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.)

 

[SiliconFly]

Yeah but MTL already looks like a poor product with the LNL leak. Jump from 6400 to 8533 MT/s RAM is reason enough to wait one more year.

Sealing the fate too early are we! What if LNC final turns out to be twice as performant as RWC and more power efficient with less die space. It's very much possible!

 

[SiliconFly]

What if there's an LNL prototype on their campus right this instant?

When hell freezes over!

 

[Tigerick]

Apple is aparently skipping the original N2.

I remember you are the one who claimed N2 is DOA, that is total BS. At least you can download the TSMC Q3's investor transcript to see what CC Wei has been saying about N2 process:-

"Our 2-nanometer technology will be the most advanced semiconductor technology in the industry in both density and energy efficiency when it is introduced in 2025. Our N2 technology development is progressing well and on track for volume production in 2025. Our N2 will adopt nanosheet transistor structure, which has demonstrated excellent power efficiency. N2 will deliver full node performance and power benefits to address the increasing need for energy-efficient computing.

As part of N2 technology platform, we also developed N2 with Backside Power Rail solution, which is best suited for HPC applications. We are targeting Backside Power Rail to be available in the second half of 2025 to customers with production in 2026. With our strategy of continuous enhancement, N2 and its derivative will further extend our technology leadership well into the future."

N2 HVM in 2025 and N2P HVM in 2026, got it?

 

[adroc_thurston]

How could you not mentioning Apple who would be the biggest buyer of N2 process?

Apple is too poor for N2.
Cloud guys can (and will) pay tho.

 

[SiliconFly]

So by going with N3B they are going to regain process leadership? That doesn't make any sense.

Also, why are they sitching their SoC CPU cores already? Guess they didn't work out as planned.

Contrary to all thats being spoken. LNL is a niche product. ARL is their main client consumer line.

Making LNL in N3B is actually a step down compared to 20A, but the speed they're trying to rush out LNL is to de-risk LNC & Battlemage for nextgen PNL. Exactly like Lakefield.

 

[Tigerick]

Apple is too poor for N2.
Cloud guys can (and will) pay tho.

Thanks for your objective view of Apple. How did AMD's launching of Turin Dense and Sarlak that based on N3E going?

 

[adroc_thurston]

Thanks for your objective view of Apple.

Yeah cheap is how they roll around, see M3 Pro.

How did AMD's launching of Turin Dense and Sarlak that based on N3E going?

Pretty well, both are on track.

 

[Thunder 57]

Okay.. I'm with you so far

Should they hold off another 6-9 months on Lunar Lake so it can be fab'd on 18A for better PR and internet points or launch their mobile processor on a very efficient TSMC N3 node? It seems you're suggesting the former.

I don't think that at all.

If they can launch Lunar Lake 6-9 months earlier than if they held of for 18A, clearly that is the right move. I really am not suggesting anything other than Intel's fab problem seems to be still ongoing. I just think it is a bit sad that the once unbeatable juggernaut that was Intel's fab advantage seems to be gone. Maybe they retake the lead. Time will tell.

Also, they way you worded the TSMC vs Samsung thing was a bit odd. I don't think anyone believes that. I just wanted to point out that dual sourcing doesn't mean AMD had no faith in TSMC.

Contrary to all thats being spoken. LNL is a niche product. ARL is their main client consumer line.

Making LNL in N3B is actually a step down compared to 20A, but the speed they're trying to rush out LNL is to de-risk LNC & Battlemage for nextgen PNL. Exactly like Lakefield.

We shall see. Maybe N3B is more pwer efficient which matters a lot for something like Lunar Lake. That would still be a bad look. I also don't think LNL is niche. It seems the low power parts sell the most units. Will there be a 15W ARL?

I do think Arrow Lake will tell us a lot about Intel. If it is a home run I think a lot of people, myself included, would feel better about Intel's future. I'm just not to sold on the idea they can reduce power consumption while increasing performace in a significant way just yet. Next year sure will tell us a lot.

 

[SiliconFly]

I'm just not to sold on the idea they can reduce power consumption while increasing performace in a significant way just yet. Next year sure will tell us a lot.

Yep. Considering the scarcity of LNC leaks & the extreme secrecy, not much can be said as of now. But when LNC was first mentioned, rumors said Intel was working on 2 designs. One with HT & the other with rentable units. And they ditched the former in favor of the latter.

Which means, they have very high confidence in ARL's LNC. That should make ARL much better than MTL both in performance & efficiency i think.

 

[DrMrLordX]

Uhh not looking good for intel foundry. Made on N3B lol

That's really odd. Why N3B? Isn't Lunar Lake in the timeframe where Intel could use N3E instead? Are they bound by contract to take only N3B wafers? Or did they just roll out the design for N3B and they're sticking with it to avoid the cost of moving the design to N3E with its different design rules?

So by going with N3B they are going to regain process leadership? That doesn't make any sense.

Oops don't mention the 800 lb gorilla in the room. Personally I figured that internal partisans were pushing them off N3B for Arrow Lake (granted this leak says nothing about Arrow Lake specifically), and that lined up well with the Macworld et al leaks claiming that Intel had bagged on N3B entirely and handed what was left of their allocation to Apple. Obviously that last bit can't be entirely true if Lunar Lake arrives on N3B and not N3E (or anything else). Which makes you wonder, what other consumer CPU tiles will be on N3B?

 

[igor_kavinski]

What if LNC final turns out to be twice as performant as RWC and more power efficient with less die space.

Intel engineer: 2x IPC gains has never been done before.

Pat: What do you need?

