Discussion Intel current and future Lakes & Rapids thread

[uzzi38]

no chia didnt block me lol. Intel usually puts out desktop parts on an annual cadence right? So that Q3'21 seems odd if Cometlake is Q1'20.

Especially odd if its just 14nm.

Normally yes. The issue is, according to Sharkbay the 10c CML-S die is later than the 6c one (i.e. probably Q2, not Q1), and on top of that, we're not exactly talking about a normal 14nm product. We're talking about a backport, something most companies of any type of product would stay away from with a 10ft pole.

 

[jpiniero]

Anything is possible when it comes to Intel these days but it may just be where the 10 core Comet S isn't on the market for very long.

 

[DrMrLordX]

Oh wait, they are anyway if RKL-S is Q3 2021 😛

Heaven help them if it's that late. That's about when Alder Lake-S is supposedly going to hit the scene, and not THAT far off from Meteor Lake either.

Intel usually puts out desktop parts on an annual cadence right?

They try. Over the last year, they went from the 9900K to the 9900KS. Not much real progress. If you had said the same thing back in 2018 I'd believe you. Things are different now.

 

[uzzi38]

If, and I really mean if, RKL-S is a TGL backport, then ADL-S will not happen. There's just no point in releasing it if RKL-S is in the same time period and Meteor Lake should be the year after.

It will just get skipped.

 

[DrMrLordX]

If, and I really mean if, RKL-S is a TGL backport, then ADL-S will not happen. There's just no point in releasing it if RKL-S is in the same time period and Meteor Lake should be the year after.

It will just get skipped.

Remember though, Alder Lake-S is supposedly Golden Cove while a TigerLake backport would be Willow Cove. It still would be awkward.

 

[uzzi38]

Remember though, Alder Lake-S is supposedly Golden Cove while a TigerLake backport would be Willow Cove. It still would be awkward.

For sure. But releasing it at the same time as backported RKL-S when 7nm desktop chips will be out in around a years time? Ones that aren't plagues with horrible yields for that matter.

Too much effort, the desktop market ain't worth all that to Intel. Not when the majority of their revenue at that point from the market will be coming from OEMs, who will stick with Intel regardless.

 

[DrMrLordX]

For sure. But releasing it at the same time as backported RKL-S when 7nm desktop chips will be out in around a years time? Ones that aren't plagues with horrible yields for that matter.

Too much effort, the desktop market ain't worth all that to Intel. Not when the majority of their revenue at that point from the market will be coming from OEMs, who will stick with Intel regardless.

Only reason to do it is if Alder Lake-S volumes are so low that they need a backup product to fill in orders. And we still don't know exactly when Meteor Lake will release . . . or in exactly what market segment.

 

[uzzi38]

Only reason to do it is if Alder Lake-S volumes are so low that they need a backup product to fill in orders. And we still don't know exactly when Meteor Lake will release . . . or in exactly what market segment.

Let's be real, that's happening regardless. ADL-S was always going to be meme tier volume.

 

[jpiniero]

Intel's going to have to go to 7 nm ASAP, even on desktop. But that definitely wouldn't be in 2021.

 

[liahos1]

Let's be real, that's happening regardless. ADL-S was always going to be meme tier volume.

what do you have that substantiates that statement?

 

[uzzi38]

what do you have that substantiates that statement?

The state of Ice and Tiger Lakes. Completely lacking in volume for Intel, and these are quad-core dies. There's no evidence suggesting yields will improve by any measurable amount by then.

It'll get funny when Ice Lake-SP 'launches'.

 

[firewolfsm]

I think, for the first time in my life, I have the confidence to short sell a stock. This discussion here, and the reality of what Intel will likely have for sale by the end of 2021, while TSMC will be solidifying its 5nm lineup and getting 3nm ready, makes it clear.

 

[Adonisds]

I think, for the first time in my life, I have the confidence to short sell a stock. This discussion here, and the reality of what Intel will likely have for sale by the end of 2021, while TSMC will be solidifying its 5nm lineup and getting 3nm ready, makes it clear.

Shorting is not good when the company is expected to have hard times ahead, it's good when you know what the market expects of the company in the future and you think the company is gonna do even worse than that. There's a reason beating the market is very hard

 

[JoeRambo]

I think, for the first time in my life, I have the confidence to short sell a stock. This discussion here, and the reality of what Intel will likely have for sale by the end of 2021, while TSMC will be solidifying its 5nm lineup and getting 3nm ready, makes it clear.

You weren't around these forums before C2D hit? Things were way more hilarious back then and according to forum members and bloggers Intel was going BK in several quarters. The rest is history I guess 🙂

The bet against Intel is a bet on their continuous process failures, 10nm is most definitely failure of epic proportions, maybe inflicted on themselves by over ambitious design goals. Hopefully we will get to know the true history of 10nm, maybe that Bohr guy will write book in 20 years, who knows.
Intel has great designs and architecture, they just completely screwed up with 10nm process. If anything arch/design guys are working extra hard to compensate and imagine what could happen if those guys design is on functioning 7nm process, giving them massive transistor budget?

 

[liahos1]

The state of Ice and Tiger Lakes. Completely lacking in volume for Intel, and these are quad-core dies. There's no evidence suggesting yields will improve by any measurable amount by then.

It'll get funny when Ice Lake-SP 'launches'.

a lot can change between now and 2021 lol.

