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.



Comparison of upcoming Intel's U-series CPU: Core Ultra 100U, Lunar Lake and Panther Lake

Model	Code-Name	Date	TDP	Node	Tiles	Main Tile	CPU	LP E-Core	LLC	GPU	Xe-cores
Core Ultra 100U	Meteor Lake	Q4 2023	15 - 57 W	Intel 4 + N5 + N6	4	tCPU	2P + 8E	2	12 MB	Intel Graphics	4
?	Lunar Lake	Q4 2024	17 - 30 W	N3B + N6	2	CPU + GPU & IMC	4P + 4E	0	8 MB	Arc	8
?	Panther Lake	Q1 2026 ?	?	Intel 18A + N3E	3	CPU + MC	4P + 8E	4	?	Arc	12

Comparison of die size of Each Tile of Meteor Lake, Arrow Lake, Lunar Lake and Panther Lake

	Meteor Lake	Arrow Lake (20A)	Arrow Lake (N3B)	Arrow Lake Refresh (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Desktop Only	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2025 ?	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	8P + 32E	4P + 4E	4P + 8E
LLC	24 MB	24 MB ?	36 MB ?	?	8 MB	?
tCPU	66.48
tGPU	44.45
SoC	96.77
IOE	44.45
Total	252.15

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.)

 

[Exist50]

When next do we find out more about Intel 4 progress? Before end of year?

I doubt we'll get much more than they covered at VSLI '22. Maybe an update when MTL launches?

 

[DrMrLordX]

All considered, I would say Intel 4 is as good as TSMC 5/4nm. And that should be good enough for 2023- their main competitor, AMD, isn't shifting to 3nm until 2024, for zen 5. And even then, we also know zen 5 is using TSMC 4nm in some cases, so....

TSMC N5 is a 2020 node. Reaching parity with it three years later isn't really a good look for Intel on the fab side. While AMD is their main competitor in the x86 space, their fab competitors are TSMC and Samsung regardless of who uses their nodes. N5-family nodes will still have a massive cost and yield/maturity advantage over Intel 4 in 2023. 2023's Meteor Lake launch will be quite a teaching moment as to the true state of Intel 4. Pay careful attention to what SKUs they do and don't release on Intel 4, and keep in mind how many products have been moved to some other node (Granite Rapids, etc.)

 

[igor_kavinski]

I doubt we'll get much more than they covered at VSLI '22. Maybe an update when MTL launches?

Wouldn't their investors demand proof that they are on track?

 

[DrMrLordX]

Wouldn't their investors demand proof that they are on track?

Take a look at what investors have been doing since 2017 (and earlier).

 

[igor_kavinski]

Take a look at what investors have been doing since 2017 (and earlier).

Surely they have wised up to Intel's tricks by now.

 

[DrMrLordX]

Surely they have wised up to Intel's tricks by now.

Maybe. But you'll have to track stock analysts and their buy/sell ratings, etc. to get some idea of what the street thinks about Intel. Rest assured that Intel isn't going to be any more forthcoming about Intel 4 than they were 10nm, especially if things are going badly.

 

[Geddagod]

TSMC N5 is a 2020 node. Reaching parity with it three years later isn't really a good look for Intel on the fab side. While AMD is their main competitor in the x86 space, their fab competitors are TSMC and Samsung regardless of who uses their nodes. N5-family nodes will still have a massive cost and yield/maturity advantage over Intel 4 in 2023. 2023's Meteor Lake launch will be quite a teaching moment as to the true state of Intel 4. Pay careful attention to what SKUs they do and don't release on Intel 4, and keep in mind how many products have been moved to some other node (Granite Rapids, etc.)

I believe the main nodes of IDM 2.0 that Intel is pushing is Intel 16?, Intel 3, and Intel 18A. Those nodes seem to be pushed because they are the further iterations of the node behind them, rather than an entirely new node. Maybe they aren't pushing Intel 4 much for IDM 2.0 could be cost, sure, but also because Intel 4 doesn't even have complete libs, it's just high perf cells.
This also connects to why Granite Rapids might have been pushed to Intel 3- not necessarily mainly because of the progress of the Intel 4 node itself. We have a couple reasons here:
In complete libraries for Intel 4 (Dr. Ian Cutress theory)
SPR delays means that SPR 2023, EMR 2023 end/ 2024 early?, GNR HAD to be pushed back regardless, and considering Intel 3 is a reiteration of Intel 4, why not just push GNR to a newer node if it's going to delayed a year or more regardless?
GNR got 'redefined'. New architecture+node as well. AMD themselves (Norrad) said they expected genoa to compete against GNR, but it's pretty clear now that GNR is going to compete against Zen 5 server products. GNR had to be redefined and improved somehow...
Intel also doesn't produce small die pipe cleaner mobile chips to sell at large scale as well, so for them yields and perf have to be better earlier than TSMC and Samsung for the same node.
What really is a bad look though is that despite Intel 4 just catching up pretty late to the party, is the fact that if Intel executes, they can catch up to TSMC relatively quickly, in the next 3 years or so.
For Intel's means and purposes, Intel 4 is pretty competitive. Especially since, even as a foundry, Intel doesn't even seem to be really pushing Intel 4.

