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

Tigerick · Aug 22, 2022

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	12 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 (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	8P + 16E	4P + 4E	4P + 8E
LLC	24 MB	36 MB ?	12 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.)

dullard · Oct 14, 2024

MS_AT said:
Different business model. Note that ARM never made any chips, and doesn't plan to make them afaik. They are selling designs. Intel was forced to license x86 to AMD to meet a supply contract for someone, I no longer remember who, so both companies would be able to make actual chips. Then AMD came up with x64 that Intel is licensing and they are locked in this weird duopoly, though there are few other companies that have/had x86 license like VIA. Since both AMD and INTEL were able to produce the actual CPUs it made little sense for them to share that cake with others back then.

It was IBM that forced the weird agreement.

IBM had a requirement for a secondary supplier for all components. That way, if one supplier had an issue, IBM's business is not severely impacted. Lots of companies have this requirement. In order to sell to IBM, which was to be a huge deal for Intel, Intel needed to find a secondary supplier of x86. So, Intel gave everything over to AMD: rights to use x86, microcode to run the chips, how to make and design the chips, etc. The whole secret sauce.

Over the years, there have been court battles. The main result is that courts have decided that the original agreement was a reciprocal agreement. Meaning that both Intel and AMD had to at least have the option to share the technology that either company develops. What isn't shared is trademarks. That is why we went from numbers (8086, 486, etc.) to marketing names.

Intel also only needed one secondary supplier. Thus, AMD's agreement to use Intel's technology is non-transferrable. Although ARM is showing that it can be quite profitable to just sit back and skim off the profit when other companies take the financial risk to make and sell ARM chips.

511 · Oct 14, 2024

x86 is just complex IP Licencesing agreement It was monopoly for a short time lol it's a duopoly now we had multiple company having x86 license before via iirc also Noyce and Sanders were friend one of the reason they got x86 licence

On a sidenot Will have a secondary CUDA Supplier when will this monopoly be broken 🤣 won't the customers force Hand of Huang?

Wolverine2349 · Oct 14, 2024

MS_AT said:
Different business model. Note that ARM never made any chips, and doesn't plan to make them afaik. They are selling designs. Intel was forced to license x86 to AMD to meet a supply contract for someone, I no longer remember who, so both companies would be able to make actual chips. Then AMD came up with x64 that Intel is licensing and they are locked in this weird duopoly, though there are few other companies that have/had x86 license like VIA. Since both AMD and INTEL were able to produce the actual CPUs it made little sense for them to share that cake with others back then.

Why doesn't VIA make any X86 chips? Or at least any that are worth while or powerful.

They had the Cyrix like 20 years ago I remember, but it was weak at a time of Athlon 64 ad Northwood P4.

GTracing · Oct 14, 2024

MS_AT said:
Different business model. Note that ARM never made any chips, and doesn't plan to make them afaik. They are selling designs. Intel was forced to license x86 to AMD to meet a supply contract for someone, I no longer remember who, so both companies would be able to make actual chips. Then AMD came up with x64 that Intel is licensing and they are locked in this weird duopoly, though there are few other companies that have/had x86 license like VIA. Since both AMD and INTEL were able to produce the actual CPUs it made little sense for them to share that cake with others back then.

Random thought I had reading this. Is x86_64 the last instruction extension from AMD? If so, since the patent on that has expired, could Intel license out newer instructions extensions like AVX2?

gdansk · Oct 14, 2024

GTracing said:
Random thought I had reading this. Is x86_64 the last instruction extension from AMD? If so, since the patent on that has expired, could Intel license out newer instructions extensions like AVX2?

It's the same reason no one makes unlicensed binary-compatible ARM32 cores despite there being a substantial market for that. These companies have other patents that cover the normal way of implementing certain features.

MS_AT · Oct 14, 2024

Wolverine2349 said:
Why doesn't VIA make any X86 chips? Or at least any that are worth while or powerful.

They had the Cyrix like 20 years ago I remember, but it was weak at a time of Athlon 64 ad Northwood P4.

http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/zhaoxins-zx-f-kx-7000-kh-40000-and-beyond.2564975/ if I am not mistaken they are in joint venture with VIA.

GTracing said:
Random thought I had reading this. Is x86_64 the last instruction extension from AMD? If so, since the patent on that has expired, could Intel license out newer instructions extensions like AVX2?

