Intel Tries To Get Its Chip Manufacturing Back on Track With 'Intel 4,' Due in 2023 (arstechnica.com) 66
Intel's chip manufacturing technology has been outpaced by rivals like TSMC and Samsung in recent years, but the company is looking to put its troubles behind it. From a report: The first step forward will be the Intel 4 manufacturing process, which Intel has shared more details about at the Institute of Electrical and Electronics Engineers' annual VLSI Technology Symposium (as reported by AnandTech and Tom's Hardware). The new manufacturing tech is on track to be used in consumer chips starting in 2023, starting with Intel's "Meteor Lake" CPU architecture. Meteor Lake will likely come to market as Intel's 14th-generation Core CPU sometime next year. Intel 4's biggest improvement is its integration of extreme ultraviolet (EUV) lithography, which uses short-wavelength ultraviolet light to etch tiny patterns into silicon wafers. TSMC and Samsung use EUV technology in their most advanced manufacturing processes. Intel says that compared to the Intel 7 process, Intel 4 will enable either 21.5 percent better clock speeds using the same amount of power or the same speeds using 40 percent less power.
After Intel 4, Intel will move on to Intel 3, which is a higher-density iteration of Intel 4 using the same EUV technology. Notably, chipmakers will be able to port designs made for Intel 4 directly to Intel 3 without having to make changes, which will hopefully allow both Intel and third-party chip designers to start using it quickly (Intel 3 will be offered to third parties through Intel Foundry Services). By making smaller jumps between process technologies -- introducing EUV lithography in Intel 4 and then optimizing for maximum density in Intel 3, rather than trying to do both at once -- Intel hopes to avoid the delays and yield problems that held the 10nm/Intel 7 process back for so many years.
After Intel 4, Intel will move on to Intel 3, which is a higher-density iteration of Intel 4 using the same EUV technology. Notably, chipmakers will be able to port designs made for Intel 4 directly to Intel 3 without having to make changes, which will hopefully allow both Intel and third-party chip designers to start using it quickly (Intel 3 will be offered to third parties through Intel Foundry Services). By making smaller jumps between process technologies -- introducing EUV lithography in Intel 4 and then optimizing for maximum density in Intel 3, rather than trying to do both at once -- Intel hopes to avoid the delays and yield problems that held the 10nm/Intel 7 process back for so many years.
Too late (Score:5, Informative)
It's nice that Intel dished on their 7nm node (renamed to Intel 4), but it's already horrifically late. Intel has moved every future product intended for the node off Intel 4 except for Meteor Lake and . . . Loihi 2?
Notably Ponte Vecchio isn't using Intel 4 tiles anymore. Granite Rapids also got pushed back and on to a different node. In fact, Intel doesn't have any future enterprise CPU products on Intel 4, which ought to tell you something.
Re: (Score:1)
Re: (Score:1)
Well, Intel has gotten fat and lazy on scamming its customer with over-priced crappy products. There is a price to pay for that eventually.
Now all we need is a real Microsoft competitor...
Re: Too late (Score:1, Funny)
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
So like 20% of Linux devices then? Cope.
Re: (Score:2)
Re: (Score:2)
PCs are not the only devices. Far from it ignoramus. PCs are in fact a small minority of devices. Cope.
Re: (Score:2)
Re: Too late (Score:2)
Or option three: they have no idea what they're talking about (seems most likely from the rest of their comment).
Re: (Score:2)
That's not really true. Unless you're talking about AMD CPUs from 2016 and earlier, which who knows, maybe you are?
Also you must be using some bargain-bin NVMe drives if you're getting speeds that slow. Aren't you also aware that spinning rust relies heavily on cache?
Re: Too late (Score:5, Insightful)
Re: (Score:2)
AMD pivoted to fabless and tried different technologies until they found some that stuck, like chiplets. They weren't complacent, they were busy changing direction.
Re: (Score:2)
AMD pivoted to fabless and tried different technologies until they found some that stuck, like chiplets. They weren't complacent, they were busy changing direction.
No. They were complacent and got their arses handed to them. They *then* did all you just said to recover from their poor position and to great effect.
Re: (Score:2)
Hey bellycrawler, AMD's main issue was illegal trustmaking activity by slithery Intel. You fit right in with those creeps.
Karma's a bitch huh?
Re: (Score:2)
You kind of forgot about Hector Ruiz, didn't you? He wanted to ride K8 for awhile longer. He was totally unprepared for Conroe and everything that came with it. Plus he engineered the Globalfoundries spinoff which made AMD a vassal of that company for quite some time, and ultimately had a negative effect on AMD. AMD wasn't able to switch to TSMC's superior nodes for its CPUs until 2017 on a commercial level.
