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.)

 

[Josh128]

In 2024, AMD made 57% of its profit from server and only 14% in Client.

It is no wonder that AMD has stated many times that it's designs are "Server First" in nature.

With AMD having CPUs and GPUs that gamers want, that is no longer the case. In their latest ER, client and gaming earnings surpassed data center/enterprise. Client alone made a record 2.5B without gaming revenue. Intel starting to lose its grip in client means massive increases in client for AMD. 9800X3D may be AMDs most profitable client CPU ever.

Segment Summary

• Data Center segment revenue was $3.2 billion, up 14% year-over-year primarily driven by strong demand for AMD EPYC™ processors more than offsetting headwinds impacting AMD Instinct MI308 shipments to China.
• Client and Gaming segment revenue was $3.6 billion, up 69% year-over-year. Client revenue was a record $2.5 billion, up 67% year-over-year primarily driven by strong demand for the latest “Zen 5” AMD Ryzen™ desktop processors and a richer product mix. Gaming revenue was $1.1 billion, up 73% year-over-year driven by an increase in semi-custom revenue and strong AMD Radeon™ GPU demand.

 

[511]

All it would take a good CPU Out of Intel and it will reverse quickly in less than the time AMD took to gain Share.

 

[RTX]

For AMD to actually take advantage of N2X, they would likely have to change the physical design of the core itself, and I also wonder how much of AMD's N2P node would already use features of the N2X node.
Conspiracy theory- the reason why Techinsights is claiming that Zen 5 is on N4X, multiple times (so not a typo), while AMD themselves only claim N4P, is because the custom node AMD uses actually already has many back-end features of N4X.

Those are the same thing according to Synopsys/Cadence.

 

[Josh128]

All it would take a good CPU Out of Intel and it will reverse quickly in less than the time AMD took to gain Share.

Thats would only happen in a vaccum. That was the past, but that doesnt ring true anymore. If AMD continues to output strong designs, a strong design from Intel will face stronger headwinds to marketshare. Why? Because AM5 users know they have at least one, possibly two (according to MLIDs latest video) CPU gens left in them without needing to upgrade mobos, and Intel has taken a hit worldwide in its perception as "quality". There are many more AM5 client users out there than LGA1851, so Intel doesnt currently have as big of a satisfied user base.

Now, if Intel has another Core 2 moment and catches AMD with its pants down, sure, things can reverse quickly, but I'd wager they dont have another Core 2 in them at this time.

 

[511]

Thats would only happen in a vaccum. That was the past, but that doesnt ring true anymore. If AMD continues to output strong designs, a strong design from Intel will face stronger headwinds to marketshare. Why? Because AM5 users know they have at least one, possibly two (according to MLIDs latest video) CPU gens left in them without needing to upgrade mobos, and Intel has taken a hit worldwide in its perception as "quality". There are many more AM5 client users out there than LGA1851, so Intel doesnt currently have as big of a satisfied user base.

Now, if Intel has another Core 2 moment and catches AMD with its pants down, sure, things can reverse quickly, but I'd wager they dont have another Core 2 in them at this time.

You are forgetting OEMs and Channel and if NVL is truly 4 Gen of socket than AM5 doesn't have anything from platform perspective for new buyers.
Intel is going to have TB5 Integrated so easily that it would be OEMs fault for not integrating it.
Intel BT/Wi-Fi Combo plus other platform features that AMD can't match (Media/Display IP) as for Core 2 only Unified Core has the chance to do this.

 

[Win2012R2]

All it would take a good CPU Out of Intel and it will reverse quickly in less than the time AMD took to gain Share.

AM5 installment base is growing and that would ensure future CPU sales for AMD, I am looking forward to upgrading to Zen 6 next year, pass current CPU to someone else on the cheap, this will all keep Intel down.

Intel will crack in the next 18 months, they are almost out of runway.

 

[gdansk]

NVL is truly 4 Gen of socket

Intel has to announce that and commit to it publicly or no one will care. So they had better be confident in that platform.

 

[511]

Intel will crack in the next 18 months, they are almost out of runway.

