most people never even heard of AMD. They know Intel, and Nvidia, because those do some advertising that gets through to your average Joe. But AMD? It always takes a bit of effort to convince them that Ryzens from AMD are the better bang for the buck.
Before Zen 3, I've only recommended Intel for pure gaming machines. With Zen 3, if you want top performance, Intel is only an option for some fringe cases where they have some proprietary stuff due to partnerships in various applications that can't be used by AMD. But then again, I've only built and sold a single machine *since Zen 3. Now I'd only recommend Intel's i5 series because of their performance/value ratio. What Intel's 12th generation can bring on the table still remains to be seen.
*Damned graphics card prices. Since I do have some integrity, conscience compels me to point people towards pre-built systems for cost effectiveness. Even though those systems are usually terrible, measured by the standards of how a well matched build could look like, because those OEMs get their graphics cards from the manufacturers at fair prices they can sell their systems at half decent prices still.
Re: (Score:2)
I wouldn't say that they're lucky. That is one of the 'benefits' of investing money into advertising; people will prefer the product they've heard about already over some unknown product
That's a really good point.
Re: (Score:2)
Hasn't AMD always been mainly about custom builds?
In 2 decades I've never worked in a corporate office with AMD desktops, always $VENDOR with Intel Inside (TM). And these days for my home I'm a cheapskate who buys second hand laptops where Thinkpads and Inspiron laptops are typically Core i5.
My one and only AMD machine lasted 16 years before I recycled it.And mainly from word of mouth in reading Slashdot who recommended an Athlon XP over Pentium 4 at the time.
Re: (Score:2)
Intel's been putting more effort into large business contracts with system-integrators that already had some significant brand recognition like Dell, Hewlett-Packard, Acer.
For that Intel had employed quite a bit of anti-competitive contracts, which we know now from various lawsuits, which are still in progress.
One of the latest examples a quick search came up with, with some references to other stuff inside: https://www.extremetech.com/co... [extremetech.com]
Intel did some questionable stuff to facilitat
Re: (Score:1)
I have been working with a few smaller companies, and where possible, have been pushing AMD based gear, simply cos they are more cost effective, especially when you consider the performance per dollar metric.
Even my own startup has a bunch of ryzen based workstations.
Not been involved with larger companies in the pass 20 or so years, so can't speak for them.
Re: (Score:2)
have been pushing AMD based gear, simply cos they are more cost effective, especially when you consider the performance per dollar metric
For small tasks maybe. For anything datacenter based where amount of memory matters, surely you joke. The CPU price hasn't taken more than ~1/4 of hardware costs for fat gear for quite a while.
Re: (Score:1)
I did mention smaller companies.
They don't really need HPC / DC class of gear generally. And if you need something with up to 4TB of RAM, AMD Epyc is still more cost effective.
Re:Bean counters (Score:4, Informative)
AMD's primary focus is on the server segment. Though their EPYC line is (at times) slower to reach the market than their desktop designs, you can clearly see that their CCD strategy has been aimed primarily at scaling core count in server/workstation CPUs without relying on massive dice. Zen4 marks a first where Genoa (a server part based on Zen4 CCDs) may reach the market before similar desktop parts.
Re: (Score:2)
and hardly anyone one can tell if their chips are inferior or not.
That is because no one dares to plug them out of the motherboard!!
Re: (Score:2)
I like the chips with a nice dip of guacamole.
Re: (Score:2)
and hardly anyone one can tell if their chips are inferior or not.
Narrator: They are.
Re:Bean counters (Score:4, Insightful)
Something like 95% of CEO income at Intel is stock, Intel was already printing money for the people who mattered. Stock buybacks give immediate predictable stock price increases, investments depends on investor sentiment.
Shareholders are retards, boards are nepotists and management are plunderers.
Re: (Score:2)
If Intel made this investment a few years back, they would be [...] ahead of everyone else [...]
Hell yeah. Decisions made by bozos before Pat were basically "let's sell parts of the company and put that cash into stock buybacks". If not for some lucky bits like eg. new memory types for datacenter gear, Intel would be in deep shit by now. But despite the bean counters' efforts, Intel still has so much more money than AMD and the rest, that there's more than enough to weather the investment gap.