And let's be honest, they also struggled with designs after it became clear that riding the K8 tra
Re: (Score:2)
Phenom was a flop
You are full of shit. There was nothing wrong with Phenom, I have a 6 core running as a game machine to this day (entirely adequate CPU power for casual gaming). It was Intel's illegal trust making that starved AMD of the operating funds it needed to capitalize on it. Now go slither back under your fucking slimy revisionist rock.
Re: (Score:2)
Perhaps, but AMD seem to be pushing hard themselves at the moment so it's going to be difficult for Intel to keep up.
There is still plenty of ground for AMD to take too, especially in the laptop market. For some reason there just aren't all that many AMD models compared to Intel ones, but that's changing.
Re: (Score:2)
Intel has a manufacturing problem. One that might be gone by 2025. We'll see.
Intel ME (Score:2)
Until Intel kills ME:
https://en.wikipedia.org/wiki/Intel_Management_Engine
I do not care what they do. I am on an old system that in theory ME can be disabled, but newer systems I heard disabling it is impossible.
To Intel: As soon as an affordable system comes you with Open RISC or another Open Chip, I will never by a brand new PC/Laptop.
Re: (Score:3)
Did you have a look at:
https://en.wikipedia.org/wiki/... [wikipedia.org]
https://www.raptorcs.com/TALOS... [raptorcs.com]
Re: (Score:3)
Re: (Score:2)
It's a power9 board+cpu starting at $4,219.99, but if you want to use all the expansion slots you need two CPUs and that starts at $4,813.27.
Re: (Score:2)
POWER9 has IPMI too, and some known vulnerabilities.
https://cve.mitre.org/cgi-bin/... [mitre.org]
https://www.ibm.com/docs/en/po... [ibm.com]
Re: (Score:3)
Until Intel kills ME:
https://en.wikipedia.org/wiki/Intel_Management_Engine
I do not care what they do. I am on an old system that in theory ME can be disabled, but newer systems I heard disabling it is impossible.
To Intel: As soon as an affordable system comes you with Open RISC or another Open Chip, I will never by a brand new PC/Laptop.
So, you are planning on using Non-X86/X64 architectures going forward?
I mean, AMD has a Security Management Processor, which does much the same thing. While current VIA X-64 processors do not have a ME analogue, their performance is (let's be diplomatic here) severely lacking for desktop or laptop use in 2022 and forward, and any new revisions will probably incorporate some sort of ME.
M1 silicon has a secure enclave, ARM processors of a performance level bigger than a Pi4 also have something similar.
Power10
Re:Intel ME (Score:4, Informative)
ARM processors of a performance level bigger than a Pi4 also have something similar.
You're thinking of TrustZone, and it exists in the Pi4, and Arms much smaller than it, and has for many years.
It isn't directly analogous to the Intel ME or AMD PSP, in that both of those will only ever run their code.
TrustZone is designed to be used by the user.
I used it extensively in some embedded work I was doing in the late 2000s.
Re: (Score:1)
Must be a pretty old system then, ME has been in all Core products since Core2Duo and unless you wrote a key, can be managed by whoever gets physical access to your device.
Where to? (Score:3)
Intel 4, Intel 3, Intel 2, Intel 1, Intel 0, Intel ?, ...
Re:Where to? (Score:4, Informative)
Intel 4, Intel 3, Intel 2, Intel 1, Intel 0, Intel ?, ...
Nope, after intel 3, there comes intel 20A. Since slashdot does not have UTF-8, i can not use the propper symbol for Angstrom here, but 1A=0.1nm, that unit is commonly used in optics and physics, but is not an SI unit.
Re: (Score:2)
Re: (Score:2)
Intel Blast Off.
Think Boeing... (Score:3)
Re: (Score:2)
Good engineering requires focusing roughly a decade down the road. But current in-style "ROI" financial models make that too far out. This is partly because the models are tuned for "typical" corporations which may sell skirts or soda.
Making soda is obviously different than making computer chips, but some investors either don't know the difference or would rather the company be milked and gutted such that the down-sides of pump-and-dump happen AFTER they sell their shares: squeeze-and-run. I'm not sure what
Re: (Score:2)
An engineer is back as the CEO and has been making massive internal changes for like a year and a half now. [intel.com]
He's part of the reason I'm still long $INTC. There are a number of others.