LOL i like this take

 

[511]

Intel has to announce that and commit to it publicly or no one will care. So they had better be confident in that platform.

Yeah also i have said before Intel 4 gen socket at max will be 4 years

 

[OneEng2]

9800X3D may be AMDs most profitable client CPU ever.

Yes, but it still pales compared to DC profit.

Now, if Intel has another Core 2 moment

I can't see this happening. Intel is in its early release window of a new architecture (Arrow Lake). Panther Lake and Nova Lake will likely be all Intel releases until 2028 IMO. In all honesty, I am not even sure why Intel is doing that. I think that there will be only very minor changes between ARL, PTL, and NVL.

I wonder if Intel is wasting its design resources (which they have a lot less of these days) doing minor tweaks and making minor process changes vs planning the next big change on the next big process.

AMD's cadence has been about 18-24 months between processor releases. This has coincided with process changes that happen about that often as well.

TSMC will be releasing A14 in 2028. That is when I expect to see AMD release Zen 7. I can't fathom Intel having a process advantage over TSMC before then (or after then either for that matter). I still believe that processor generations only make sense when your process advances enough to increase your transistor budget and/or your PPA. Otherwise, you are just wasting engineering resources and validation on a "new" processor that has little to no improvement over the previous one.

 

[511]

I can't see this happening. Intel is in its early release window of a new architecture (Arrow Lake). Panther Lake and Nova Lake will likely be all Intel releases until 2028 IMO. In all honesty, I am not even sure why Intel is doing that. I think that there will be only very minor changes between ARL, PTL, and NVL.

we got Razer Lake for 2027 Intel does yearly releases and you are wrong on the minor changes part there are huge changes underneath from ARL -> PT-> NVL

I wonder if Intel is wasting its design resources (which they have a lot less of these days) doing minor tweaks and making minor process changes vs planning the next big change on the next big process.

tweaking a design takes a lot less than designing new arch

TSMC will be releasing A14 in 2028. That is when I expect to see AMD release Zen 7. I can't fathom Intel having a process advantage over TSMC before then (or after then either for that matter). I still believe that processor generations only make sense when your process advances enough to increase your transistor budget and/or your PPA. Otherwise, you are just wasting engineering resources and validation on a "new" processor that has little to no improvement over the previous one.

A14 Mass Production is set for H2 2028 same way N2 Mass production is H2 2025 so don't expect before Q2-Q3 next year of MP

 

[OneEng2]

we got Razer Lake for 2027 Intel does yearly releases and you are wrong on the minor changes part there are huge changes underneath from ARL -> PT-> NVL

Yes, Intel plans yearly releases. This is yet another example (IMHO) of how Intel wastes its money.

... and making "huge changes underneath" every year is incredibly risky .... if that is indeed what they are doing.

From what I can see so far, PTL isn't looking like much more than a new stepping of ARL with a bit more clock breathing room.

 

[511]

Yes, Intel plans yearly releases. This is yet another example (IMHO) of how Intel wastes its money.

It's what OEMs want so Intel does it

 

[jpiniero]

I think the Raptor Lake fiasco hurt more than Intel wants to admit.

 

[reaperrr3]

You are forgetting OEMs and Channel

Those are kinda losing trust in Intel, too.

and if NVL is truly 4 Gen of socket than AM5 doesn't have anything from platform perspective for new buyers.

The very point is that existing AM5 users can just upgrade the CPU, while NVL buyers would have to buy a new mobo as well.

Intel is going to have TB5 Integrated so easily that it would be OEMs fault for not integrating it.
Intel BT/Wi-Fi Combo plus other platform features that AMD can't match (Media/Display IP)

Maybe, but most desktop perf CPU users are gamers, and care mostly about absolute gaming perf or gaming perf/$, everything else is secondary.
WiFi just needs to be good enough, media/display usually comes from dGPU in a gaming system.

You can argue exceptions, but for the bulk of the market the above holds true.

 

[511]

Maybe, but most desktop perf CPU users are gamers, and care mostly about absolute gaming perf or gaming perf/$, everything else is secondary.
WiFi just needs to be good enough, media/display usually comes from dGPU in a gaming system.

