Discussion Intel current and future Lakes & Rapids thread

[dacostafilipe]

Before 2018, Skylake was 14nm until they changed it to 14nm+ for Cooperlake's release

https://images.anandtech.com/doci/15971/Section%201%20%2815%29.jpg

You want them to rename their old products to this?
Broadwell 14nm, Skylake 13nm, Kabylake 12nm, Coffeelake 11nm, Cooperlake 10nm
Cannonlake 9nm, Icelake is 8nm, Tigerlake 7nm, Sapphire Rapids 6nm, Raptorlake 5nm?
Meteorlake 4nm, Lunarlake 3nm

Did you not get the message? "nm" is dead.

 

[Ajay]

Except now Intel gets to try to bamboozle investors. Look, we shipped 7nm product in 2021! Yeah, okay.

The dumb ones, yes. Of course, large investors are playing an entirely different game where they make can make money on Intel whether it's stock is up or down.

 

[eek2121]

Honestly, I don't care if Intel names their process nodes '10 flying butt monkeys'. If they get competitive again, great!. If they don't, rats! Real competition is the only price control we have. If Intel were to offer 20% more performance/$ the next time I upgrade; I'd consider Intel. Same with AMD (as was the case this time, plus better perf/watt).

Gelsinger hasn't had hardly enough time to make any real changes at Intel. Though, obviously he's applying pressure since he got the process/manufacturing team to commit to this timeline. So while it is to some extent it's a 'dog and pony show', it's also a commitment that the press and investors will hold Intel to. That's not nothing.

FINALLY someone who is intelligent enough to understand.

 

[gdansk]

I suspect they must be happy with ADL if they're rebranding its process. Otherwise it would make their shiny new name look like trash right out of the gate. Why rebrand if you're going to tarnish the new naming scheme immediately?

But maybe I'm reading too much into it.

 

[geegee83]

Anyone counted the MTL dies from the wafer picture? Looks extremely small to me.

 

[Zucker2k]

Intel's explanation for the adoption of the new node nomenclature.

Intel 7 Summary of Enhancements

Another Big Jump in Performance and a New Name; This is Intel 7

Author: Chris Auth, Vice President and Director, Logic Technology Development

medium.com

 

[IntelUser2000]

Anyone counted the MTL dies from the wafer picture? Looks extremely small to me.

It does look really small at the initial glance but it should be a lot larger. Looking at the picture more closely you can see that there are two distinct patterns. Likely they are different parts of the die.

Since Intel revealed how Meteor Lake package looks like, the compute tile can't be that small. The three "tiles" take up quite a bit of the package size.

Another Big Jump in Performance and a New Name; This is Intel 7

Author: Chris Auth, Vice President and Director, Logic Technology Development

medium.com

Great technical coverage partially negated by following the stupid conventions of the industry.

The article is also saying 10-15% is greater than a full node gain.

Marketing essentially.

 

[Asterox]

I suspect they must be happy with ADL if they're rebranding its process. Otherwise it would make their shiny new name look like trash right out of the gate. Why rebrand if you're going to tarnish the new naming scheme immediately?

But maybe I'm reading too much into it.

In short, it is marketing move.

"If you have good power efficient CPU", numerical marketing gimmicks are not required.

https://twitter.com/x/status/1420069663879409669

 

[coercitiv]

The article is also saying 10-15% is greater than a full node gain.

Just for accuracy, they're stating that SF + ESF gains are equivalent to a full node gain.

 

[DrMrLordX]

The dumb ones, yes. Of course, large investors are playing an entirely different game where they make can make money on Intel whether it's stock is up or down.

It was the same deal when BK pushed through Cannonlake in a few lame 2c laptops sold only in China to fulfill his promise of 10nm in 2017.

 

[semiman]

Somehow people are quite sarcastic here. Intel presentation this week was quite different, compared to previous ones. Now, Intel is using words like 'parity', meaning that they're behind. At least Intel accepts the truth. A year ago, their presentation was like "yeeeey, we're the best, everything's on schedule" but hiding every detail then later they say "hey look, financial numbers' all good anyway" but not this year.