Probably someone more knowledgeable would need to jump in here, but since x64 was introduced AMD and Intel entered into multiple cross-licensing agreements so if you are asking if AVX2 could be licensed by Intel to AMD for additional fee, it is already licensed via these. If you mean could they license AVX2 to anyone? I don't know, but AVX2 seems too tied into x64 to be useful without it.

GTracing · Oct 14, 2024

MS_AT said:
Probably someone more knowledgeable would need to jump in here, but since x64 was introduced AMD and Intel entered into multiple cross-licensing agreements so if you are asking if AVX2 could be licensed by Intel to AMD for additional fee, it is already licensed via these. If you mean could they license AVX2 to anyone? I don't know, but AVX2 seems too tied into x64 to be useful without it.

More of the latter. I guess my question makes more sense in terms of x86 emulation. As I understand it, random companies still can't make x86 chips, but they can emulate x86 code in software.

For example, could Intel license out AVX2 to allow Microsoft to emulate it? Can they do it without AMD's involvement or approval? Or if Intel decides to create a new ISA, can they emulate x86 with AVX10 and APX? Allowing them to keep backwards compatibility and the x86 "moat" while creating a new and improved ISA.

poke01 · Oct 14, 2024

GTracing said:
For example, could Intel license out AVX2 to allow Microsoft to emulate it?

AVX2 patent expired this year. You don’t need a license to emulate AVX2

Doug S · Oct 14, 2024

poke01 said:
AVX2 patent expired this year. You don’t need a license to emulate AVX2

When Apple added AVX2 support to Rosetta 2 this was mentioned by some as the reason it didn't have it originally. Interestingly, I recall reading where someone stated that the SSE2 patents didn't expire until 2020, which just so happens to be when Apple Silicon Macs were launched. Since SSE2 was a base requirement to meet the x86-64 spec, he suggested the timing of the launch may have been based on that patent expiration, not waiting on Apple Silicon performance to be good enough.

OneEng2 · Oct 14, 2024

Doug S said:
Doesn't it have a lot to do with expectations? There were some hyping Zen 5 seeing a 40% increase in IPC, and when it only hit managed the 15% AMD had actually promised it was a "failure" because of the inflated expectations. On the other hand there was a lot of talk that Apple was washed up after GW III and friends left and some people were claiming they would get a 0% IPC increase - just keep re-using basically the same core, running on a better process to get more frequency as the only performance increase. Then M4 hit (early) and it was seen here as anything but a failure, because of those very low expectations.

A few months ago there wasn't much faith Intel's next core would increase performance much (but it wouldn't matter anyway because Zen 5's 40% IPC gain would make it irrelevant) Then rumors started hitting about how much of a performance gain was coming in the E core, so expectations were raised to the point where a 9% IPC increase is now seen as a "failure" by some. Especially because too many on this board believe gaming performance is the only thing that really matters, and were horrified to discover it cut frequency and their precious games wouldn't get any faster! As someone else mentioned earlier in this thread, Intel is targeting laptops and dense servers, because that's where the money is, the gaming market is an afterthought financially.

Me. Laptops and data center is where the volume and growth are, not desktop, and certainly not high end desktop/gaming.

GTracing said:
Where did you see 5.05mm for redwood cove? The only place i see that is a reddit post that doesn't cite a source. The one die shot annotation I've seen says 5.33mm^2 for redwood cove in meteor lake. That's 28% or 22% bigger depending on which number is correct. Even 22% is a sizeable difference. Plus Strix Point has a decent performance lead over Meteor Lake.

Any way you slice it, Intel's P cores have bad PPA.

Can't find it, but still, it isn't like it is double. Noted that it is bigger though and does not best Zen 5 at a similar density.

Thunder 57 said:
I think you need to read up on Zen 1. It competed with Intel's 6900k which was HEDT at the time BTW. In gaming and more single threaded things Intel still had the upper hand beacuse they had Skylake and Kaby Lake out. Those were generations ahead of Broadwell that Zen 1 competed against. Zen just wasn't great at gaming until Zen 2, and truly until Zen 3.

From what I read of reviews when it was released, it did very poorly in games compared to Intel (but way better than Bulldozer), and was generally in the ball park (20-40% lower) than the best Intel offerings of the time. Sadly, this was a HUGE improvement over Bulldozer. My point is that a brand new architecture from the ground up may not show its chops in its first release. I am fairly confident that Lion Cove and Skymont implementations this release will improve greatly over the next few years.