The adage about CEOs goes: 1. engineer, 2. MBA, 3. accountant, 4. receiver. Pat means a jump back from 3 t
Re: (Score:1)
Good grief! It's not just your hideous typeface that's annoying, but your baffling use of the hyphen.
x86 is dead (Score:1)
Re: (Score:2)
It is the Post-Intel era.
Re: (Score:2)
ARM is dead, so are tens (!) of less-known archs. Take a look at what's happening with new designs in embedded -- it's all RISC-V. As for bigger machines, x86 is still the king, but there's a reason Intel is doing that SiFive partnership thingy. In my opinion, there's way too much foreplay involved, but let's see what happens...
In the meantime, as Linus said, "fuck nVidia" -- so let's do so, in a non-sexual way, all their acquisitions included.
Re: (Score:2)
Take a look at what's happening with new designs in embedded -- it's all RISC-V
"All RISC-V" is a big claim. Got links?
Re: (Score:2)
Right, I've exaggerated -- apologies. But, the only new embedded arch that got added to Linux in the recent years was c-sky, with plenty getting removed.
I don't know much about stuff too small to run Linux, but rumors I hear also say that.
Re: (Score:2)
Hyperbole aside, I'm very interested in what kind of takeup RISC-V is getting.
Re: (Score:1)
The X86 God is powerful. It even defeated Intel's precision Itanium chip that was in every way better than the X86. Mediocre managers wouldn't buy it any more than they'd switch to the superior digital Alpha chip.
It'll take something to finally kill X86. Hope I live to see what it is. I remember when I thought it was dead in the 1980s to the superior 88000. One of the few things I was ever wrong about.
Re: (Score:2)
I don't know why that's a problem.
It isn't, if it works. They're betting the company.
Tens of billions is being invested while the CEO is telling investors to expect low margin years. They're putting thousands of engineers on the payroll; labor costs are climbing. They execute or they will be in a very bad place.
Fortunately the plans make sense and the new leadership appears to know what Intel is for, so it will probably work.
Incompetent ... journalists (Score:4, Insightful)
Plus, only two? Samsung is still very much in the game... Sure, they got other priorities than Intel and TSMC, but they're keeping up. And for specific device types other manufacturers are more than keeping up.
Re: (Score:3)
Umm, I'm pretty certain the article was implying Samsung and TSMC, ie: Intel has been left behind, the same as GlobalFoundries.
Re:Incompetent ... journalists (Score:4, Insightful)
Intel has a growing number of fabs that do nothing because of their vertical integration. Their older fabs become part time fabs, and then get taken apart entirely.
This is why Intel has to split up, into an Integrated Ciruit design company (like ARM and AMD) and Rent-A-Fab company (like TSMC.)
The former can continue vertical integration, but the later has to be a neutral player ("Pure Play" in the industry - foundry model.)
Re: (Score:2)
And that pertains to incompetent journalists how?
Re: (Score:3)
Re: (Score:2)
It certainly had no meaning in the context of the subject under discussion. So, unless red-herrings are your thing, no valid points.
Re:Incompetent ... journalists (Score:4, Insightful)
Re: (Score:2)
Based on what I know about the history of Commodore and MOS Technologies (later CSG, or Commodore Semiconductor Group), it seems there are lots of people who think that vertical integration is awesome but rarely works out in practice, at least in the long run.
A lot of people don't know this, but a major reason why Commodore fell so far behind was because they insisted on using their own fab to manufacture their custom chips. Their fab made no money selling to other vendors, so it was essentially unprofitab
Re: (Score:2)
Re:Incompetent ... journalists (Score:5, Interesting)
The real limitation to digital CMOS, as has always been the case, is power density. And I'd say Intel is far further along than TSMC at making a great stride forward in that field. Additionally, most yield comparisons are garbage due to the questionable statistics they're based on, all manufacturers keep these numbers under wraps and it is very much design dependent, you also lose a surprising number of chips during singulation and packaging in most instances. We have an inkling of what the yields are for Intel and AMD their CPUs, and a lot of analysis of Intel vs. TSMC are done by fan-boys based on those as a result. But those comparisons are flawed at best, AMD split up their CPU in multiple discrete dies to get around the thermal problems. So AMD's drop-out during packaging is significantly higher, meanwhile Intel will lose more devices during planar processing. But this in no way reflects the technological capability of TSMC vs. Intel in terms of CMOS manufacturing.