Re: (Score:2)
show me (Score:3)
Re: (Score:3)
Perhaps this time they won't take that the proverbial "left turn at Albuquerque" from the rest of the industry.
Re: (Score:2)
Surely TSMC won't allow Intel to do that, given that it would cause AMD and Nvidia to invest serious money with other fabs and make a long term move away from TSMC.
It would be crazy to screw over two of their biggest customers in favour of a third one that is actively building a new fab so they can stop giving TSMC business.
Re: (Score:2)
I don't think it would happen, but TSMC might be all too happy to give Intel a better outsourcing experience than they are getting out of their own fabs. It wouldn't be enabling a competitor, but embarrassing the part of Intel that they compete with in front of the part of Intel they would love to have as a client.
Eventually, the higher ups at Intel might decide their in-house fab team just isn't going to do it and, like AMD did before it, offload it and just become mostly a TSMC customer.
The longer that In
Re: (Score:2)
Re: (Score:2)
Biggest joke about Intel is that ticktock had something to do with their success. As witness ticktock man Gellsinger parachuting in to wake up fading virtualization vendor VMWare but just sinking it further. Ticktock outed as just just another bullshit cargo cult.
Now illegal trust making activity... yeah, that's been worth a $trillion to Intel.
Re: (Score:2)
Re: (Score:2)
Who cares? (Score:2)
Re: (Score:3)
Re: (Score:2)
I presume commenter's point is that investment toward more advanced processes are wasted due to bottleneck of throughput of still relevant older processes. I presume the commenter would advocate that the resources spent going to Intel 4 would be better used to expand older process capacity, though financially I think that wouldn't pan out too well.
Re: (Score:2)
It's been a long time since Intel made MSI scale chips so anything they spend has no bearing on discrete logic or opamps.
I'm not aware of anyone that uses larger than 6" wafers for opamps and such, but I haven't looked into it in years. Analog ICs/opamps use a totally different layer stack and doping profiles than logic devices as well. It's not like Intel could just switch out the masks and crank out 741 opamps all day.
Will.i.am and Gaga on an Orange County Chopper (Score:2)
It's good to see all of those partnerships with media celebrities finally paying off,
but the way Intel marketing chases trends, they will be announcing an
Intel electric car on their new free speech social media site later this year.
Intel's Long Game Strategy (Score:2)
EUV ASML (Score:3)
"biggest improvement is its integration of extreme ultraviolet (EUV) lithography" finally, which TSMC and Samsung have been working with for quite some time. And I assume this means they will be using ASML gear as well, unless there's some other EUV option out there I haven't heard of. ASML can make only so much of their extremely complicated (and gigantically expensive) stuff, and these days that is a considerable bottleneck.
Re: (Score:2)
Intel does not own ASML.
Re: (Score:2)
Re: (Score:3)
Intel is the first to get the next gen EUV machines (called NA EUV)
https://www.anandtech.com/show... [anandtech.com]
https://wccftech.com/intel-sha... [wccftech.com]
I have been reading that they will have a year's head start with the new machines, compared to TSMC, Samsung, etc.
So yes, they will get the new NA EUV machines before anyone else. Now the question is, will they be able to "tune" them to the max performance before the other fabs get theirs.
Re: (Score:3, Interesting)
From https://www.nextplatform.com/2... [nextplatform.com]
"Clearly, the G
Re: (Score:2)
Re: (Score:2)
Don't know how much they still own, haven't seen any news of them selling it off.
Re: (Score:2)
Intel owns ASML so I am sure they will get priority.
No they do not. Intel is just a customer. They have no priority and join a queue like everyone else.
Fortunately though for Intel they were the *first* customer to order High NA EUV equipment from ASML.
Re: (Score:2)
It's possible that Intel owns some ASML stock.
https://finance.yahoo.com/news... [yahoo.com]
"TSMC eclipsed its rivals by integrating ASML's EUV machines first in the late 2010s - a mistake Intel CEO Pat Gelsinger has vowed not to make again with High-NA."
"ASML controls more than 90% of the lithography market. No competitor is attempting to build an EUV system, citing high development costs."
For all the people ragging on Intel (Score:1)
As much as
Re: (Score:2)
That is why TSMC has to be persuaded to build and staff a major fab in the US, helped by US taxpayer money. If TSMC can continue its business from the US after China invades, not only will that save its customers, but it will remove one of the major incentives for China to invade Taiwan.
Think of it like a colony from Earth on another planet to escape the coming apocalypse.
Really? (Score:2)
Kid Slashdot finally got over the hurdle of designing the Intel even-numbered processors.... so what are they going to do next?