DavidC1 said:
@OneEng2 Yes and Lion Cove doesn't do anything special. Skymont has far more room to grow.

And it has nothing to do with me being grumpy. You said I could have "pounded you more" correct? No, you were incorrect and I merely corrected you.

Because while 9% gain itself is okay, they aren't operating in a vacuum. It's competing against AMD, which has a significantly smaller core and achieves the same thing. I would think very much so the engineers and managers are not happy needing N3B to do something the competition could do with cheaper and easier to design N4P. At least in the Raptorlake generation it had a clock advantage!

It's competing with the team within the same company, and the core that was merely an "Atom" core and almost considered a meme is now threatening it's very existence.

And it's competing with the ever growing ARM threat. Compared to the off the shelf ARM cores it's merely equal to Cortex X3 or X4, and that isn't even their latest.

The excuses continue to grow though. Way way back, the excuse was that ARM couldn't reach x86 levels of performance, because it's meant for "low power". It's beating x86 cores per clock, and some are beating it in absolute performance against everything except the 125W -S chips. What excuses are there now?

What about the excuse that the x86 and ARM ecosystems are different? It's even hard to accept this because x86 has been living in it's own RDoF, the same field that Apple enthusiasts are said to have. And it has been protected many many times by flawed law and lawyers. You want true competition? Then x86 should have opened up completely 30 years ago. That's competition.

Aren't some people saying already the Windows compatibility doesn't really matter and WoA will just take over the market? Look at the Qualcomm thread, that's what they claim!

Intel/AMD both need to wake up and realize their cushy x86 bubble will not last. P cores, especially Intel P cores is a laughable attempt.

Every since x86 processors were pipelined and superscalar, it has necessitated that all the instructions be equal length vs x86 variable length instruction and variable length data. As a result, it is my understanding of cpu architecture that x86 instructions are decoded into RISC style equal length instructions prior to going into the pipeline.

In other words, there really isn't much fundamental difference between x86 and RISC architectures today. x86 continues to exist only because of backwards compatibility with decades of code.

Finally my young jedi, you are greatly mistaken about a great many things. Turin dense using Zen 5c is knocking the pants off of its arm competition. See the latest benchmarks. Last time I checked, Zen 5c is still an x86 chip.

It would be more accurate to say that ARM chips showed Intel AND AMD that performance within a price and power envelope was worth designing for and was, in fact, the future. It seems to me that both companies are on track to have taken that lesson to heart.

Hulk said:
Wow, lots of hating on Intel suddenly. Pat was the lead architect for the 486, he knows a lot about microprocessor design, no doubt about that. Intel is huge and has a lot of moving parts. I'm sure there are huge tradeoffs/risks to be had at every decision to be made. I would be careful armchair quarterbacking and being so sure you are somehow wiser than Pat and others at Intel. Rerunning the plays correctly is always easier after they've been run. Also, keep in mind that when negative inside information comes out about a company it's usually from disgruntled employees to I'd take it with a grain of salt. In addition, it's not like Pat was handed a company in good shape and at the top of their game, it was after 5 or so years of Skylakes and ++++++++++ processes. Since he's been there we've seen Alder Lake, Raptor Lake, Raptor Lake Refresh (yes I know that's not really an accolade), Meteor Lake, Lunar Lake, and now Arrow Lake. I realize most of those were in the pipe long before he got there but there has been quite a bit of architectural "action" since he got there.

Before we shovel the dirt into Intel's grave and write about all of the stupid things they did that were obviously stupid and how we could have done better let's wait for ARL reviews.

I have personally shoveled that dirt before myself (Netburst and Itanic days) and learned my lesson. Intel is not beyond pulling a rabbit out of its hat when needed ...... and as I stated, Lion Cove and Skymont are no where near the disaster Netburst and Itanic were. Lets see how Arrow Lake for laptops and the 18A next gen designs turn out before we start ordering the head stone for Intel.

DavidC1 · Oct 14, 2024

@igor_kavinski You seem to be letting your emotions control you. It didn't pan out as the most optimistic projections, but they are going through a lot. Wishing death indicates that you probably should get off the keyboard and enjoy the outside.

GTracing said:
More of the latter. I guess my question makes more sense in terms of x86 emulation. As I understand it, random companies still can't make x86 chips, but they can emulate x86 code in software.

