Discussion Intel current and future Lakes & Rapids thread

[IntelUser2000]

Oops! Good catch, yes I was referring to Intel nodes. We need a new abbreviation system IN4 for Intel, SN4 for Samsung and TN4 for TSMC?? IDK…

Remember i486 on 486 CPUs?

i4, t4, s4?

 

[uzzi38]

I really don't think on-die PCH is going to be the panacea you're hoping for. I mean, it's nice to have, sure, but I would be pleasantly surprised if Meteor Lake has radically better battery life than Tiger Lake (using as baseline in light of the Alder Lake regression). Lunar Lake would hopefully show more of the setup's potential, but if not that, assuredly Nova Lake.

Yeah I have to agree with this. I had really high expectations for bringing the PCH on package for ADL-P/ADl-H (however you want to call it) vs TGL-H, and even now I'd imagine that the power savings from moving it on package would dwarf that of moving it on die as well.

But even with that, ADL-P's battery life is more or less in line with TGL-H at best. Even if that did lessen the impact of the former's relatively worse uncore due to the growing pains of moving to a heterogeneous architecture, it's clear that the gains aren't massive by any means.

Intel (and if Nuvia pull through on their performanxe claims, so will AMD) needs to do more on top of that, I'm afraid. Slightly better performance but way worse battery life isn't a good look, to say the least.

 

[igor_kavinski]

Paul Otellini rejected making Smartphone chips for Apple because he severely underestimated the volume.

Maybe Apple themselves had no idea of the volume. They must have given Intel some number that Otellini didn't consider worth their while. Also, being as proud as Intel is, he probably thought no one else could supply the volume Apple wanted.

 

[DrMrLordX]

Intel can get around with 12 EUV systems for 45k wpm. Intel needs around 12 layers for its Intel 4 process. 1 EUV layer requires 1 EUV system for 45k wpm.
I don't think EUV capacity will be the bottleneck for Intel 4.

Do they even have 12 EUV systems? Some other reports (albeit from 1-2 years ago, based on present and future shipments of machines through 2023; ASML has been booked for awhile) indicated a peak of 20 kwpm for Intel 4 in 2023.

I am not 100% sure but based on context, @Ajay is saying Intel 4 when he says "N4" and Intel 3 when he says "N3"?

Not sure! Might need to ask him.

 

[Ajay]

Do they even have 12 EUV systems? Some other reports (albeit from 1-2 years ago, based on present and future shipments of machines through 2023; ASML has been booked for awhile) indicated a peak of 20 kwpm for Intel 4 in 2023.

Not sure! Might need to ask him.

Yeah, I did make things confusing. @IntelUser2000 suggested a simple abbreviation systems so we don’t have to write Intel 4, etc. Probably won’t catch on, but it would be convenient.

As far as Intel EUV lithography capacity goes, it does seem that old report underestimated the amount that Intel had or was acquiring. That said, I don’t think the actual number is being publicly shared. Maybe BK didn’t actually cancel EUV orders as was reported.

 

[dttprofessor]

Computer tile is so small,12 EUV systems can produce more chips than ever.

 

[JasonLD]

Do they even have 12 EUV systems? Some other reports (albeit from 1-2 years ago, based on present and future shipments of machines through 2023; ASML has been booked for awhile) indicated a peak of 20 kwpm for Intel 4 in 2023.

Same report that indicated a peak or 20kwpm for Intel 4 in 2023 also reported Intel is project to have 20 EUV systems by 2023.

 

[igor_kavinski]

I read somewhere that EUV machines require considerably more maintenance downtime so that's gotta reduce output by quite a bit too.

 

[bearmoo]

Saphire Rapids delayed again??

https://twitter.com/x/status/1537457755690532864

 

[igor_kavinski]

They might cancel it altogether because possibly the next gen should be ready by then.

 

[jpiniero]

They might cancel it altogether because possibly the next gen should be ready by then.

Emerald is presumably decently bigger though. They might have to cancel that too.

 

[Doug S]

What happened about Intel and mobile and most think it should have been done: Paul Otellini rejected making Smartphone chips for Apple because he severely underestimated the volume. Sure, the cost per chip was low but volume was great. Then the solution was if he accepted making chips for Apple right?

What should have happened: Apple should have never needed to ask Intel in the first place. Their Atom project would have had super low platform power necessary to fit into Smartphones and Intel would have done it organically, out of naturally expanding the computing market. Paul Otellini didn't listen because Intel was absolutely nowhere near the point of getting chips in mobile. Medfield, the first chip that showed that x86 can compete with ARM in battery life, was introduced in 2012, 4 years after the first Atom and 5 years after the iPhone.

If the battery life competitive chip was Silverthorne, even if it was a year after iPhone's introduction, Intel would have took the place of Qualcomm. The performance was way better than the iPhone first gen so Apple might have took Intel.

You seem to be assuming Apple was asking Intel about x86 chips for iPhone. There is no evidence of that. Apple would have looking around for someone to provide them SoCs a couple years before iPhone went on sale, so probably 2005 or so. Intel had its ARM based Xscale CPUs, based on the StrongARM tech they acquired from Digital in the 90s, which Intel sold off in 2006.

 

[nicalandia]

I read somewhere that EUV machines require considerably more maintenance downtime so that's gotta reduce output by quite a bit too.

