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.)

GTracing · Oct 21, 2024

DavidC1 said:
C&C doesn't have everything correct, and they also miss things(such as existence of the OD-ILD. Heck they don't even mention it!). Since they had macro op execution in Silvermont, it would have been quite a change to go back to micro op execution in successors. Agner Fog's x86 optimization manual implies even Tremont keeps the same scheme.

Chips and Cheese got that piece of info from an Intel marketing slide. I would be quite surprised if it was false.

Hulk · Oct 21, 2024

AcrosTinus said:
Currently Raptor Lake even at an IPC disadvantage can get more done on the P-Cores due to HT. Hence we need more P-Cores not less.

The Best setup was 8P+8E the next best one would be 10P+8E or 10P+16E.

If what you mean by "get more done" is quickly degrading the P cores, then yes, you are correct.
Just kidding of course.
But we'll have to agree to disagree. I side with the Intel engineers on this one.

AcrosTinus · Oct 21, 2024

Hulk said:
If what you mean by "get more done" is quickly degrading the P cores, then yes, you are correct.
Just kidding of course.
But we'll have to agree to disagree. I side with the Intel engineers on this one.

I like the snark, I trust Intel too but a reduction in P-Cores will never happen.
The P-Cores have lost throughput due to the lack of HT and need a higher IPC increase (size of HT benefit) to compensate for that. I hope they go the higher P-Core count road rather than the higher E-Core route, we will see...

DavidC1 · Oct 21, 2024

GTracing said:
Chips and Cheese got that piece of info from an Intel marketing slide. I would be quite surprised if it was false.

I would like further clarification if that's the case, because Intel gets many things in their own slide wrong.

Meteorlake being up to 192EU, when Exist50 said that it never even existed, and it was simply a mistake.

What about the time the server head of Intel paraded a Sapphire Rapids wafer as a Sierra Forest one? Remember that one? We were wrapping our heads on how many cores that "SRF" die had?

AcrosTinus said:
I like the snark, I trust Intel too but a reduction in P-Cores will never happen.
The P-Cores have lost throughput due to the lack of HT and need a higher IPC increase (size of HT benefit) to compensate for that. I hope they go the higher P-Core count road rather than the higher E-Core route, we will see...

Lack of HT is better countered by having another E core cluster, since each E cores are much more powerful than HT anyways, and from a program point of view, it'll take less threads to accomplish the same task, thus more benefit.

Or, the better solution of a peak performing core derived off Skymont or Arctic Wolf.

Hulk · Oct 21, 2024

AcrosTinus said:
I like the snark, I trust Intel too but a reduction in P-Cores will never happen.
The P-Cores have lost throughput due to the lack of HT and need a higher IPC increase (size of HT benefit) to compensate for that. I hope they go the higher P-Core count road rather than the higher E-Core route, we will see...

Who knows? On one hand I understand the hybrid approach. On the other AMD seems to be doing more than okay with the more traditional approach.

OneEng2 · Oct 21, 2024

SiliconFly said:
C'mon, if Intel wins in a bench, the bench sucks cos it is doctored or optimized for Intel. But if Intel loses, Intel sucks! No two ways about it.

Fair comment. I would say that Cinebench comparisons that I have seen don't paint Intel processors in a very good light compared to Zen5:

AMD Ryzen 9 9950X Review - The New Flagship

With 16 cores and 32 threads, the Ryzen 9 9950X, powered by AMD’s Zen 5 architecture, is the fastest desktop processor we've ever tested. In our review, it breezed through application workloads and delivered high FPS rates in gaming. But at $650, it doesn’t come cheap.

www.techpowerup.com

If it is true that Arrow Lake has equal MT to 14900K, then it would seem to me that Arrow Lake is going to get beaten in MT by Zen 5. I must be missing something in this discussion.