 

[uzzi38]

If RKL-S is actually a backport, then that's a big neon sign Intel don't believe that's the case.

 

[IntelUser2000]

I am trying to find icelake SMT coefficient, anyone ?

if I am not wrong, Icelake has much better SMT coeficient

any reasons why is that? except wider core?

There's so many unknowns there that its impossible to say.

On a simple look, Comet Lake looks to have better scaling. 1134/402(2.82x) versus 1168/430(2.72x).

But again, you'll only find real results on the desktop after they fix frequencies.

 

[NostaSeronx]

It should be:
CML-S -> TGL-S on the 2020(LGA1200) platform <== PCIe 3.0, USB 3.2 2x1
RKL-S -> ADL-S on the 2021(LGA1200v2?) platform <== PCIe 4.0, USB 4.0 2x2

TGL-S is 10nm++ and ADL-S is 7nm. After ADL-S is MTL-S? on 7nm++(First consumer CXL w/ PCIe 6.0[planned but not confirmed]), and after that one is a 5nm -S.

EXE:5000B will only be used on critical layers for 5nm.

 

[beginner99]

I think, for the first time in my life, I have the confidence to short sell a stock. This discussion here, and the reality of what Intel will likely have for sale by the end of 2021, while TSMC will be solidifying its 5nm lineup and getting 3nm ready, makes it clear.

But then the stock market and big money investors isn't rational. (reminder Theranos). The risk I see here is 2-fold. Point one as I recently learned is that big OEMs are firmly in intels hand. Want a threadripper workstation? You can really only get it from a more local provider, not from HP, Dell, etc the ones which companies buy from. Have you ever seen a PC at a company that wasn't intel? I can tell you that the IT departement buying the stuff has 0 clue about the actual hardware and has to buy from the OEM they have a contract with. HP doesn't offer threadripper at all, dell has it in their alienware branding but also absent from workstations.
Besides that intel probably has >95% of the laptop market share and since AMD is kind of focusing on server (and desktop as a side effect, that won't change anytime soon because they lack the process node advantage for now). Not to mention that I highly doubt TSMC even has the capacity to supply that many wafers to AMD. AMD should probably fab the APU at a different foundry if they really want to put a dent into intel.

Second point is that Intel wasn't asleep in these 14nm skylake years. They made cannon core, sunny cove, willow cove and golden cove. If all of them add 10% IPC, it adds up. Once intel is on track with 7nm, it will be quickly back to business as usual and I would not be surpirsed of a "conroe light" moment.

 

[uzzi38]

It should be:
CML-S -> TGL-S on the 2020(LGA1200) platform <== PCIe 3.0, USB 3.2 2x1
RKL-S -> ADL-S on the 2021(LGA1200v2?) platform <== PCIe 4.0, USB 4.0 2x2

TGL-S is 10nm++ and ADL-S is 7nm. After ADL-S is MTL-S? on 7nm++(First consumer CXL w/ PCIe 6.0[planned but not confirmed]), and after that one is a 5nm -S.

EXE:5000B will only be used on critical layers for 5nm.

TGL-S is dead. RKL-S is on LGA1200. ADL-S is unknown for now, but likely also dead.

 

[mikk]

For Rocket Lake-S, the iGPU will be off-die entirely

You have to prove this. Where is the proof?

 

[jpiniero]

You have to prove this. Where is the proof?

Only makes sense since that is what it looks like they are doing with U. It would give them the option to not include the IGP with K parts too.

 

[IntelUser2000]

You have to prove this. Where is the proof?

Rocketlake might be too large to manufacture cheaply with 8 Willow Cove cores and 32EU Gen 12 iGPU. If they enlarge the L3 to 3MB and do a split L2 to extend it to 1.25MB, we might end up having monolithic 8 core RKL being close to 300mm2, which will not help their supply issues with 14nm chips.

RKL is still disappointing even with Willow Cove cores. I guess it makes little more sense on the desktop at least. Laptops are just as hopeless as if it were Skylake cores.

But then the stock market and big money investors isn't rational.

It is rational in that its all about trust. Trust takes years to build. This is why consistent execution is needed to take solid marketshare, not just 1-2 years and coasting afterwards.

Also recent rise of the share value of mega corporations are likely related to the fact that we have decade of record low interest rates and people can't rely on savings accounts anymore. Overall value of the market has been rising for the past few years.

 

[DrMrLordX]

You have to prove this. Where is the proof?

Like I said . . . that's the assumption. Every leak seems to point in that direction.

 

[NostaSeronx]

we might end up having monolithic 8 core RKL being close to 300mm2

With the track height optimization present with Willowcove, that is unlikely to happen.

Skylake on 14nm uses the tallest height cells for std cells. Which is around ~672-nm(Intel 14 nm FinFET Kaby Lake CPU 12.92-Track Library Standard Cell). While, Willowcove in 14nm should be using the least tallest track height. Which is targeting Samsung's 11LPP 6.75T logic cell ~324-nm. Which I believe will be the 6T cell(~312-nm) or 7.67T(~399-nm).

RKL should also be on the 14nm equivalent of 22FFL+. So, 14nm FinFET using the 10nm FinFET technologies.

CML/RKL -S/-H are meant to be cost-sensitive relative to the 10nm TGL/ADL -S/-H solutions.

RKL-S will be the Pentium to ADL(/TGL)-S Core i's.