Discussion Intel current and future Lakes & Rapids thread

[jpiniero]

N3 was never the plan for MTL, but I look forward to the rumor mongers attempting to claim otherwise when it launches. To cover their failed predictions, if nothing else.

So what do you think the IGP is then, N4?

 

[Exist50]

So what do you think the IGP is then, N4?

N5 or N4, yes.

 

[repoman27]

N5 or N4, yes.

So what was Intel planning on using N3 for if not the MTL GPU tile? We obviously know ARL was originally planning on using it for the GT3 GPU and 6+8 CPU, but that wasn't slated to hit volume production until the last week of April 2023.

 

[Doug S]

A16 is supposed to be N3 node.[/URL]

iPhone 14 coming in four models without ‘mini’ version, more - 9to5Mac

According to 9to5Mac sources, iPhone 14 will be available in 6.1-inch and 6.7-inch versions, and the "mini" version will be discontinued.

9to5mac.com

The articles you link state specifically A16 will not be on N3. I'm a bit skeptical of the claim it will be on N5P, as there seems to be no reason for Apple to use that when N4 is available but Kuo is one of the more accurate Apple rumor guys so who knows.

 

[mikk]

N3 was never the plan for MTL, but I look forward to the rumor mongers attempting to claim otherwise when it launches. To cover their failed predictions, if nothing else.

Didn't Intel say MTL iGPU will be on TSMC 3nm?

 

[JasonLD]

Didn't Intel say MTL iGPU will be on TSMC 3nm?

Even their own slide said N3.

 

[Exist50]

So what was Intel planning on using N3 for if not the MTL GPU tile? We obviously know ARL was originally planning on using it for the GT3 GPU and 6+8 CPU, but that wasn't slated to hit volume production until the last week of April 2023.

Well that's a major flaw in the rumor to begin with. But why limit the scope to client? Their server roadmap has also slipped. Who's to say they didn't intend for something to use N3? Or as you point out, ARL originally intended to use N3 in '23, but is now a '24 product. Would that not cover most of what the rumor claims? And that's if you buy into the premise to begin with.

Even their own slide said N3.

Intel has never said what process MTL is on besides the CPU tile. Where did you see N3?

 

[repoman27]

This is based on ASML Financial reporting, # of EUV systems per year :

2015-2017	16​
2018	18​
2019	26​	TSMC has half of all EUV machines (statement in Aug 2020) = 30
2020	31​
2021	32​	Based on 2x first half 2021
Total	123​

Even if TSMC had bought every single EUV machine after their statement in 2020 of having half of the ones made to that point, they'd still only have about 90. A lot more likely they have about 60-70 right now. The earlier machines cannot do below 5nm (TSMC N7/N5/N4).

TSMC would need a 3600D for 3nm, and ASML delivered its first one in mid 2021. Earlier models were 7 / 5nm (TSMC N7 / N5 / N4).

Intel also purchased a number of these earlier units and reported have at least a couple of them up and running in 2019, but yields were not as good as expected. In 2015 Intel had ordered 15 EUV systems from ASML, which is half of what was produced up until 2019.

If we assume ASML is producing the 3600D as fast as other models were produced at 32/year starting mid 2021, then there would at most be 30-35 in existence right now. TSMCs goal was to get 60 systems shipped by 2023.

I would bet there is a 40/40/20 split between TSMC / Intel / Samsung on the new models. That would be like 12/12/8 3nm capable machines each.

This means you've got way too many credited to TSMC for the N3 node.

This should not be surprising given we know TSMC delayed Apple in getting 3nm last year and this year only the top iPhone models will have the 3nm A16, the rest will continue to use the 5nm A15.

The $340M EUV systems Intel recently ordered (Jan) were the 5200 model, which doesn't exist as a shipping system yet. That is for 2024/2025.