MS_AT said:
MSVC is objectively the worst compiler of the big three C++ compilers (GCC, MSVC, CLANG) when it comes to generating performant code... The reason people use it when Clang is available directly from Visual Studio with two button clicks is still a puzzle to me

To be fair the full quote from C&C is this:

In other words in SPECint aarch64 has the advantage, in SPECfp, x64 has the advantage. Overall, ARM has the advantage. Still I would approach these results with caution as I doubt Chester looked at generated assembly to also check the quality of generated code. I would expect the 16 architectural GP registers are hurting x64 compared to aarch64 in SPECint where it will have to spill more often compared to ARM. In SpecFP, since this is usually more math oriented code, wider SIMD registers could explain x64 advantage. But once again, I am just guessing, I don't have access to SPEC sources so I cannot compile it myself to check

I completely agree. Clang (being more open source and community optimized) tends to generate very performant binaries. I would say that for some number of years, Intel has relied on a constant flood of new instructions (which AMD would then be a generation behind on) and specific benchmark optimizations to take advantage of these new instructions, in order to make their processors look better than they would be in general usage. This seems to have slowed

Hulk said:
Who knows? On one hand I understand the hybrid approach. On the other AMD seems to be doing more than okay with the more traditional approach.

I tend to believe that a heterogeneous approach will prevail in the long run as specific processing for specific workloads can be handled much more quickly with hardware designed to do that work.

In the short run, operating systems are not able to easily handle this new complexity so a more homogeneous chip architecture is likely to do better.

My example is how fast a digital mixer can perform digital signal processing. 48 channels of sound being modified, fed into busses that are modified, summed to a pair of outputs and sent to speakers all in near real-time (< 1mSec). All of this is done with DSP chips designed to handle sound.

AcrosTinus · Oct 21, 2024

Hulk said:
Who knows? On one hand I understand the hybrid approach. On the other AMD seems to be doing more than okay with the more traditional approach.

I still cannot understand how AMD got to this performance level, Zen is a mystery to me, held back on consumer by memory bandwidth, latency and their chipset (also chiplet to IO-Die bandwidth, on Server they now have 2 links per chiplet). If they solve these points, I might be gone...

Hulk · Oct 21, 2024

OneEng2 said:
My example is how fast a digital mixer can perform digital signal processing. 48 channels of sound being modified, fed into busses that are modified, summed to a pair of outputs and sent to speakers all in near real-time (< 1mSec). All of this is done with DSP chips designed to handle sound.

I do a lot of multitrack audio work. Playing a gig tomorrow actually that I will be recording and I run sound as well. Soundcraft UI24R stagebox mixer, controlled via wireless, 24 tracks, fx, etc... running the sytem in real time with no perceptible latency to the musicians, 8 aux mixes, feedback control...AND the ability to multitrack record all 24 tracks. It's nuts. 30 years ago it would have taken a good sized box truck full of racks of gear to do that.

AcrosTinus said:
I still cannot understand how AMD got to this performance level, Zen is a mystery to me, held back on consumer by memory bandwidth, latency and their chipset (also chiplet to IO-Die bandwidth, on Server they now have 2 links per chiplet). If they solve these points, I might be gone...

I am going to take a good, hard look at the 9950X vs Arrow Lake if I upgrade this generation. I know, I know, there isn't much additional performance to be had from my 14900K. But the sad reality is there are more upgrades behind me than ahead of me if I upgrade at my old rate. This is a fun hobby for me and I splurge more on it these days.

What is the stock all-core clock for the 9950X BTW?

jdubs03 · Oct 21, 2024

Hulk said:
I do a lot of multitrack audio work. Playing a gig tomorrow actually that I will be recording and I run sound as well. Soundcraft UI24R stagebox mixer, controlled via wireless, 24 tracks, fx, etc... running the sytem in real time with no perceptible latency to the musicians, 8 aux mixes, feedback control...AND the ability to multitrack record all 24 tracks. It's nuts. 30 years ago it would have taken a good sized box truck full of racks of gear to do that.

I am going to take a good, hard look at the 9950X vs Arrow Lake if I upgrade this generation. I know, I know, there isn't much additional performance to be had from my 14900K. But the sad reality is there are more upgrades behind me than ahead of me if I upgrade at my old rate. This is a fun hobby for me and I splurge more on it these days.

What is the stock all-core clock for the 9950X BTW?

The question then becomes what features are you looking for in an upgrade from a 1 year old top-end CPU (that doesn’t exhibit any instability)?

Maybe the answer to that will be clearer in 3 days 😄

You’d have to deal with the switching costs too, and probably a depressed resale market for that 14900K.

Oh and that’s a lot of sound equipment. Is that typical of a band that does local/regional gigs?

