Discussion Intel current and future Lakes & Rapids thread

Page 848 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Geddagod

Golden Member
Dec 28, 2021
1,343
1,435
106
This should kill any rumors of Intel using TSMC N3 for uh...anything: https://arstechnica.com/gadgets/202...ons-on-chips-thanks-to-unique-deal-with-tsmc/
No?
The 'Reports' that Apple bought up all of TSMC 3nm manufacturing capacity in the 'short term' (whatever short term means) is ... from Digitimes.
As for exclusivity, aren't the Apple 3nm chips supposed to launch at the end of this year? And ARL at the end of 2024... so ye, exclusivity for ~ a year. And at what point was this exclusivity deal made, before TSMC 3nm was delayed?

Dude, Pat Gelsinger himself said they are using TSMC N3 for uh... something:
The 3nm programs are on track, both that with TSMC as well as our internal Intel 3 programs Granite Rapids and Sierra Forest in particular (Intel Capital Allocation Update conference call)
And here's an SS of an publicly announced Intel roadmap of 'external' N3 being used somewhere in MTL or ARL:
1691447297745.png
Idk why people are so against in believing Intel is going to use TSMC N3 despite Intel making it pretty clear they will be using it. And it's heavily implied that it's going to be for ARL, and in ARL, the only two tiles it would make sense to use N3 on would be CPU or GPU. Rumors say it's the CPU tile that would use N3. This isn't anything so unbelievable, so alien... which is why I find it so baffling that only a couple months ago when we discussed this rumor, some people were acting as if this rumor is akin to some conspiracy theory.
 

SpudLobby

Golden Member
May 18, 2022
1,041
701
106
No?
The 'Reports' that Apple bought up all of TSMC 3nm manufacturing capacity in the 'short term' (whatever short term means) is ... from Digitimes.
As for exclusivity, aren't the Apple 3nm chips supposed to launch at the end of this year? And ARL at the end of 2024... so ye, exclusivity for ~ a year. And at what point was this exclusivity deal made, before TSMC 3nm was delayed?

Dude, Pat Gelsinger himself said they are using TSMC N3 for uh... something:

And here's an SS of an publicly announced Intel roadmap of 'external' N3 being used somewhere in MTL or ARL:
View attachment 84159
Idk why people are so against in believing Intel is going to use TSMC N3 despite Intel making it pretty clear they will be using it. And it's heavily implied that it's going to be for ARL, and in ARL, the only two tiles it would make sense to use N3 on would be CPU or GPU. Rumors say it's the CPU tile that would use N3. This isn't anything so unbelievable, so alien... which is why I find it so baffling that only a couple months ago when we discussed this rumor, some people were acting as if this rumor is akin to some conspiracy theory.
Agreed, the resistance to even Lunar Lake in particular using N3 was comical, because it's arguably more believable there if they want an accelerated release for a premium M sku of ARL (which is in a sense what LNL is apparently).

And I think the material is confusing there because in theory it implies LNL uses an external foundry for a tile and Intel 18A for the other, e.g. it does not seem to imply this is merely an option. It only names Lunar Lake and then says "and beyond" so I think the confusion is warranted but afaict I think they released some new graphic that was a bit more clear.
 

SiliconFly

Golden Member
Mar 10, 2023
1,925
1,284
106
Last edited:

eek2121

Diamond Member
Aug 2, 2005
3,384
5,011
136
Idk why people are so against in believing Intel is going to use TSMC N3 despite Intel making it pretty clear they will be using it. And it's heavily implied that it's going to be for ARL, and in ARL, the only two tiles it would make sense to use N3 on would be CPU or GPU. Rumors say it's the CPU tile that would use N3. This isn't anything so unbelievable, so alien... which is why I find it so baffling that only a couple months ago when we discussed this rumor, some people were acting as if this rumor is akin to some conspiracy theory.
Intel is using TSMC…for the GPU (as they always have) and a few other bits. They have not made any announcement about moving client chips to TSMC for the compute tile. No credible leaker has made such a claim either. If you have evidence to the contrary please post a source.

People are skeptical because outside of a few know-nothing youtubers and 1-2 forum members here, nobody is saying Intel will abandon their internal fabs for TSMC.
 

uzzi38

Platinum Member
Oct 16, 2019
2,746
6,653
146
Intel is using TSMC…for the GPU (as they always have) and a few other bits. They have not made any announcement about moving client chips to TSMC for the compute tile. No credible leaker has made such a claim either. If you have evidence to the contrary please post a source.

People are skeptical because outside of a few know-nothing youtubers and 1-2 forum members here, nobody is saying Intel will abandon their internal fabs for TSMC.
Nobody's saying they're abandoning their internal fabs? Using TSMC for ARL - even the CPU tile - is not the same as abandoning their fabs entirely. The whole point of the optionalities thing is that Intel could use whatever they felt would work best for them, you're blowing things out of proportion.

And besides yeah, turns out they are still using 20A for one of the CPU tiles anyway. So yeah, definitely no abandoning happening.
 

SpudLobby

Golden Member
May 18, 2022
1,041
701
106
Nobody's saying they're abandoning their internal fabs? Using TSMC for ARL - even the CPU tile - is not the same as abandoning their fabs entirely. The whole point of the optionalities thing is that Intel could use whatever they felt would work best for them, you're blowing things out of proportion.