Quarter	EUV systems recognized in sales
Q4'13	1
Q1'14	1
Q2'14	1
Q3'14	2
Q4'14	1
Q1'15	0
Q2'15	1
Q3'15	0
Q4'15	0
Q1'16	0
Q2'16	2
Q3'16	1
Q4'16	1
Q1'17	0
Q2'17	2
Q3'17	4
Q4'17	5
Q1'18	1
Q2'18	7
Q3'18	5
Q4'18	5
Q1'19	4
Q2'19	7
Q3'19	7
Q4'19	8
Q1'20	2
Q2'20	7
Q3'20	14
Q4'20	8
Q1'21	7
Q2'21	9
Q3'21	15
Q4'21	11
Q1'22	3
Q2'22	12
Total:	154

You can feel free to go through the quarterly reports yourself to double check my numbers, but that's where I got them. Systems sold prior to 2017 were NXE:3300B and NXE:3350B prototype models. However, most, if not all of them, have since been field upgraded to mass production systems. Likewise many NXE:3400B models have had overlay and throughput upgrades to make them closer in capability to the NXE:3400C.

Furthermore, while older machines may not be able to resolve the finest features or achieve the tightest pitches, they're perfectly capable of handling many of the metal layers where EUV is used, even for N3. Also, ASML EUV installs (when they're recognized as sales) tend to be heavily weighted towards the back half of the year. As far as my tables were concerned, the high-NA machines aren't even in the picture, because I was focusing on the Meteor Lake timeframe in particular.

And seeing as TSMC has told us that commercial wafer-outs on N3 won't happen until Q1'23, I'm pretty sure the A16 won't be on N3.

edit: added Q2'22 shipments to table

 

[JasonLD]

Well that's a major flaw in the rumor to begin with. But why limit the scope to client? Their server roadmap has also slipped. Who's to say they didn't intend for something to use N3? Or as you point out, ARL originally intended to use N3 in '23, but is now a '24 product. Would that not cover most of what the rumor claims? And that's if you buy into the premise to begin with.


Intel has never said what process MTL is on besides the CPU tile. Where did you see N3?

From their own slide.

 

[Exist50]

View attachment 64790

From their own slide.

That says "Meteor Lake & Arrow Lake". So unless you think Meteor Lake uses 20A...

 

[JasonLD]

That says "Meteor Lake & Arrow Lake". So unless you think Meteor Lake uses 20A...

Hard to say if they are going to use N3 on both Meteor Lake and Arrow Lake or just Arrow Lake just going by the slide. I wouldn't completely rule out N3 on Meteor Lake.

 

[Exist50]

Hard to say if they are going to use N3 on both Meteor Lake and Arrow Lake or just Arrow Lake just going by the slide. I wouldn't completely rule out N3 on Meteor Lake.

That's your speculation, fine, but that's very different than Intel confirming it. The only process-related detail that Intel's officially confirmed is that the compute tile will be on Intel 4.

 

[IntelUser2000]

That says "Meteor Lake & Arrow Lake". So unless you think Meteor Lake uses 20A...

Okay then the part I don't get is why it's only saying "External N3" rather than N4+N3?

Also is there a reason the ordering is changed on Lunar Lake? Why External fist rather than internal?

 

[uzzi38]

N3 was never the plan for MTL, but I look forward to the rumor mongers attempting to claim otherwise when it launches. To cover their failed predictions, if nothing else.

I have no clue if it ever was or wasn't - that part of what I wrote was honestly assuming that the former rumours were correct, but I can definitely agree that right now it's not what OEMs think is coming right now.

 

[Exist50]

Okay then the part I don't get is why it's only saying "External N3" rather than N4+N3?

Also is there a reason the ordering is changed on Lunar Lake? Why External fist rather than internal?

Keep in mind that the 3rd section is "Lunar Lake & Beyond". If you go by rumors, which I happen to believe, then Lunar Lake is on N3. Then probably Panther Lake on 18A later. Contrast this with Arrow Lake, where they appear to be dual sourcing between N3 and 20A. But in general, I don't think Intel's slides merit that degree of inspection. They've been terribly sloppy with marketing lately.