 

[Doug S]

Surely they have wised up to Intel's tricks by now.

Most Intel investors, or investors in most large tech companies (other than the few with a big retail component like Tesla) are fund managers. They don't have the bandwidth to go in depth enough to 'wise up' to Intel's tricks. If they did they would have seen the writing on the wall with 10nm with that paper launch in 2018, but Intel was still putting out terrific quarters for several years after that. The covid laptop buying frenzy gave them an extra year before the warts really started to get exposed. I used to think a lot of that chart watching stuff like 50 DMA crossing 200 DMA meaning X was bull, but it turns out a lot of it is true because the fund managers subscribe to it and will change allocations based on such chart action!

Institutional investors mostly react to Intel's results and price moves, and their public statements like saying they will be negative cash flow for the next few years. They aren't doing comparative assessments of Intel's process roadmap against TSMC's and trying to determine how accurate each will turn out to be to see what Intel's 2026 profits will be. They aren't looking anywhere near that far ahead, so that stuff simply doesn't matter to them. If Intel was to regain process leadership, or even parity, it would improve their profitability and that is what would drive up their stock price and that's what would bring the institutional money back in.

I've been low key been watching Intel for the past 18 months or so trying to find a good entry point. The stock has FINALLY reached a point where I feel it is fairly priced, but I'm hoping for some overreaction to the downside based on bad quarters they will continue to have while watching for clues they might address their process issues. People paying attention as closely as we are here will see those signs long before they create effects that will drive the stock price up. I'm thinking I should at least be able double up over 2-3 years if they right the ship, but if I can get in a beaten down price (say 30% lower than today's) I'll make money even if they continue to stumble.

 

[moinmoin]

The stock has FINALLY reached a point where I feel it is fairly priced, but I'm hoping for some overreaction to the downside based on bad quarters they will continue to have while watching for clues they might address their process issues.

Macro aside I'm expecting Intel to fall some more if only for the fact that even once process issues are said to be addressed the results as in product output and financial results come significantly later.

 

[ericlp]

hmm, well, the stock market is bouncing up and down, but, I feel more pain is on the way. I will hold out ... An old quote, when the market crashes, cash is king. Intel is probably a safe bet, but I'd diversify and look at Nvidia, Qualcomm, Apple, TCSM, etc...etc...

 

[DrMrLordX]

I believe the main nodes of IDM 2.0 that Intel is pushing is Intel 16?, Intel 3, and Intel 18A.

Intel competes vs. TSMC and Samsung with all their nodes, not just the ones featured by IFS 2.0 (not IDM 2.0; IDM stands for Integrated Device Manufacturer).

 

[Exist50]

Intel competes vs. TSMC and Samsung with all their nodes, not just the ones featured by IFS 2.0 (not IDM 2.0; IDM stands for Integrated Device Manufacturer).

"IFS 2.0" isn't a term they've used.

 

[moinmoin]

"IFS 2.0" isn't a term they've used.

Maybe not Intel, but the press ran with it.

Edit: Why can't this board escape quotes in search strings on first try instead cutting them off??

 

[Geddagod]

Intel competes vs. TSMC and Samsung with all their nodes, not just the ones featured by IFS 2.0 (not IDM 2.0; IDM stands for Integrated Device Manufacturer).

oop sorry IFS 2.0
Regardless IFS is not set to be a major aspect of Intel, at least for a long while. The core of Intel is still chips, and so saying Intel 4 being a competitive node is definitely fair, since their major competitors for their core business would be on a similar node. Intel doesn't really make small chips for electronics such as smart phones, where the die size is much smaller, meaning that when they start HVM manufacturing of a node, they might need higher yields/better clocks than what TSMC and Samsung might need.
This is all just pedantic anyway

 

[Exist50]

Maybe not Intel, but the press ran with it.

If you tried linking something, it didn't work. Think you might be mixing IFS up with IDM 2.0.

 

[moinmoin]

If you tried linking something, it didn't work. Think you might be mixing IFS up with IDM 2.0.

Board was cutting the search string at the quotes. Link is https://www.google.com/search?q=intel+"ifs+2.0" with plenty results.

 

[DrMrLordX]

Intel doesn't really make small chips for electronics such as smart phones

For cell phones? No, not since they got out of the modem business. They do make some small chips though.

This is all just pedantic anyway

Fair enough.

 

[IntelUser2000]

Intel also doesn't produce small die pipe cleaner mobile chips to sell at large scale as well, so for them yields and perf have to be better earlier than TSMC and Samsung for the same node.