Markfw · Oct 21, 2024

Hulk said:
I do a lot of multitrack audio work. Playing a gig tomorrow actually that I will be recording and I run sound as well. Soundcraft UI24R stagebox mixer, controlled via wireless, 24 tracks, fx, etc... running the sytem in real time with no perceptible latency to the musicians, 8 aux mixes, feedback control...AND the ability to multitrack record all 24 tracks. It's nuts. 30 years ago it would have taken a good sized box truck full of racks of gear to do that.

I am going to take a good, hard look at the 9950X vs Arrow Lake if I upgrade this generation. I know, I know, there isn't much additional performance to be had from my 14900K. But the sad reality is there are more upgrades behind me than ahead of me if I upgrade at my old rate. This is a fun hobby for me and I splurge more on it these days.

What is the stock all-core clock for the 9950X BTW?

I am not sure what the spec is is anybody's else's result. I am getting 4.6 @ 100% load

Kepler_L2 · Oct 21, 2024

Josh128 said:
Thats worse than Zen 1 mem latency.

Infinity Fabric haters in shambles.

511 · Oct 22, 2024

Amazing Slides

coercitiv · Oct 22, 2024

511 said:
Amazing Slides

I'm having trouble reconciling that slide with... reality. How did they manage to score 1370 on the XPS 9350 /w U7 256V?

Here's what a quick search turned out from a laptop review outlet:

511 · Oct 22, 2024

coercitiv said:
I'm having trouble reconciling that slide with... reality. How did they manage to score 1370 on the XPS 9350 /w U7 256V?

Here's what a quick search turned out from a laptop review outlet:
View attachment 109954

I know how but it is a guess they could have disabled Turbo Boost

desrever · Oct 22, 2024

Maybe they put it in the oven during the benchmark.

coercitiv · Oct 22, 2024

Looking at other slides from the presentation, I think it has to do with performance while unplugged. The XPS laptop may be too aggressive with throttling while on battery.

I don't have time to go through all the slides and notes, but at least this would make sense in terms of measurements. (how that Dell is configured from the factory is another issue altogether)

511 · Oct 22, 2024

ARL CPU Die

https://twitter.com/x/status/1848573032799080586

igor_kavinski · Oct 22, 2024

Gee, Arrow Lake, look at the size of Ps you got!

Seriously, that's gotta be the disabled AVX-512 and HT circuitry in all of that silicon real estate.

Kocicak · Oct 22, 2024

Such a beautiful CPU cannot be bad!

coercitiv · Oct 22, 2024

Kocicak said:
Such a beautiful CPU cannot be bad!

Beauty can be deceiving.

Kocicak · Oct 22, 2024

igor_kavinski said:
Gee, Arrow Lake, look at the size of Ps you got!

They seem to be about three times larger than E-cores. There are massive yet to be revealed performance reserves in all that surface. They are like sleeping volcanos.

DrMrLordX · Oct 22, 2024

igor_kavinski said:
Seriously, that's gotta be the disabled AVX-512 and HT circuitry in all of that silicon real estate.

Does Lion Cove even have such "disabled AVX-512 and HT circuitry"?

Saylick · Oct 22, 2024

DrMrLordX said:
Does Lion Cove even have such "disabled AVX-512 and HT circuitry"?

I would imagine it would be easy to pick out if Lion Cove had full AVX512 since it is usually a large structure that’s typically comprised of two identical-looking AVX256 pipes…

SiliconFly · Oct 22, 2024

DrMrLordX said:
Does Lion Cove even have such "disabled AVX-512 and HT circuitry"?

In the Intel P core team video, they said it's relatively easy to turn features on or off. And since Lion Cove has no server parts, it's likely they might have not included it.

511 · Oct 22, 2024

It may have been present but is disabled considering the die size for ARL-S

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

Senior member

Attachments

Member

Diamond Member

Member

Golden Member

Diamond Member

Senior member

Member

Diamond Member

Golden Member

Moderator Emeritus, Elite Member

Senior member

Golden Member

Diamond Member

Golden Member

Senior member

Diamond Member

Golden Member

Lifer

Golden Member

Diamond Member

Golden Member

Lifer

Diamond Member

Golden Member

Golden Member