As for a node renaming, it's more like they're following foundries norm, and some for marketing. OEMs don't like it if AMD counterpart is branded as TSMC 7nm when Intel's using 10nm SF. That's what foundry started(half node transition numbering etc), but now that's a norm anyway.

Also, now they're opening detailed Foveros and EMIB schedules as well with certain numbers(50um 35um...) with product schedules. PowerVia as well. They're not putting some shiny 1~2 year unrealistic shiny schedules and self-praising(of course there are some) to satisfy investors. They shared hardships ahead and provided how they're planning to overcome with a mix of different technologies. I think we shouldn't really undermine Intel simply because they re-branded their nodes.

Of course, execution is what really matters. But at least, I see that there are some good people in Intel who want to change and there are "paranoids" out there who still love Intel(maybe Inside again?). Don't forget that Pat was the one who started IDF. 2017 was a quite shocking year for me, and now he's planning to bring it back(IntelON). Probably this will end in vain, but without trying, they're already done for.

 

[coercitiv]

Somehow people are quite sarcastic here. Intel presentation this week was quite different, compared to previous ones. Now, Intel is using words like 'parity', meaning that they're behind. At least Intel accepts the truth. A year ago, their presentation was like "yeeeey, we're the best, everything's on schedule" but hiding every detail then later they say "hey look, financial numbers' all good anyway" but not this year.

I think you need a refresh on what Intel was saying last year in March about 10nm to investors in a Morgan Stanley conference call:

this isn't just going to be the best node that Intel has ever had. It's going to be less productive than 14nm, less productive than 22nm, but we're excited about the improvements that we're seeing and we expect to start the 7nm period with a much better profile of performance over that starting at the end of 2021 .

... but the fact is that I wanted to be clear what was happening during the 10nm generation. The fact is, it isn't going to be as strong a node as people would expect from 14nm or what they'll see in 7nm.

Not only were they already admitting being behind, they also made public plans to achieve parity with their 7nm node and regain the lead on 5nm. Here's a quote from ComputerBase:

In the 7 nm production at the end of 2021, Intel wants to reach the level of parity with the competition and be significantly stronger than with 10 nm, in 5 nm the manufacturer then wants to take the lead again.

Sounds familiar?

 

[Ajay]

It was the same deal when BK pushed through Cannonlake in a few lame 2c laptops sold only in China to fulfill his promise of 10nm in 2017.

Yes, the dumb ones just said 'yes, 10nm is out!'. The smart ones said 'not' and their hedge fund managers had already figured out how to play this either way. Or, they just backed totally away from Intel and sought better prey. So, yes you are correct; I should have left off the bit about investors holding Intel accountable.

 

[blckgrffn]

Yes, the dumb ones just said 'yes, 10nm is out!'. The smart ones said 'not' and their hedge fund managers had already figured out how to play this either way. Or, they just backed totally away from Intel and sought better prey. So, yes you are correct; I should have left off the bit about investors holding Intel accountable.

And let's face it, the computer algos don't care about rebranding exercises that any company employs.

Many of these large hedge funds and the like already have models for so much of this stuff at the macro and micro level they just keep plugging in data and going with it.

If anything, they are betting on how the market reacts to news being shared then what the news even is in any detail.

Retail investors are a little piece of the pie and I know there are the GME's of the world but really that's the exception to the rule.

 

[Gideon]

A bit tangental to the topic, but still relevant as puts certain rumors very much ib doubt.

People here were claiming that TSMC will just blindly give most leading edge wafers blindly to Intel with no regard to current customers ...

Well straight from Morris Chang:

https://twitter.com/x/status/1420406263997878274

In this forum [timestamp: 01:05:28], Morris Chang said (paraphrasing):

"TSMC has 3 main competitors - GlobalFoundries, Samsung and Intel. Samsung and Intel are 700-lb gorillas. Although Intel is an indirect competitor, the threat posed by Intel is the same and just as serious."

"Intel competes with our customers, who depend on us for manufacturing and process technology. If we fall behind or fail to enable our customers to compete with Intel, then TSMC is finished; we will have no customers. TSMC's mission is to help our customers defeat Intel."