From: https://www.cnbc.com/2021/12/10/asm...y-machine-is-set-to-transform-chipmaking.html

"Chipmakers looking to print smaller features onto chips have had to rely on double or triple patterning techniques, which is time-consuming," he added. "With a High NA EUV machine, they are able to print these features in a single layer, resulting in a faster turnaround time and better process flexibility."

Chipmakers will have to strike a balance between better performance and the greater costs associated with more sophisticated machinery, Alam said.


"This is particularly true with High NA EUV machines where larger lenses imply higher acquisition and maintenance costs," he said."

So they even out(faster time vs longer maintenance time)

 

[JasonLD]

Emerald is presumably decently bigger though. They might have to cancel that too.

I kinda don’t think it is related to yield at all. Considering ADL execution has been fine.

 

[nicalandia]

I kinda don’t think it is related to yield at all. Considering ADL execution has been fine.

They are having issues validating their whole stack, they have issued multiple Pre-production steppings to try to fix them, but they are still not production ready. Intel first try at Chiplets are harder than expected. If they iron these issues they will have better results for Emerald Rapids.

Due to this we will see Raptor Lake-S being released earlier than Sapphire Rapids(due to Validation issues rather than yield issues), While this will keep Intel competitive at the desktop and Laptop market, this will hurt them at the high profit end since Ice Lake Xeons were design to compete with Rome based EPYC and yet it has been tasked to compete with Milan and Genoa for the mean time which it makes Intel Look Bad at the Highest Profit Margin segment

 

[DrMrLordX]

That said, I don’t think the actual number is being publicly shared. Maybe BK didn’t actually cancel EUV orders as was reported.

Pretty sure it was being tracked by ASML order reports.

Saphire Rapids delayed again??

Intel's enterprise division is in shambles. Truly sad.

 

[uzzi38]

Saphire Rapids delayed again??

https://twitter.com/x/status/1537457755690532864

FFS DCG. This is getting embarrassing.

 

[ashFTW]

According to this SemiWiki article from April, there is a significant gap (”shortage of 18 tools in 2022, 12 tools in 2023 and 20 tools in 2024”) between supply and demand for EUV machines. From the graph below, Intel needs 15, 25, and 15 over this period to satisfy their global foundary service ambitions. Some of these will be High-NA, which they will need for 18A manufacturing in 2025.

Depending on how this “gap“ is shared with TSMC, and Samsung, they may have 20+ EUV, machines by 2nd half of ‘23 when Intel 4 ramps.

Intel and the EUV Shortage - Semiwiki

In my “The EUV Divide and Intel Foundry Services” article available here, I discussed the looming EUV shortage. Two days ago, Intel announced their first EUV tool installed at their new Fab 34 in Ireland is a tool they moved from Oregon. This is another indication of the scarcity of EUV tools. I...

semiwiki.com

 

[Exist50]

Saphire Rapids delayed again??

https://twitter.com/x/status/1537457755690532864

I'm surprised. Is there any other source that agrees? Another quarter or so of delay would be one thing, but a full year on top of what they've already announced? Seems difficult to believe.

 

[ashFTW]

FFS DCG. This is getting embarrassing.

If true, it‘s a lot more than embarrassing. It’s an extinction level event!!

 

[IntelUser2000]

You seem to be assuming Apple was asking Intel about x86 chips for iPhone. There is no evidence of that. Apple would have looking around for someone to provide them SoCs a couple years before iPhone went on sale, so probably 2005 or so. Intel had its ARM based Xscale CPUs, based on the StrongARM tech they acquired from Digital in the 90s, which Intel sold off in 2006.

I heard that the execution of the StrongARM team was pretty bad.

There are prototypes of iPads based on Pentium M. Either they would have gone with Intel's x86 CPUs or they wouldn't have gone with Intel at all. Their fabs until now hasn't been suitable for foundry service. Sure they had process lead but could use only for their chips.

I'm surprised. Is there any other source that agrees? Another quarter or so of delay would be one thing, but a full year on top of what they've already announced? Seems difficult to believe.

A year is indeed devastating. It'll make Sapphire Rapids like a server Cannonlake launch. We'll see what the future holds.

 

[DrMrLordX]

A year is indeed devastating. It'll make Sapphire Rapids like a server Cannonlake launch. We'll see what the future holds.

Kinda makes you wonder if they're just going to cancel Sapphire Rapids for anyone but select customers and move straight to Emerald Rapids.

 

[Ajay]

Kinda makes you wonder if they're just going to cancel Sapphire Rapids for anyone but select customers and move straight to Emerald Rapids.

AT this point, that would make the most sense, IMHO - unless ER is in the gutter as well .

 

[nicalandia]

AT this point, that would make the most sense, IMHO - unless ER is in the gutter as well .

Someone has to pay for being a Beta Tester for SPR.. Emerald Rapids is like SPR+, but without the buggs. I have yet to see on on the Wild(no ES or QS Sample yet)

 

[Exist50]

Kinda makes you wonder if they're just going to cancel Sapphire Rapids for anyone but select customers and move straight to Emerald Rapids.

It's not like that would really change things though. A few percent boost is nothing compared to the gap to Genoa/Bergamo. I wonder if Kuo is getting something mixed up? Launch vs volume?

Edit: typo