Intel engineer: Lots and lots of drugs!

 

[Doug S]

Apple is too poor for N2.

Can't tell if you're trolling or actually dumb enough to believe that.

 

[dullard]

If they can launch Lunar Lake 6-9 months earlier than if they held of for 18A, clearly that is the right move.

I'm not sure of why anyone would be surprised that Intel is using TSMC for Lunar Lake.

1A) Intel said Lunar Lake and Beyond would be use two nodes: external nodes AND internal. https://www.intel.com/content/www/u...ogy-roadmaps-milestones.html#client-computing

1B) Intel clairified Lunar Lake and Beyond is External nodes and 18A: https://cdn.mos.cms.futurecdn.net/kNDC8wGsXbuxPxf9Rz7oJo-650-80.png.webp

2) Intel said Lunar Lake is to be released in 2024: https://cdn.mos.cms.futurecdn.net/ma2JM6VHHsj5wPGmokXq4U-650-80.png.webp

3) Intel said 18A will be market ready in 2025: https://www.intel.com/content/www/u...-update-internal-foundry-model.html#gs.0nr1ph

Put all three together and you get that Lunar Lake is an external node. That is because 18A won't be ready yet. I believe the latest rumor is that it will be Panther Lake that uses 18A in 2025.

 

[Geddagod]

I'm not sure of why anyone would be surprised that Intel is using TSMC.

1A) Intel said Lunar Lake and Beyond would be use two nodes: external nodes AND internal. https://www.intel.com/content/www/u...ogy-roadmaps-milestones.html#client-computing

1B) Intel clairified Lunar Lake and Beyond is External nodes and 18A: https://cdn.mos.cms.futurecdn.net/kNDC8wGsXbuxPxf9Rz7oJo-650-80.png.webp

2) Intel said Lunar Lake is to be released in 2024: https://cdn.mos.cms.futurecdn.net/ma2JM6VHHsj5wPGmokXq4U-650-80.png.webp

3) Intel said 18A will be market ready in 2025: https://www.intel.com/content/www/u...-update-internal-foundry-model.html#gs.0nr1ph

Put all three together and you get that Lunar Lake is an external node. That is because 18A won't be ready yet. I believe the latest rumor is that it will be Panther Lake that uses 18A in 2025.

Dang you actually laid all the logic out. Iswtg people who keep arguing that Intel is never going to use N3 for their CPU tiles need to see this.
Also PTL using 18A in 2025 is confirmed by Intel

 

[adroc_thurston]

Can't tell if you're trolling or actually dumb enough to believe that

10% chip density for a solid chungus wafer price bump is not something anyone outside of DC can afford.
Only products like venice-dense are viable on vanilla N2.

 

[Goop_reformed]

Intel engineer: 2x IPC gains has never been done before.

Pat: What do you need?

Intel engineer: Lots and lots of drugs!

Careful, the intel defense force might take it as a personal attack. These slides are to be believed but not the OTHER slides . Still not looking good for IF. When will they start fabbing their own client chips?

 

[FlameTail]

I mean when others have reached parity in perf/watt. I know that despite the marketing, Apple's real aim was vertical integration.

I very heavily doubt Intel can reach perf/Watt parity with Apple.

 

[Goop_reformed]

I very heavily doubt Intel can reach perf/Watt parity with Apple.

They can, but only when comparing 2025-2026 chips to a 2021 chip.

 

[Thunder 57]

I'm not sure of why anyone would be surprised that Intel is using TSMC for Lunar Lake.

1A) Intel said Lunar Lake and Beyond would be use two nodes: external nodes AND internal. https://www.intel.com/content/www/u...ogy-roadmaps-milestones.html#client-computing

1B) Intel clairified Lunar Lake and Beyond is External nodes and 18A: https://cdn.mos.cms.futurecdn.net/kNDC8wGsXbuxPxf9Rz7oJo-650-80.png.webp

2) Intel said Lunar Lake is to be released in 2024: https://cdn.mos.cms.futurecdn.net/ma2JM6VHHsj5wPGmokXq4U-650-80.png.webp

3) Intel said 18A will be market ready in 2025: https://www.intel.com/content/www/u...-update-internal-foundry-model.html#gs.0nr1ph

Put all three together and you get that Lunar Lake is an external node. That is because 18A won't be ready yet. I believe the latest rumor is that it will be Panther Lake that uses 18A in 2025.

I'm confused as to why Intel is using TSMC for the compute tile. I very much expected graphics and/or I/O dies to be made by TSMC. I don't get how people are spinning this as being just fine. What happened to 5 nodes in 4 years?

 

[Thunder 57]

Dang you actually laid all the logic out. Iswtg people who keep arguing that Intel is never going to use N3 for their CPU tiles need to see this.
Also PTL using 18A in 2025 is confirmed by Intel

ISWTG? What does that mean?

 

[FlameTail]

.

 

[FlameTail]

Thanks for your objective view of Apple. How did AMD's launching of Turin Dense and Sarlak that based on N3E going?

Sarlak is on N3E? I thought 4nm.

 

[Geddagod]

ISWTG? What does that mean?

I swear to god
dunno where the w came from, but it just felt right lol

 

[FlameTail]

Can't tell if you're trolling or actually dumb enough to believe that.

Yeah. Very unusual thing to say, when analysts reported that Apple spent $1 Billion on just the tape out of M3, M3 Pro and M3 Max.

 

[FlameTail]

So LNL is using a 4P+4E configuration.

Rather unusual. That wouldn't be very strong in Multithreaded. Are we expecting huge ST gains?