Transmeta went through the hoops of making a dedicated software/hardware translation layer to make it perform. They still have to agree to Intel that they would never do that again.

Nvidia had to settle with Intel regarding Denver, even though again it was translated.

This is what I mean when I say x86 should have been opened, so anyone can develop it. Transmeta nor Nvidia would have needed to go through bothering with translation layers and could have just created a competitor to Intel.

The performance of the original Crusoe sucked but they did a real good job of creating a low power core. Of course they had to deal with compatibility issues too. Imagine what they could have done without the performance killing layer and with compatibility issues?

@OneEng2 Skymont is actually pretty damn good. Lion Cove is veering into bad territory. 50% extra core for same clock and perf per clock and needing more power than competition is proof. The problems are tamed because it's much more conservative, but that's like comparing a table on fire versus the whole kitchen.

Lion Cove design is a dead end. Skymont derivatives will carry on.

Hulk · Oct 14, 2024

Wolverine2349 said:
Why is it that so many can make ARM chips when ARM is owned by ARM.

Where as only intel and AMD can make X86 chips.

I mean Qualcomm and Apple and others can make ARM chips despite ARM arch being owned by ARM, X86 was owned only by intel and AMD hacked it and somehow got a license through the courts with Intel.

Why does ARM license and willing to so many other companies than Intel or AMD will allow X86 to go?

If I remember correctly AMD got in trouble because they were copying exactly Intel CPU's. Once they started on their own designs they were fine from a legal point of view.

Lots have tried to make x86 CPU's and lots of failed. That was 25 years ago, now it's even hard to make up the lost ground. IBM tried. Cyrix has some decent parts if you didn't care about fp performance. VIA tried. NexGen had some hype for 5 minutes. Those are the ones I remember.

It's the same reason you don't see new automobile (ICE) manufacturers being able to compete with GM, Ford, Chrysler or whatever they're called, Toyota, Honda, Nissan, the ones who have been around forever. Hyundai is the last one to be able to pull if off and it was touch-and-go for a while but the South Koreans are amazingly industrious. Remember the Suzuki jeep thing? Ain't gonna see one of those in the US.

Now I know what you are thinking. What about all of the start up electric vehicles? Yes, there was a gap in the marketplace that Tesla was the first to really exploit. Others, like GM tried before the tech was ready. Their first attempt used lead acid batteries! But many of these electric start ups have already failed or in the process of failing as the big companies gear up.

Hulk · Oct 14, 2024

OneEng2 said:
Me. Laptops and data center is where the volume and growth are, not desktop, and certainly not high end desktop/gaming.

Can't find it, but still, it isn't like it is double. Noted that it is bigger though and does not best Zen 5 at a similar density.

From what I read of reviews when it was released, it did very poorly in games compared to Intel (but way better than Bulldozer), and was generally in the ball park (20-40% lower) than the best Intel offerings of the time. Sadly, this was a HUGE improvement over Bulldozer. My point is that a brand new architecture from the ground up may not show its chops in its first release. I am fairly confident that Lion Cove and Skymont implementations this release will improve greatly over the next few years.

Every since x86 processors were pipelined and superscalar, it has necessitated that all the instructions be equal length vs x86 variable length instruction and variable length data. As a result, it is my understanding of cpu architecture that x86 instructions are decoded into RISC style equal length instructions prior to going into the pipeline.

In other words, there really isn't much fundamental difference between x86 and RISC architectures today. x86 continues to exist only because of backwards compatibility with decades of code.

Finally my young jedi, you are greatly mistaken about a great many things. Turin dense using Zen 5c is knocking the pants off of its arm competition. See the latest benchmarks. Last time I checked, Zen 5c is still an x86 chip.

It would be more accurate to say that ARM chips showed Intel AND AMD that performance within a price and power envelope was worth designing for and was, in fact, the future. It seems to me that both companies are on track to have taken that lesson to heart.

I have personally shoveled that dirt before myself (Netburst and Itanic days) and learned my lesson. Intel is not beyond pulling a rabbit out of its hat when needed ...... and as I stated, Lion Cove and Skymont are no where near the disaster Netburst and Itanic were. Lets see how Arrow Lake for laptops and the 18A next gen designs turn out before we start ordering the head stone for Intel.

Yes. Neburst was literally smoked by the Athlon. The less said about titantic the better...

AMD was/is definitely the comeback kid and an amazing success story!