And besides yeah, turns out they are still using 20A for one of the CPU tiles anyway. So yeah, definitely no abandoning happening.
It's borderline insanely tiresome how absurd people are about Intel's fabs in either direction at this point. I'm not claiming it's wise to be arbitrarily neutral as some midpoint between pessimism or wild optimism - it is what it is, no reason to subscribe to perfect balance and it depends on what, exactly, we're measuring/forecasting - yet it's still clear when commentary on this just hinges on abstract corporate tribalism, very minimal understanding of hardware much less the economics of foundries, etc.

For ex: the above or like the excess optimism about 18A and 20A being "manufacturing ready" or ignorance about the sheer volume of dice TSMC ships on the other end with early production, comparing Intel's HP libraries to TSMC's HP libraries, comparing peak densities alone, stuff like that.
 

FangBLade

Senior member
Apr 13, 2022
203
399
106
Trolling is not permitted. Please read and adhere to the forum rules.

Henry swagger

Senior member
Feb 9, 2022
509
312
106
Personal attacks are not allowed. Attack the content, not the person writing it.
Trol
Impressive by Intel's standards, but not so much for more successful companies like Nvidia, AMD, and Apple. When it comes to Intel, we're satisfied even with a 1% performance improvement, as long as they stay on the same motherboard, lol.
Troll lol
 
  • Like
Reactions: reb0rn

Geddagod

Golden Member
Dec 28, 2021
1,343
1,435
106
Intel is using TSMC…for the GPU (as they always have) and a few other bits. They have not made any announcement about moving client chips to TSMC for the compute tile. No credible leaker has made such a claim either. If you have evidence to the contrary please post a source.

People are skeptical because outside of a few know-nothing youtubers and 1-2 forum members here, nobody is saying Intel will abandon their internal fabs for TSMC.
Raichu, Bionic, Xino, even Exist50 here on this forum
 

SiliconFly

Golden Member
Mar 10, 2023
1,925
1,284
106
Raichu, Bionic, Xino, even Exist50 here on this forum
True. They just can't abandon their own foundry (20A) for TSMC. If they use TSMC for ARL cpu tile, that would be direct admission of failure which would spell disaster for IFS. But I don't think it's gonna happen considering Pat has repeatedly said that the nodes are progressing well and are even ahead of schedule!

And considering Intel 4 is clocking well in it's first iteration itself, I think they're doing a fairly good job with their nodes. There's absolutely no necessity to use TSMC for their client cpu tile.
 

H433x0n

Golden Member
Mar 15, 2023
1,224
1,606
106
He s not in the enginering teamsn but for sure that Intel want to extract some infos about Zen 5 expected perfs and subsequent designs, a desperte move to catch up by using some kind of industrial spying...
I think you’re reading too far into this.

You’d be amazed how many engineers jump from AMD <-> Intel <-> Nvidia. They don’t need some marketing guy for insider info.
 

Saylick

Diamond Member
Sep 10, 2012
3,923
9,142
136
I think you’re reading too far into this.

You’d be amazed how many engineers jump from AMD <-> Intel <-> Nvidia. They don’t need some marketing guy for insider info.
Agreed. Even if they get insider info, it's already too late for Intel to do anything about Zen 5 at this point. All they can tweak is clocks and prices.
 
  • Like
Reactions: Tlh97 and Geddagod

Thunder 57

Diamond Member
Aug 19, 2007
3,805
6,407
136
True. They just can't abandon their own foundry (20A) for TSMC. If they use TSMC for ARL cpu tile, that would be direct admission of failure which would spell disaster for IFS. But I don't think it's gonna happen considering Pat has repeatedly said that the nodes are progressing well and are even ahead of schedule!

And considering Intel 4 is clocking well in it's first iteration itself, I think they're doing a fairly good job with their nodes. There's absolutely no necessity to use TSMC for their client cpu tile.

What's the standard bet here? If you are wrong you eat a can of cat food? Intel will not use TSMC for the CPU tile. If they did it would be a huge admission of failure.
 

mikk

Diamond Member
May 15, 2012
4,291
2,379
136
And I think the material is confusing there because in theory it implies LNL uses an external foundry for a tile and Intel 18A for the other, e.g. it does not seem to imply this is merely an option. It only names Lunar Lake and then says "and beyond" so I think the confusion is warranted but afaict I think they released some new graphic that was a bit more clear.

18A refers to the successor of LNL and also the slide refers to the CPU tile, you can see it on Meteor Lake and by the fact there are no GPU tile nodes included like TSMC N5 and N4 for MTL+ARL. eek2121 is just wrong on this.

This is what the slide says when it comes to the CPU tile:

Meteor Lake= Intel 4
Arrow Lake= Intel 20A+TSMC N3

Lunar Lake= External
Beyond= 18A
 

Ajay

Lifer
Jan 8, 2001
16,094
8,112
136
18A refers to the successor of LNL and also the slide refers to the CPU tile, you can see it on Meteor Lake and by the fact there are no GPU tile nodes included like TSMC N5 and N4 for MTL+ARL. eek2121 is just wrong on this.

This is what the slide says when it comes to the CPU tile:

Meteor Lake= Intel 4
Arrow Lake= Intel 20A+TSMC N3

Lunar Lake= External
Beyond= 18A
Oh, I missed that LNL is external (TSMC?). Hmm, I thought that was Intel's next gen arch designed to take on latter AM5 CPUs (desktop).