Their highest volume products are Core i3 class parts. For example back in 2017, the dual core 100mm2 chips were the majority of sales. Yes, Core is highest volume client category, not Pentium/Celeron since a decade ago, but you are still talking about the lower end.

That could also be the reason why Alderlake volumes are lower than expected, because the Core i3 based laptops weren't widespread until early summer.

For Alderlake that's 1210U/1215U/1220P.

I wouldn't call Apple's 100mm2+ chips "small" either. Likely the higher end of Apple chips are larger than mainstream Intel chips like I outlined above.

Of course we're talking about TSMC, so in regards to volume then you have to count lower end mobile chips in the 70-90mm2 range from Samsung, Qualcomm, and Mediatek.

(Note I'm disregarding desktops here. They are quite lower volume and also more competitive than on mobile)

 

[Exist50]

Their highest volume products are Core i3 class parts.

You sure about that? Thought it was the i5s.

 

[poke01]

For cell phones? No, not since they got out of the modem business. They do make some small chips though.

very big mistake selling modem business. Apple got a great buy.

Modems are also used in tabs and laptops. If Apple release a celluar macbook business folks will be all over it.

 

[IntelUser2000]

You sure about that? Thought it was the i5s.

Where did you hear that? They managed to move it to Core sometime in the Sandy Bridge era, but late as 2017 it was Core i3. The dies might be small, but the volume such they can almost justify cutting down higher chips in purpose to sell as Core i3, and the yield needed to be very high. Dual Core GT2 was the king for the longest time even during the quad core Kabylake R era.

Also, this might be anecdotal, but the difference in mindset between "us" and average joes are huge. Core i5 might seem like common thing for us, but wait until you are surrounded by people who are average joes and family man types. Then the price range for laptops fall squarely in the Core i3 range. They spend enough not to go down into the Pentiums of course.

 

[Geddagod]

Their highest volume products are Core i3 class parts. For example back in 2017, the dual core 100mm2 chips were the majority of sales. Yes, Core is highest volume client category, not Pentium/Celeron since a decade ago, but you are still talking about the lower end.

That could also be the reason why Alderlake volumes are lower than expected, because the Core i3 based laptops weren't widespread until early summer.

For Alderlake that's 1210U/1215U/1220P.

I wouldn't call Apple's 100mm2+ chips "small" either. Likely the higher end of Apple chips are larger than mainstream Intel chips like I outlined above.

Of course we're talking about TSMC, so in regards to volume then you have to count lower end mobile chips in the 70-90mm2 range from Samsung, Qualcomm, and Mediatek.

(Note I'm disregarding desktops here. They are quite lower volume and also more competitive than on mobile)

That may be true, but to my knowledge they don't make i3s first and then sell the rest of the lineup at the same time, they make the higher end skus of a product stack and then sell the lower end parts later. So I think Intel still has to have higher perf/yield expectations for them to launch a product using a node compared to TSMC and Samsung, who can first release mobile chips, and then have customers buying larger and larger chips as yields improve. I will admit though that this is just MY speculation, could be wrong.

 

[Doug S]

very big mistake selling modem business. Apple got a great buy.

Modems are also used in tabs and laptops. If Apple release a celluar macbook business folks will be all over it.

Who needs a cellular modem in their laptop or tablet these days, when they can tether to their phone?

Before most carriers allowed tethering and/or when type phone data plans were measured in MB instead of GB this was more useful, but even then a modem in a USB stick makes more sense than building it in - because the way Qualcomm charges for patent licenses based on a percentage of device price the OEM's payments to them are much less for cellular in a USB LTE/5G modem cellular built in to a $1000 laptop or tablet.

The time for this to matter has passed - so much so I would bet against Apple ever offering a Mac with built in cellular. They won't be building the modem into Apple Silicon like they will the iPhone SoC.

Intel didn't lose anything selling their modem business to Apple, because Apple was the only possible customer for a discrete smartphone modem. Everyone else used an SoC with a modem built in.

 

[IntelUser2000]

So I think Intel still has to have higher perf/yield expectations for them to launch a product using a node compared to TSMC and Samsung, who can first release mobile chips, and then have customers buying larger and larger chips as yields improve. I will admit though that this is just MY speculation, could be wrong.

I think so too, but there's also the mentality of "they'll never launch" because some have the forever bear mentality and that they'll never ever escape the 10nm issue and it'll be doomed forever. Some of that might be hope, some might be just refusal to change mindset.

 

[Exist50]

Where did you hear that?

Where did you hear most of the volume is i3s?

I think so too, but there's also the mentality of "they'll never launch" because some have the forever bear mentality and that they'll never ever escape the 10nm issue and it'll be doomed forever. Some of that might be hope, some might be just refusal to change mindset.

I think we've seen several times over how this cycle goes... I wouldn't bring expect logic to sway an emotional position.

Though Meteor Lake certainly isn't likely to do Intel's reputation any favors, regardless of the health of Intel 4.