Re: (Score:3)
TSMC and Samsung are shipping 5 nm.
Re: (Score:3)
Careful, Samsung's L5PE isn't that much different from 7LPP. TSMC and Samsung are both very careful not to use the letters "nm" in their node names anymore.
Regardless, in terms of performance and transistor density, Intel can't hold a candle to TSMC. N5P in particular is leaps and bounds better than "Intel 7", Intel 10SF, Intel 10nm+, or any other node they're schlepping.
Samsung is improving. 7LPP is somewhat competitive with TSMC N7 and Intel 10nm+. 5LPE is a small but notable improvement. 3GAE goes i
Re: (Score:2)
But where do you actually get this information? Most of the computer/IT enthusiast websites have no clue what they're writing about and take the marketing garbage at face value instead of interpreting what's actually being said, and don't get me started on the gaming news channels. I can only recommend you take the actual PDKs for the latest generations and compare them based on a real digital design. We
Re: (Score:2)
Really
https://fuse.wikichip.org/news... [wikichip.org]
Intel "claims" 106 MTr/mmA^2 but if you do die area vs. transistor comparisons between, say, 10nm+ and TSMC N7P or N7+ (which had density improvements over N7) then you will notice that Intel's claims don't stack up all that well in commercial silicon.
Anyway N5 is far-and-away more dense than 10nm+. 10SF and 10ESF aren't that much better or worse than 10nm+ in terms of overall density. N5 has been in production for awhile, and N5P is in full production now as well.
Re: (Score:2)
As to whether or not you think Samsung is "behind", just take a look at the Snapdragon 888 reviews and see for yourself how well 7LPP worked out for Qualcomm. What a disappointment that chip was. Reference:
https://www.anandtech.com/show... [anandtech.com]
One hopes Snapdragon 898 on 4LPE will work out better.
Re: (Score:2)
Re: (Score:2)
The real limitation to digital CMOS, as has always been the case, is power density. And I'd say Intel is far further along than TSMC at making a great stride forward in that field.
What does that mean? Intel is currently behind there. Until they actually make a stride forward they're just behind.
Re: (Score:2)
Re: Incompetent ... journalists (Score:2)
At Least! (Score:2)
Well, at least they're still good for something!
Intel Inside.
--
Beauty has no boundaries, no rules, no colors. Beauty is like a religion. You can include everything inside it. - Alessandro Michele
Investment (Score:2)
Re: (Score:2)
What makes you think they can?
Just two? (Score:4, Insightful)
Which two are ARS talking about?
Intel, Global Foundries, Samsung, and TSMC?
Or maybe they are talking about technology companies making advanced chips so is it these two:
Intel, AMD, ARM, Apple, Google and the RISC-V founding members?
Even if Ars is talking only about CPUs used in actual computers with desktop operating systems there's more than 2.
Re: (Score:3, Interesting)
What about all the Chinese semiconductor companies?
I don't know if Taiwan's TSMC sells to mainland china and everyone uses them, or there is something else like this: https://www.protocol.com/china... [protocol.com]
Re: (Score:3)
Well I assume they were specifically talking about the upper of the high end, such as CPUs or custom silicon (while somehow I left NVIDIA off the list). By all accounts china's highest end self designed and self fab'd CPU is kind on par with a 5+ year old mobile phone.
But you are right as well, they may be playing catchup but are doing it at an incredible pace.
Re: (Score:2)
Yes, just two. TSMC and Samsung.
Re: (Score:2)
With every season turn! turn! turn! (Score:2)
All about Fed $ (Score:2)
This is just to cash in on Federal $ for Chip Makers. Either this will be another Wisconsin or Intel will sell the plant as soon as they get the Fed $
See: https://www.marketwatch.com/st... [marketwatch.com]
Re: (Score:2)
Intel wants military contracts. Specifically, RAMP-C. They're in in for the long haul.
Canada's Quebec would be great for a chip plant (Score:2)
Dancheezee (Score:1)