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.)

 

[Geddagod]

If NovaLake is 16+32, I suspect there may be 8 P cores + a unit for rent. It would fit the codename.

If rentable units are true, which knowing it's coming from MLID means it's 50/50 lol, but in that case I guess it might mean having 16 "P" cores each with one thread that can use core fusion in pairs of 2 to get 8 "cores" and 8T. So in applications like games and such, you still have the 8T from the P-cores like ARL is likely to have, but each of those 8T are significantly stronger.
Again, this is so far into the future, and RYC looks to be such a large shift in architecture design, that it is kinda useless to discuss about it other than for fun and complete speculation. Which isn't to say it's not fun to do so, but who knows what's going to happen lol

If Intel is running 20A and Intel 3 in parallel I would assume they each have advantages and disadvantages or they would be going with one or the other? Why do you think these nodes are running in parallel production?

Think it's more to do with Intel not wanting to use 20a for server due to capacity constraints or risk.

 

[Geddagod]

Wonder what the successor to LNL in 2025 will be

 

[cebri1]

View attachment 86996
Wonder what the successor to LNL in 2025 will be

Isn't Panther Lake supposed to scale from mobile to desktop?

 

[Hulk]

Isn't Panther Lake supposed to scale from mobile to desktop?

According to MLID Panther will be mobile and Nova for desktop for 17th Gen.

 

[Ajay]

According to MLID Panther will be mobile and Nova for desktop for 17th Gen.

What principle process node will Intel 17th gen CPUs be manufactured on ?

 

[Geddagod]

Isn't Panther Lake supposed to scale from mobile to desktop?

Apparently not desktop anymore, but LNL is more special than the typical mobile platform of Intel's.
As in combined CPU + GPU tiles, and IIRC a SLC according to rumors as well.
So if PTL replaces it, then I expect either
a) the core + node to be so much better that it's still better than LNL with a worse tile setup
b) the tile setup has changes vs ARL
That's not to mention the SLC aspect

What principle process node will Intel 17th gen CPUs be manufactured on ?

So ARL is "15th gen", PTL "16th gen" so I'm assuming ur asking abt NVL?
Logically, it would make sense it's either as Intel 18A/variant or Intel 18A next. Also, by Intel 18A next, I expect a half decent node shrink, not the psuedo "shrinks" of Intel 4>Intel 3, or Intel 20a to Intel 18a.
Or tsmc/samsung lol

 

[Ajay]

So ARL is "15th gen", PTL "16th gen" so I'm assuming ur asking abt NVL?
Logically, it would make sense it's either as Intel 18A/variant or Intel 18A next. Also, by Intel 18A next, I expect a half decent node shrink, not the pseudo "shrinks" of Intel 4>Intel 3, or Intel 20a to Intel 18a.
Or tsmc/samsung lol

Well, that's what I was looking for. If Intel can get a good architecture together (better efficiency for one) and 18A turns out as well as is expected (since it's going to be a key foundry node). Then I really think they could be back in the game. If so, I might take a look at them, even though I've become an AMD fan. Of course, that's allot of ifs. Panther-Lake, given lead times, should be telling in terms of how well Intel's design teams are progressing (it should be from after the shake up that Keller started - too bad he didn't finish). Also, some rumors point to PTL being on I18A - this is probably predicated on Intel's comments that it has successfully accelerated the 18A node timetable.

 

[dullard]

View attachment 86996
Wonder what the successor to LNL in 2025 will be

That graphic is pretty much in line with the timelines that the rumors have been saying. But, I have two thoughts about the graphic.

1) If it starts with 2022, how old is this? We are almost in 2024. Is this an old graphic, or did someone just think it would be useful to include 2022 to display all of Alder Lake-U9 without displaying any other mobile chip?

2) WTF is Raptor Lake Refresh-H doing in 2025? Is Intel just trying to produce as much as they can in all lines to keep up with expected demand? Why not a Meteor Lake-H Refresh?