 

[moinmoin]

But in general, I don't think Intel's slides merit that degree of inspection. They've been terribly sloppy with marketing lately.

I disagree, all of Intel's design innovation were put into creating all these sleek shiny marketing slides as of late. /s

 

[Henry swagger]

Intel Meteor Lake's "Redwood Cove" and "Crestmont" architectures confirmed - VideoCardz.com

Meteor Lake = Redwood Cove + Crestmont A confirmation on the microarchitecture used by Meteor Lake CPU series. The rumors about Redwood Cove architecture are dated as far back as November 2020, when it was first mentioned by Moore’s Law is Dead. This rumor is finally finding its confirmation...

videocardz.com

I think xe lpg uses battlemage for meteor lake gfx

 

[jpiniero]

https://twitter.com/x/status/1550101506426376192

Looks like Intel is moving Xeon W to their favorite numbering scheme. The W9 3495 is 56/112.

 

[Heartbreaker]

Intel Meteor Lake's "Redwood Cove" and "Crestmont" architectures confirmed - VideoCardz.com

Meteor Lake = Redwood Cove + Crestmont A confirmation on the microarchitecture used by Meteor Lake CPU series. The rumors about Redwood Cove architecture are dated as far back as November 2020, when it was first mentioned by Moore’s Law is Dead. This rumor is finally finding its confirmation...

videocardz.com

I think xe lpg uses battlemage for meteor lake gfx

Anything known about Crestmont E-Cores. It seems like Intel needs it's lower power cores to support AVX-512 (even though that uses a lot of power) so they can stop disabling AVX-512 on the their P-cores, so they should be working toward AVX-512 on E-Cores.

 

[jpiniero]

Anything known about Crestmont E-Cores. It seems like Intel needs it's lower power cores to support AVX-512 (even though that uses a lot of power) so they can stop disabling AVX-512 on the their P-cores, so they should be working toward AVX-512 on E-Cores.

https://images.anandtech.com/doci/13699/1-Roadmap.jpg

Course there's no guarantee that the "Next Month" is even Crestmont.

 

[nicalandia]

https://twitter.com/x/status/1550101506426376192

Looks like Intel is moving Xeon W to their favorite numbering scheme. The W9 3495 is 56/112.

Xeon W is for Workstations, that 3495 has to be based on SPR Quad Tile System instead of the yet to be seen Monolithic die.

 

[jpiniero]

instead of the yet to be seen Monolithic die.

I still don't think it exists, at least for big boy 8 channel. It'd be too big for 10 nm. Doesn't help that Golden Cove Server is very much a chonker.

I also still don't think it even makes sense for 4 channel HEDT because you should be able to get enough salvage to handle the low volume. But that would at least be realistic if you don't get too crazy with the core count.

 

[Henry swagger]

Intel 13th Gen Raptor Lake CPUs & Z790 Motherboards Rumored To Launch on 17th October, H770 & B760 Motherboards at CES 2023

Intel 13th Gen Raptor Lake Desktop CPUs & Z790 motherboards are rumored to launch on 17th October followed by mainstream SKUs at CES 2023.

wccftech.com

Intel review embargo will be weeks early before the release.. they must be very confident 😁

 

[dullard]

Intel 13th Gen Raptor Lake CPUs & Z790 Motherboards Rumored To Launch on 17th October, H770 & B760 Motherboards at CES 2023

Intel 13th Gen Raptor Lake Desktop CPUs & Z790 motherboards are rumored to launch on 17th October followed by mainstream SKUs at CES 2023.

wccftech.com

Intel review embargo will be weeks early before the release.. they must be very confident 😁

Intel review embargos have often been weeks before the actual release. This is true even when the product is nothing special.

 

[FangBLade]

Intel review embargos have often been weeks before the actual release. This is true even when the product is nothing special.

Like RKL, it was shat from the day 1, but many Intel supporters hyped it like nothing before.