Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

[Tigerick]

As Hot Chips 34 starting this week, Intel will unveil technical information of upcoming Meteor Lake (MTL) and Arrow Lake (ARL), new generation platform after Raptor Lake. Both MTL and ARL represent new direction which Intel will move to multiple chiplets and combine as one SoC platform.

MTL also represents new compute tile that based on Intel 4 process which is based on EUV lithography, a first from Intel. Intel expects to ship MTL mobile SoC in 2023.

ARL will come after MTL so Intel should be shipping it in 2024, that is what Intel roadmap is telling us. ARL compute tile will be manufactured by Intel 20A process, a first from Intel to use GAA transistors called RibbonFET.





Intel Core Ultra 100 - Meteor Lake



As mentioned by Tomshardware, TSMC will manufacture the I/O, SoC, and GPU tiles. That means Intel will manufacture only the CPU and Foveros tiles. (Notably, Intel calls the I/O tile an 'I/O Expander,' hence the IOE moniker.)

 

[adroc_thurston]

I’ll add a profile pic to this account if 18A surpasses 5.7GHz

not happening

 

[igor_kavinski]

I’ll add a profile pic to this account if 18A surpasses 5.7GHz

That's not high stakes enough.

Put a profile pic that @511 chooses for you

 

[OneEng2]

I don't like "to big to fail" but if Intel spunoff their fabs that might be a good investment.

I think that the fab spinoff is virtually guaranteed at this point.

The big question is who will buy it, and what will they do with it?

My personal belief is that Intel simply tried to "shoot for the moon" with 18A. It was a HORRIBLY risky node on several fronts.

Samsung got the a** handed to them trying to pull off GAA. Seems like TSMC took their sweet time getting there with N2, but the slow and steady hand seems to be paying off.

Intel's 18A not only goes to GAA (first for Intel), but includes BSPDN (first for Intel). At least they are getting a handle on their EUV equipment. At least they backed off the idea of ALSO using High NA EUV for this node as well.

TSMC on the other hand is going to dip its toe SOOOOO slowly into BSPDN. So much so that it can make the same design on either N2 or N16 .... just in case a design goes sound with BSPDN.

I hope that Intel's foundry can not only be saved, but be made "industry leading" again.

Which Intel won't cause the shareholders won't allow the milking cow to be separate so it's going to be a difficult thing it's possible but Intel products would demand good Money from US Government for this lol.

How do you think Intel is going to pay the bills if the DON'T sell off the foundry?

Also if Panther Lake is having such poor yields then imagine what's happening with Clearwater Forest and Diamond Rapids.

While I expect CWF and Diamond Rapids to have larger die sizes than PTL, they will also be clocking much lower. 18A might well be a very good fit for DC parts. Time will tell.

Well if PTL is between 4.5 GHz - 5Ghz it's salvageable if it's below 4Ghz it's embarrassing and can't be salvaged. 😂

With an IPC increase of no more than 5% a drop in clock speed to only 5Ghz (down from 5.7Ghz) would put PTL south of ARL in performance. Not sure how that would work.

 

[poke01]

With an IPC increase of no more than 5% a drop in clock speed to only 5Ghz (down from 5.7Ghz) would put PTL south of ARL in performance. Not sure how that would work.

He maybe referring to mobile ARL-H

 

[511]

How do you think Intel is going to pay the bills if the DON'T sell off the foundry?

They will stop pursuing leadi

With an IPC increase of no more than 5% a drop in clock speed to only 5Ghz (down from 5.7Ghz) would put PTL south of ARL in performance. Not sure how that would work.

PTL is mobile only also if we believe rumors from reuturs 10-15% SKU Archives targets so they should clock good majority Will be i5

 

[511]

Why would you count filler tiles for these 70% ?
That is really being very economical with the truth for a CEO.
But I guess par for the course for intel.

If you count other SKU it's easily greater than 70% as the GPU Tile is Intel.

 

[DrMrLordX]

Well if PTL is between 4.5 GHz - 5Ghz it's salvageable if it's below 4Ghz it's embarrassing and can't be salvaged. 😂

This isn't 2019, Intel can't roll out a 4.5 GHz part in their notebooks and expect to sell them for anything but scraps. Especially with the relatively poor improvements anticipated for the P-cores over Arrow Lake. If Intel manages to ship 4.5-5 GHz parts that are ultra low power for what they do, then they may get some design wins, but that's a stretch.

I think that the fab spinoff is virtually guaranteed at this point.

The big question is who will buy it, and what will they do with it?

Uncle Sam, and (if they had any brains) they'd make it a consortium project involving other American or pseudo-American companies. But I've posted on this idea before and I'm starting to repeat myself. Bonus points if they go international and invite Samsung into the consortium.

My personal belief is that Intel simply tried to "shoot for the moon" with 18A. It was a HORRIBLY risky node on several fronts.