The reality of the situation is that it is looking very clear that both ARL and Zen 5 will be great products and very competitive with one another. We keep hoping for a KO from one or the other but they remain evenly matched.

Jan Olšan · Oct 14, 2024

Wolverine2349 said:
Why doesn't VIA make any X86 chips? Or at least any that are worth while or powerful.

They had the Cyrix like 20 years ago I remember, but it was weak at a time of Athlon 64 ad Northwood P4.

The design company (Centaur Tech) architecting their cores was a relatively small team. The were only able to design a totally new CPU core architecture once per 5-10 years, plus the never got the chops and resources to be able to go for a hich-frequency design (and by that I don't mean Netburst, but a physical design-optimised core that will clock decently, which is always a requirement to be competitive).
To make matter worse, VIA same as Zhaoxin also tended to deliver later, but the core issue was lack of resources, likely.

Look at the story of Centaur CHA / CNS as the latest illustration. Comes late (years since anouncements) or never, and kept down by clocks slightly above 2 GHz.

gdansk · Oct 14, 2024

There is a documentary about Centaur called Rise of the Centaur which was somewhat informative and entertaining. But since then Henry retired and the team was sold to Intel and the IP licensed to Zhaoxin.

I guess they were in the same city as the Intel 'mont team at the time too?

DavidC1 · Oct 14, 2024

Hulk said:
Lots have tried to make x86 CPU's and lots of failed. That was 25 years ago, now it's even hard to make up the lost ground. IBM tried. Cyrix has some decent parts if you didn't care about fp performance. VIA tried. NexGen had some hype for 5 minutes. Those are the ones I remember.

Via made such a chip at a small fraction of Intel's cost and team size. And it was a time when cheap computers started having a market.

Transmeta on the other hand made a pretty damn good low power chip despite needing a translator layer. Actually despite being on the older 90nm process it had power profile similar to the original 45nm Atom. Yes, the Atom team came a long, long way since then.

But let's say they didn't have to emulate?

I also know regarding Denver Intel was quite threatened by Nvidia. A few times they said it public on IDF that Nvidia was their real threat. They were right of course. It would have made them shudder if they were allowed full development. Here's the AT piece on it: https://www.anandtech.com/show/4122/intel-settles-with-nvidia-more-money-fewer-problems-no-x86/2

Here's part of the settlement and what Nvidia doesn't get.

x86 License, Including Rights To Make an x86 Emulator

Can't even make an emulator. Tell me there's no legal shenanigans and why Intel and x86 was a "baby" being protected by legal entities.

Who knows? Maybe x86 would have had a way bigger low power market decades ago if it had TRUE competition.

Hulk · Oct 14, 2024

We're off topic but this reminds me of 3dfx. They had it all and lost it all.

DavidC1 · Oct 14, 2024

I'm waiting for the latest x86 optimization manual updated to include Lion Cove and Skymont. Real juicy details are in there.

OneEng2 · Oct 14, 2024

DavidC1 said:
@igor_kavinski You seem to be letting your emotions control you. It didn't pan out as the most optimistic projections, but they are going through a lot. Wishing death indicates that you probably should get off the keyboard and enjoy the outside.

Transmeta went through the hoops of making a dedicated software/hardware translation layer to make it perform. They still have to agree to Intel that they would never do that again.

Nvidia had to settle with Intel regarding Denver, even though again it was translated.

This is what I mean when I say x86 should have been opened, so anyone can develop it. Transmeta nor Nvidia would have needed to go through bothering with translation layers and could have just created a competitor to Intel.

The performance of the original Crusoe sucked but they did a real good job of creating a low power core. Of course they had to deal with compatibility issues too. Imagine what they could have done without the performance killing layer and with compatibility issues?

@OneEng2 Skymont is actually pretty damn good. Lion Cove is veering into bad territory. 50% extra core for same clock and perf per clock and needing more power than competition is proof. The problems are tamed because it's much more conservative, but that's like comparing a table on fire versus the whole kitchen.

Lion Cove design is a dead end. Skymont derivatives will carry on.

Agree. x86 should have been opened. Intel has been hiding behind (IMO) that license, and creating new instruction sets (and their accompanying software/firmware optimizations) that they can have exclusively before AMD (and others now) are able to implement them as others don't get the specifications or roadmap ahead of time and Intel keeps all the software companies that get the heads up quiet with NDA's. As Intel loses market share, this phenomena should change as it will no longer make as much sense for OS and software companies to do up-front work to utilize instruction sets that are only supported by Intel (I could be wrong on this one though).