So they are well aware what giving most of their 3nm production to Intel would result in

 

[Dayman1225]

A bit tangental to the topic, but still relevant as puts certain rumors very much ib doubt.

People here were claiming that TSMC will just blindly give most leading edge wafers blindly to Intel with no regard to current customers ...

Well straight from Morris Chang:

https://twitter.com/x/status/1420406263997878274

So they are well aware what giving most of their 3nm production to Intel would result in

Morris Chang is no longer the boss/ceo of TSMC iirc. So whether or not that is still what people think at TSMC is another question.

 

[JoeRambo]

I am pretty much sure that Intel had to pay steep price for TSMC foundry access. Kinda strange that noone in this thread commented, but future Intel nodes involve rather little of EUV*, and that's where Intel's "price" probably was - if that rumor we heard earlier about Intel cancelling their EUV machine orders that were in queue.
* as in starting in metal layers that currently require quad patterning, then moving to using EUV to draw fins of transistors and then somehow making leap to that 4-Ribbon FET all in ~3 years with minor amount of EUV machines they will have? Sounds exactly like setup for minor production while buying from TSMC to outlast till supposed High-NA revolution.

So TSMC was happy to buy extra EUV machines now, while Intel is ordering wafers from TSMC and betting the farm on High-NA EUV machines that might come or might not come in time. Sounds fair to both companies to me from business perspective. And I am 100% sure TSMC knows about High-NA timelines enough to make informed decisions for themselves.

 

[Doug S]

Intel might have always had orders booked for the first high NA EUV machines, it isn't necessarily TSMC sacrificing anything, they might not have been ordering as far ahead as Intel. Intel has a much bigger cash pile, and until recently much of it was "stranded" overseas for tax reasons.

So piling up orders with ASML many years in advance (especially back when they thought their roadmap was humming along) made perfect sense. Its zero risk because you know if you order too many others will want them so you can cancel/trade/sell.

 

[A///]

Intel can afford to take the gamble now. It's a 50/50 split at the moment between success and failure.

 

[Asterox]

In short, it is classic paper lounch.

95 percent of users do not need the most expensive motherboards, and especially not the most expensive CPU models.

Intel Alder Lake S Launch - Only Enthusiast CPUs and Z690 Chipset between Oct. 25 and Nov. 19 in 2021, the rest is coming later | igor´sLAB

Disclaimer: The following article is machine translated from the original German, and has not been edited or checked for errors. Thank you for understanding!

www.igorslab.de

 

[jpiniero]

In short, it is classic paper lounch.

95 percent of users do not need the most expensive motherboards, and especially not the most expensive CPU models.

Intel Alder Lake S Launch - Only Enthusiast CPUs and Z690 Chipset between Oct. 25 and Nov. 19 in 2021, the rest is coming later | igor´sLAB

Disclaimer: The following article is machine translated from the original German, and has not been edited or checked for errors. Thank you for understanding!

www.igorslab.de

We are talking about DIY, which K is what Intel pushes.

 

[Shivansps]

Thats sounds like Broadwell with extra steps. (and a bad IGP)

 

[jpiniero]

Thats sounds like Broadwell with extra steps. (and a bad IGP)

Intel's done this a lot. It's quicker getting the chips into the retail market than through OEMs who then have to assemble the PCs and then ship it (and they prefer by boat which takes awhile)

Usually they do include the lowest i5 locked as well though.

 

[LightningZ71]

I'm more interested in their troubles getting the market to move to PCIe 5. I can see where Intel would gain from having a PCIe 5 link between their CPU and Chipsets as 4 lanes of that would keep almost anything short of massive HEDT rigs fed.

 

[DrMrLordX]

In short, it is classic paper lounch.

It's not really a paper launch if they have the chips ready to meet demand on launch day, or at least can ship in volume (assuming they find demand like AMD did with Vermeer; you can't say Vermeer was a paper launch just because it was sold out). That they choose to meet the needs of a narrow segment of the market does say something about the total volume they intend to ship.