 

[jpiniero]

Why not a Meteor Lake-H Refresh?

OEM rejection is the most likely answer. If this is legit, it should confirm that Meteor Lake and even Arrow Lake volume won't be that high.

 

[igor_kavinski]

Consider a hypothetical 8+32 Raptor Lake part

I can see that being possible if Intel puts a restriction on moving threads between P-cores and E-cores. Think about it: if the Thread Director (or developer through hints to the Windows Scheduler) has determined that a job needs to run on the E-cores, then that's where it should run until completion. If that execution policy is implemented, the inter-core communication latency issue is bypassed and the E-core clusters can be worked on independently by different teams and maybe the E-cores can use the most cutting edge node with the smallest transistor sizes because they don't need to run at 5+ GHz. This should allow Intel to cram multiple really tiny E-core clusters adjacent to the P-core clusters and reach the 32 core count.

For very rare cases where the P-cores are too busy to take on an additional thread, there will be need for thread movement from E to P cores but once the P core is free, it gets the thread and finishes it and the thread never goes back to the E core.

 

[SiliconFly]

I can see that being possible if Intel puts a restriction on moving threads between P-cores and E-cores. Think about it: if the Thread Director (or developer through hints to the Windows Scheduler) has determined that a job needs to run on the E-cores, then that's where it should run until completion. If that execution policy is implemented, the inter-core communication latency issue is bypassed and the E-core clusters can be worked on independently by different teams and maybe the E-cores can use the most cutting edge node with the smallest transistor sizes because they don't need to run at 5+ GHz. This should allow Intel to cram multiple really tiny E-core clusters adjacent to the P-core clusters and reach the 32 core count.

For very rare cases where the P-cores are too busy to take on an additional thread, there will be need for thread movement from E to P cores but once the P core is free, it gets the thread and finishes it and the thread never goes back to the E core.

Having separate tiles for P cores & E cores sounds terrific, but I think it increases the cost of the CPU way too much. Rumors already suggest that Intel might reduce the tile count in future by merging the CPU tile with the SoC tile, but its unlikely considering it drastically increases the turn around time.

 

[igor_kavinski]

View attachment 86996
Wonder what the successor to LNL in 2025 will be

Where is that graphic from?

RPL-R mobile in 2025 means that Intel 7's yields are so good and Intel 4's are so bad that Intel has no choice. Seems converting existing Intel 7 fabs to the Intel 4 process is gonna be a really slow process. Also, Intel has a long way to go before they master the HVM of disaggregated tile arch coz they wouldn't be relying on Intel 7 so much otherwise. It's not like we still got supplemental 14nm stuff after ADL launched for satisfying volumes.

 

[dullard]

Where is that graphic from?

4:12 in this video: https://www.bilibili.com/video/BV1GC4y157D2/ If you can translate the rest, that would be great.

 

[SiliconFly]

View attachment 86996
Wonder what the successor to LNL in 2025 will be

But we're getting RPL refresh-refresh on Intel 10++++++ in 2025.

❤️ Intel ❤️

 

[Geddagod]

Panther-Lake, given lead times, should be telling in terms of how well Intel's design teams are progressing

Hmm, maybe PTL is going to be the first CPU designed in Intel's new "no more 56 steppings" plan where they actually try using more simulation before testing out actual silicon

WTF is Raptor Lake Refresh-H doing in 2025? Is Intel just trying to produce as much as they can in all lines to keep up with expected demand? Why not a Meteor Lake-H Refresh?

Perhaps it's a mistake in the chart, because I'm sure some variant of RPL will be sold alongside MTL. Doubt it has enough volume, at least at first.

If this is legit, it should confirm that Meteor Lake and even Arrow Lake volume won't be that high.

I mean kinda, but if MTL volume isn't that high, why are there no RPL parts in the graph with MTL, but there is RPL with ARL? Weird. I'm assuming it's a mistake, and they should have added RPL in the bracket with MTL.

 

[jpiniero]

I'm assuming it's a mistake, and they should have added RPL in the bracket with MTL.