Not certain I agree that Lion Cove is dead. I am pretty sure that Skymont architecture as a base wont have the chops to take on full sized Zen 5 any time in the near future (years). I am guessing a modified Lion Cove will show great improvement in the next round.

Hulk said:
Yes. Neburst was literally smoked by the Athlon. The less said about titantic the better...

AMD was/is definitely the comeback kid and an amazing success story!

The reality of the situation is that it is looking very clear that both ARL and Zen 5 will be great products and very competitive with one another. We keep hoping for a KO from one or the other but they remain evenly matched.

I mostly agree. Seems like AMD still hold the data center crown by a decent margin .... just not the "night-and-day" situation they had the past couple of years.

While it is costing Intel their margins, ARL looks to be pretty competitive to Zen 5. Intel is definitely looking good in the thin and light market with Lunar Lake.

Hulk said:
We're off topic but this reminds me of 3dfx. They had it all and lost it all.

Hard to fathom for sure. They literally created the 3D graphics card industry (I owned one for quite some time).

Back on topic, Intel hasn't just screwed the pooch on their stalled process technology with the 10+++++++ silliness, they also missed the boat with AI .... and integrated graphics. They got dragged into 64 bit by AMD, and were also dragged into multi-core by AMD (on the desktop). They tried to finagle the entire industry into paying license fees for RAMBUS memory and then had to drop back to DDR when that finally failed miserably.

I see Lion Cove and Skymont as a continuation of sanity (after abandoning Netburst for Core 2, adding 64 bit, and going back to DDR). This next gen architecture moves to focus on efficiency at an architectural level. This should allow Intel to create a competitive high core product in the data center, as well as less expensive processors for the desktop and laptop markets moving forward.

I don't think Intel's current ARL is going to displace Zen 5 in the desktop, or Turin in the server (yet), but it is thinking in the right direction IMO.

I am not sure about the Intel tile idea though (as compared to AMD's Chiplets). Seems like that extra silicon is going to cause some inefficiencies in price. At any rate, Intel is going to have a few growing pains while they iron out how to handle the chip to chip latencies that they didn't have to deal with on their monolithic desktop designs IMO.

I certainly am not buying into the "Intel is doomed" line of thinking. That is just silly thinking. Just to bring in politics for a moment here ......

There is no way on Earth with the current "America First" movement, and the "De-risk" movement away from China, and the "Tariff everything" mentality in the USA right now that the US government would let Intel die. Not a chance. It will likely be a decade if ever before it would become politically allowable again for something like that to happen.

Hulk · Oct 14, 2024

I wonder if Intel considered something like 2+10 for Lunar Lake. Given the huge increase in IPC for Skymont coupled with the fact that the clockspeed advantage of Lion Cove is quite small in this segment, 2+10 seems like it could have been a nice configuration and probably even a bit less area then 4+4.

gdansk · Oct 14, 2024

Hulk said:
I wonder if Intel considered something like 2+10 for Lunar Lake. Given the huge increase in IPC for Skymont coupled with the fact that the clockspeed advantage of Lion Cove is quite small in this segment, 2+10 seems like it could have been a nice configuration and probably even a bit less area then 4+4.

Given that its predecessor in the low power segment was 2+8 RPL-U I'm sure they considered doing 2+8 again.

Hulk · Oct 14, 2024

What are the advantages and disadvantages of Skymonts 3x3 decoder vs. Lion Cove's 8 simple decoders?

dttprofessor · Oct 14, 2024

Hulk said:
What are the advantages and disadvantages of Skymonts 3x3 decoder vs. Lion Cove's 8 simple decoders?

More wide, more simple,more future!

dttprofessor · Oct 15, 2024

Epyc 9755 vs xeon 6980p, tsmc n4p vs intel 3 , intel p core looks not so bad except size. E core has future.

dttprofessor · Oct 15, 2024

Sorry, the same.

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

Senior member

Attachments

Elite Member

Golden Member

Senior member

Member

Diamond Member

Senior member

Member

Platinum Member

Platinum Member

Senior member

Golden Member

Diamond Member

Diamond Member

Senior member

Diamond Member

Golden Member

Diamond Member

Golden Member

Senior member

Diamond Member

Diamond Member

Diamond Member

Member

Member

Member