You can argue exceptions, but for the bulk of the market the above holds true.

More PCI-E lanes as well

The very point is that existing AM5 users can just upgrade the CPU, while NVL buyers would have to buy a new mobo as well.

but what about new buyers and won't this be true for buyers going forward?

I only hate one aspect of both NVL and Zen 6

 

[511]

@igor_kavinski your 2nd favorite 12P Core CPU on LGA1700

https://twitter.com/x/status/1956762888951247091

 

[Josh128]

@igor_kavinski your 2nd favorite 12P Core CPU on LGA1700

https://twitter.com/x/status/1956762888951247091

Bart Simpson Lake. A waste of silicon, I cant believe they are wasting resources on another 10nm SKU.

 

[igor_kavinski]

Bart Simpson Lake. A waste of silicon, I cant believe they are wasting resources on another 10nm SKU.

They wasted so much sand on Raptor Bust. Why not some more?

And that's the only thing they seem to be capable of doing really. Intel 3, Intel 4, Intel 20A and now Intel 18A. All Pat nodes that they haven't been able to mass produce with any success. You don't see CPUs made on the Pat hyped processes flooding the market because they haven't really learned how to make these processes cheap enough to replace Intel 7.

 

[DrMrLordX]

Intel 3 is okay, but for some reason they skipped over it for desktop CPUs. Had there been a bridge product between Raptor Lake and Arrow Lake on Intel 3, it could have been moderately successful, especially if it didn't have the design baggage of Arrow Lake-S. It could be down to volume and/or the need to use Intel 3 for Granite Rapids and Sierra Forest.

 

[igor_kavinski]

Had there been a bridge product between Raptor Lake and Arrow Lake on Intel 3, it could have been moderately successful, especially if it didn't have the design baggage of Arrow Lake-S.

Arrow Lake tiles are supposed to be process agnostic (or so they claim). Why aren't they frickin' backporting ARL to Intel 7 and running it at 6+ GHz? Sure, it will run a lot hotter but plenty of folks don't mind the heat generated by 14900K. It would at least help them make some "easy" profit by utilizing their own fabs. Manufacture tons of non-K Arrow Lake Intel 7 CPUs to flood the market and get rid of the stigma of Raptor Bake. Unless they don't trust Intel 7 anymore...

 

[511]

Arrow Lake tiles are supposed to be process agnostic (or so they claim). Why aren't they frickin' backporting ARL to Intel 7 and running it at 6+ GHz? Sure, it will run a lot hotter but plenty of folks don't mind the heat generated by 14900K. It would at least help them make some "easy" profit by utilizing their own fabs. Manufacture tons of non-K Arrow Lake Intel 7 CPUs to flood the market and get rid of the stigma of Raptor Bake. Unless they don't trust Intel 7 anymore...

Uhh remember Rocket Lake SNC ported to 14nm ? also Intel 7 is capacity constrained they don't have enough capacity for Intel 7

 

[igor_kavinski]

also Intel 7 is capacity constrained they don't have enough capacity for Intel 7

Capacity constrained making what exactly? That's the most capacity they have of any of their processes.

 

[511]

Capacity constrained making what exactly? That's the most capacity they have of any of their processes.

SPR/EMR/RPL IO tiles for GNR/SRF alder lake N and bunch of other stuff

 

[DrMrLordX]

Arrow Lake tiles are supposed to be process agnostic (or so they claim). Why aren't they frickin' backporting ARL to Intel 7 and running it at 6+ GHz? Sure, it will run a lot hotter but plenty of folks don't mind the heat generated by 14900K. It would at least help them make some "easy" profit by utilizing their own fabs. Manufacture tons of non-K Arrow Lake Intel 7 CPUs to flood the market and get rid of the stigma of Raptor Bake. Unless they don't trust Intel 7 anymore...

They'd still have the crappy tile layout legacy from Meteor Lake. They'd have been better off doing a large monolithic die on Intel 3 with updated P and E cores. But by how much we'll never know.