Sort of reminiscent of 10nm no?

Samsung got the a** handed to them trying to pull off GAA.

Samsung was having problems even before 3GAE. They rushed into GAA because they knew they couldn't compete with TSMC on EUV finfet nodes.

While I expect CWF and Diamond Rapids to have larger die sizes than PTL, they will also be clocking much lower. 18A might well be a very good fit for DC parts. Time will tell.

One of the things I'm not clear on is: how do poor parametric yields affect parts that don't necessarily need high clocks, assuming defect rates are low (or at least, not so high as to prohibit full production)? Are they going to get poor leakage characteristics compared to a hypothetically nearly-identical version of 18A that (for whatever reason) didn't have poor parametric yields and could operate well in the 5-6 GHz range?

With an IPC increase of no more than 5% a drop in clock speed to only 5Ghz (down from 5.7Ghz) would put PTL south of ARL in performance. Not sure how that would work.

Yeah that would be really bad. It would have to sip power to make any sense.

 

[adroc_thurston]

Sort of reminiscent of 10nm no?

Oh no the statement is wrong, 18A was derisked in multiple steps.
Intel just kinda sucks.

Samsung was having problems even before 3GAE. They rushed into GAA because they knew they couldn't compete with TSMC on EUV finfet nodes.

ding-ding-ding, it's been a yield struggle session ever since 7LPP.

 

[511]

This isn't 2019, Intel can't roll out a 4.5 GHz part in their notebooks and expect to sell them for anything but scraps. Especially with the relatively poor improvements anticipated for the P-cores over Arrow Lake. If Intel manages to ship 4.5-5 GHz parts that are ultra low power for what they do, then they may get some design wins, but that's a stretch.

I meant for volume U5/U7 they can clearly get few U9/U7 clocking higher Halo SKU it's still doable but anything beyond that is not salvageable.

 

[DrMrLordX]

I meant for volume U5/U7

It's all guesswork as to where these things top out clockspeed-wise, but if 90% or so of the product tops out at 4.5 GHz then the prices they can command for these products will be pretty low.

 

[511]

that is true it all depends on the yields physical and parametric

 

[LightningZ71]

Oh no the statement is wrong, 18A was derisked in multiple steps.
Intel just kinda sucks.

ding-ding-ding, it's been a yield struggle session ever since 7LPP.

The sad part is that they aren't just late, they aren't even eventually fixing the yields on many of their older nodes. Instead of figuring out and rectifying what went wrong, they seem to just stumble along into the next node (or half step), which flops, and only mature previous nodes a few percentage. I don't know of a single 5nm or below Samsung node that has ever achieved PPA targets while also yielding well enough to break even with any sort of real production volume.

 

[eek2121]

I think that the fab spinoff is virtually guaranteed at this point.

The big question is who will buy it, and what will they do with it?

My personal belief is that Intel simply tried to "shoot for the moon" with 18A. It was a HORRIBLY risky node on several fronts.

Samsung got the a** handed to them trying to pull off GAA. Seems like TSMC took their sweet time getting there with N2, but the slow and steady hand seems to be paying off.

Intel's 18A not only goes to GAA (first for Intel), but includes BSPDN (first for Intel). At least they are getting a handle on their EUV equipment. At least they backed off the idea of ALSO using High NA EUV for this node as well.

TSMC on the other hand is going to dip its toe SOOOOO slowly into BSPDN. So much so that it can make the same design on either N2 or N16 .... just in case a design goes sound with BSPDN.

I hope that Intel's foundry can not only be saved, but be made "industry leading" again.

How do you think Intel is going to pay the bills if the DON'T sell off the foundry?

While I expect CWF and Diamond Rapids to have larger die sizes than PTL, they will also be clocking much lower. 18A might well be a very good fit for DC parts. Time will tell.

With an IPC increase of no more than 5% a drop in clock speed to only 5Ghz (down from 5.7Ghz) would put PTL south of ARL in performance. Not sure how that would work.

If they can’t make it work, they are shutting that part of the business down completely, not spinning it off. There will be no IFS. It will be dead and buried. Intel needs the process to work and they need customers. They will get things working, however, the lack of customers is a huge problem. They bet billions and they are losing that bet…

If they get things working properly on time, they might be able to turn it around. If not…

 

[511]

If they can’t make it work, they are shutting that part of the business down completely, not spinning it off. There will be no IFS. It will be dead and buried. Intel needs the process to work and they need customers. They will get things working, however, the lack of customers is a huge problem. They bet billions and they are losing that bet…

hang on they are not shutting down IFS they are shutting further investment in 14A and beyond IFS will exists and they will milk 18A like they did with 14nm/10nm and now 18A anyone's guess how many '+' 18A will get

 

[eek2121]

hang on they are not shutting down IFS they are shutting further investment in 14A and beyond IFS will exists and they will milk 18A like they did with 14nm/10nm and now 18A anyone's guess how many '+' 18A will get

No, they will exit the business completely.