I was thinking that too. But I suppose that could be explained by Intel originally planning on only doing a Raptor Refresh 2+8 die. OEMs will surely be able to buy the original Raptor 6+8 die but they might not get Raptor Refresh for that until the end of 2024.

 

[adroc_thurston]

WTF is Raptor Lake Refresh-H doing in 2025?

Cheap (same as RPL-282, really).

Why not a Meteor Lake-H Refresh?

Expensive and not really better.

 

[SiliconFly]

Isn't Panther Lake supposed to scale from mobile to desktop?

At this point, we only have the MLID info leak to go with. And it looks like, Panther Lake is more like Lunar Lake. A specialty line of processors that cater to low-power devices which usually doesn't concern us. Why do they even make it? 😫

It also mentions that the primary line of Intel CPUs for 2025 will be ARL refresh on 20A (and also possibly on N3B). But i doubt N3B.

 

[Tigerick]

iPhone 16 Series Chip Technology Moves to Volume Production

Apple has already given order commitments to chip supplier TSMC for its second-generation 3nm chip fabrication process known as N3E, which is expected to be used in all four models of next year's iPhone 16 lineup. TSMC's next 3nm node upgrade to N3E is less expensive and has improved yield...

forums.macrumors.com

Hmm, seems like Intel has not placed orders on N3B process and based on article Apple has 100% of N3B capacity...

 

[DrMrLordX]

Hmm, seems like Intel has not placed orders on N3B process and based on article Apple has 100% of N3B capacity...

That's what Macworld said maybe a month ago. Which is very interesting if anyone thinks that Arrow Lake-S will have a compute chiplet on N3B.

 

[FangBLade]

https://videocardz.com/newz/intels-...thread-performance-gain-leaked-slide-suggests

I can't wait to see Henry's excitement when he sees these results

 

[Gideon]

https://videocardz.com/newz/intels-...thread-performance-gain-leaked-slide-suggests

I can't wait to see Henry's excitement when he sees these results

That shouldn't really be news?

All it references is this month old post:

Page 105 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

Page 105 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

@uzzi38 look, you're on Videocardz!

 

[Saylick]

@Tigerick, you're famous too!

Intel Arrow Lake-S Desktop CPUs Deliver Up To 5% IPC & 15% Multi-Threaded Performance Uplift

Intel Arrow Lake-S Desktop CPUs are expected to feature a 5% IPC & 15% multi-threaded performance gain over Raptor Lake CPUs.

wccftech.com

 

[eek2121]

https://videocardz.com/newz/intels-...thread-performance-gain-leaked-slide-suggests

I can't wait to see Henry's excitement when he sees these results

FYI, Arrow Lake has a significant drop in power consumption. From recent leaks, it is more efficient than Zen 4. PL2 is said to be down to 177W.

If that 5%/15% is over Raptor Lake Refresh, that puts the chip well ahead of the competition in terms of both absolute performance as well as perf/watt.

Intel is also said to be launching Battlemage GPUs.

Will Zen 5 bridge the gap? Probably. I stand by my performance projections. Zen 5 and Arrow Lake will be extremely close in terms of performance. I won’t speculate perf/watt competitiveness, because there have been claims Intel is using TSMC for some parts, and that muddies the waters. I suspect Intel may end up with a slight edge, but we will see.

I will be considering Intel for my next upgrade if these rumors turn out to be true.

 

[eek2121]

That shouldn't really be news?

All it references is this month old post:

Page 105 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

Page 105 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

@uzzi38 look, you're on Videocardz!

@Tigerick, you're famous too!

Intel Arrow Lake-S Desktop CPUs Deliver Up To 5% IPC & 15% Multi-Threaded Performance Uplift

Intel Arrow Lake-S Desktop CPUs are expected to feature a 5% IPC & 15% multi-threaded performance gain over Raptor Lake CPUs.

wccftech.com

Welp, time to change your usernames. 🤣