12nm is technically smaller then 14nm lol

I'm not sure that is an accurate statement

GlobalFoundries, TSMC and Samsung have all traditionally used marketing numbers for node size, whereas Intel has tended to be more conservative and closer to ITRS guidelines.

GlobalFoundries 12nm process is probably similar to Intel's 14nm process, maybe even a little larger.

Intel's 10nm process - if it had been successful - would likely have been comparable in size to TSMC's 7nm process.


Which begs the question, why would Intel even want GloFo? Aren't they widely considered a second tier also-ran of a player? Maybe they hope to learn from GloFo's abandoned 7nm process?

But buying a failed process to help fix your failed process doesn't seem like it would be a winning strategy...
 
Last edited:
The article mentioned it already. Intel doesn't really have a contract fab business while GlobalFoundries is all contract fab. It is likely not trivial for Intel to use its existing fab capacity for contract manufacturing.
 
damn, Intel and GloFo? Didn't see that coming.
Makes sense tho. Intel could focus their fabs on leading processes while having GloFo fabs as a less expensive option for things what don't require the densest litho.
 
Zarathustra[H] said:
Which begs the question, why would Intel even want GloFo? Aren't they widely considered a second tier also-ran of a player? Maybe they hope to learn from GloFo's abandoned 7nm process?

But buying a failed process to help fix your failed process doesn't seem like it would be a winning strategy...
Click to expand...

It's totally not about that.

What Intel would plausibly want it for is get GloFo's contract fab business. They've stated a desire to expand into that segment, and buying GF would get them in the door in a big way. From there they'd be in a much better spot to court the subset of GF's current customers who are interested in upgrading to newer, but still behind leading edge, processes. The customers would still have to do a fair amount of work to transition from GF to Intel design rules; but the same would be true if switching to TSMC or Samsung.
 
DanNeely said:
It's totally not about that.

What Intel would plausibly want it for is get GloFo's contract fab business. They've stated a desire to expand into that segment, and buying GF would get them in the door in a big way. From there they'd be in a much better spot to court the subset of GF's current customers who are interested in upgrading to newer, but still behind leading edge, processes. The customers would still have to do a fair amount of work to transition from GF to Intel design rules; but the same would be true if switching to TSMC or Samsung.
Click to expand...

That makes sense. So they are really in it to acquire the customer base, not necessarily the technology or the fabs.
 
Youn said:
it's not about length, it's about girth
Click to expand...
AMD tries to cheat with its condom.
1626458050487.png
 
Interesting rumor. Title would have been better named "Intel rumored buyout of GloFo"

Aside from what has been mentioned about existing contracts, this would potentially steal AMD's older chip/IoT/backup fab from AMD. A good snooker move by Intel.

Additionally, there are a lot of things that go into improving chip design processes. Even though these are older processes, there may be things that intel does not already know and could use to help develop smaller nodes. Plus the infusion of new ideas from GloFo personnel.

Also, not sure, but buying a private company may not have to go through the regulatory requirements of a corporation, so if GloFo is rumored to do an IPO, intel better act fast.

Will be interesting to see if they do it. They have to start doing something with that dwindling supply of profit to get back on top.
 
Zarathustra[H] said:
That makes sense. So they are really in it to acquire the customer base, not necessarily the technology or the fabs.
Click to expand...

Honestly I think it is all 3. If they weren't trying to expand into contract fabbing the deal would never happen; but if they buy GFs current customers, they need GFs current fabs to produce chips GFs current processes for those customers. Having access to all of GF's old processes also gives them a lot more scope to sell to customers wanting to make simple chips on cheap processes.

Intel hasn't kept significant amounts of older process capacity around; their current fabs are currently a mix of 22nm to 5nm processes on 300mm wafers with 22/14nm and 14/10/7(/5)nm being the most common configurations; they have a single 32/45nm fab and a single 65nm 200mm wafer one.

GF has 3x 300mm fabs, ones doing 14nm, one doing 48/28/22/12nm, and one doing various 40-130nm processes. They also have a pair of 200mm fabs, one doing 350-180 nm processes and one doing 600-350nm ones. They had several others but have since sold off or merged them into other facilities.

What I don't think is driving any part of this potential deal would be access to the 7nm that GF was working on. GF stopped work on it several years ago; so there'd be an extended re-startup period; and this deal would take a long time to get regulatory signoff (assuming they'd sign off on it at all); so it'd probably be the mid 2020's before a GF 7nm process could be available. At that point it would be another legacy process in the collection; the only reason I could see Intel doing that would be if they had major GF 12/14nm customers who were only willing to stay with their current fab if they got a newer process that was easier to migrate to than one of Intel in house ones would be. But if there was an economic case for that now, there probably would've been a much stronger one several years ago when GF threw in the towel. And of course, there's the question of if GF had a clear path to a production 7nm process but just couldn't sell enough to cover the costs; or if blaming it on that was just a face saving way to cover up failure.

https://en.wikipedia.org/wiki/GlobalFoundries#300_mm_fabrication_facilities
https://en.wikipedia.org/wiki/List_of_Intel_manufacturing_sites
 
Zarathustra[H] said:
Intel's 10nm process - if it had been successful - would likely have been comparable in size to TSMC's 7nm process.
Click to expand...
FYI while it wasn't a success in terms of coming out when they wanted it, it is not something that doesn't exist. The i7-11800H is an 8 core mobile CPU that is 10nm, and available in systems now. Ends up being competitive performance wise with Zen 3 mobile CPUs, but is more power hungry. It's really Intel 10nm though, and is really available in products in consumer hands.
 
Sycraft said:
YI while it wasn't a success in terms of coming out when they wanted it, it is not something that doesn't exist. The i7-11800H is an 8 core mobile CPU that is 10nm, and available in systems now. Ends up being competitive performance wise with Zen 3 mobile CPUs, but is more power hungry. It's really Intel 10nm though, and is really available in products in consumer hands.
Click to expand...

Yeah, they have released a few, but the CPU's it yields still aren't of the quality the original plan called for back in 2016...

So I still wouldn't call the process "successful".
 
Sycraft said:
YI while it wasn't a success in terms of coming out when they wanted it, it is not something that doesn't exist. The i7-11800H is an 8 core mobile CPU that is 10nm, and available in systems now. Ends up being competitive performance wise with Zen 3 mobile CPUs, but is more power hungry. It's really Intel 10nm though, and is really available in products in consumer hands.
Click to expand...

Yeah, they have released a few, but the CPU's it yields still aren't of the quality the original plan called for back in 2016...

So I still wouldn't call the process "successful".
 
I mean depends on what you mean by "successful". The implication of the 12nm headline, and your talking about the node as though it was still theoretical would imply it had failed to materialize. However that's not the case, it is in mass production right now. It failed to be at the level they wanted when they wanted, it wasn't a business success, but it is real and can be compared against other real, in production, nodes.
 
Zarathustra[H] said:
Yeah, they have released a few, but the CPU's it yields still aren't of the quality the original plan called for back in 2016...

So I still wouldn't call the process "successful".
Click to expand...
But from what I understand of Intel’s processes they are also now using the designs they had originally planned to launch with the 10nm process back in 2016.

2022 is when we should start seeing their new architecture’s on the 10 and 7nm processes. As well as on TSMC’s 5 and 3nm nodes as well. Supposedly Intel has a good lineup of mobile and HEDT parts itching to go on TSMC 3nm.
 
You must log in or register to reply here.
Back
Top