The statement he made was carefully worded to not spook investors even more than they are. If IFS can’t become profitable quick, all future products will be made at TSMC and IFS will cease to exist.

 

[511]

No, they will exit the business completely.

The statement he made was carefully worded to not spook investors even more than they are. If IFS can’t become profitable quick, all future products will be made at TSMC and IFS will cease to exist.

Wrong there was a letter published they are shutting down future node not packing and present nodes this would just kill the company

 

[Magio]

Regarding PTL, I'm not gonna say who but I DMd a reliable Twitter leaker, and in their opinion the performance of PTL shouldn't be the worry at this stage and if there's trouble it will be in the volume and pricing Intel will deliver. So fear of 4.5GHz parts might not be the correct fear to have.

Perhaps their parametric yields are so bad because the performance targets themselves are sky high.

Wrong there was a letter published they are shutting down future node not packing and present nodes this would just kill the company

I have trouble believing IFS can continue to exist on legacy nodes alone. Sooner or later their own products wouldn't require those nodes, and if they struggle to compete in the foundry space at the cutting edge I don't fancy their chances on mature nodes.

Cutting edge was their big opportunity, that Pat Gelsinger rightly identified. A market where only TSMC currently matters, and one that only they and maybe Samsung had any chance at all to join the Taiwanese giant in. A market that comes with the added benefit that even if you can't quite compete on cost with TSMC, you can still compete with their wafer prices and retain decent margins. A market where success helps both their products and foundry services.

Maybe Gelsinger only had the right idea and not the right execution but so far LBT seems to have neither. It's really hard to imagine Intel succeeding on any front in the mid-long term with the current approach.

 

[511]

have trouble believing IFS can continue to exist on legacy nodes alone. Sooner or later their own products wouldn't require those nodes, and if they struggle to compete in the foundry space at the cutting edge I don't fancy their chances on mature nodes.

Global Foundries exist and the nodes they use are way past anything 18A/AP is going to be relevant for way longer than 12/14nm was

 

[OneEng2]

Uncle Sam, and (if they had any brains) they'd make it a consortium project involving other American or pseudo-American companies. But I've posted on this idea before and I'm starting to repeat myself. Bonus points if they go international and invite Samsung into the consortium.

I kinda agree with you. I would definitely put it in the "national interest" catagory.

Sort of reminiscent of 10nm no?

Frightenly so. One would have thought they would have learned from that experience.

Samsung was having problems even before 3GAE. They rushed into GAA because they knew they couldn't compete with TSMC on EUV finfet nodes.

Sort of like 18A's BSPDN?

One of the things I'm not clear on is: how do poor parametric yields affect parts that don't necessarily need high clocks, assuming defect rates are low (or at least, not so high as to prohibit full production)? Are they going to get poor leakage characteristics compared to a hypothetically nearly-identical version of 18A that (for whatever reason) didn't have poor parametric yields and could operate well in the 5-6 GHz range?

Agree. My assumption was that the problem is limited to getting a high clock part to yield and that at lower clocks, it would sip energy.

Global Foundries exist and the nodes they use are way past anything 18A/AP is going to be relevant for way longer than 12/14nm was

I was going to make the same comment. The idea that Intel would simply close the door on such a valuable part of the company and not try to sell it is just pure unadulterated business stupidity. There is less than 0 chance of this happening IMO.

 

[igor_kavinski]

but so far LBT seems to have neither.

His primary objective is stopping the bleeding of cash and actually helping Intel turn a profit. The mess created by Pat is just too insane:

trying to sell existing Raptor Lake inventory (even Alder Lake is still not sold out),

dealing with RPL RMAs,

Intel 7 still being a harsh reality to deal with because those fabs were not converted to Intel 3 or 4,

trying to woo back OEMs because Pat's decisions led them to look to AMD to fulfill their mobile products demand,

trying to trim Pat's fat engineering teams to make them more lean and agile

and so many other problems that are probably not even public and which contributed to Pat's sudden termination.

 

[msj10]

PTL is 5.1GHz apparently

https://www.zhihu.com/question/1936359085550318318/answer/1936380500546683598

 

[511]

Same as Lunar Lake/Meteor Lake

 

[Kepler_L2]

View attachment 128431
PTL is 5.1GHz apparently

https://www.zhihu.com/question/1936359085550318318/answer/1936380500546683598

Not as bad as I thought, but also lmao@Semiwiki and everyone else who parroted "18A is faster than N2P"

 

[511]

Same as Lunar Lake/Meteor Lake

 

[poke01]

Not as bad as I thought, but also lmao@Semiwiki and everyone else who parroted "18A is faster than N2P"

awww, i really wanted a profile pic