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	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 (20A)	Arrow Lake (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	4P + 4E	4P + 8E
LLC	24 MB	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.)

 

[igor_kavinski]

While I can only guess, based on their own press materials, they wanted to save transistor budget. After all you need to duplicate some structures to support SMT

So now they have two Lion Cove designs, one with SMT for servers and one without?

 

[MarkPost]

This is my 13900K with four E-Cores disabled to simulate a 14700K:



So, with V-Ray 265K would be barely faster (1-2%) than 14700K. More than a poor improvement, to say the least

 

[MS_AT]

So now they have two Lion Cove designs, one with SMT for servers and one without?

Rumors are pointing in that direction. Server people this way won't complain as they will get what they are used to. Marketing has already prepared an easy explanation with maximizing throughput not energy saving to justify the difference and Client has simpler scheduler model.

 

[jdubs03]

Mm that does make sense they are the same chip. Maybe more KF test were done under OC conditions?

Choosing the median makes the 14700K get a little closer to the KF, but that doesn’t solve the potential scenario you mentioned there.

This is my 13900K with four E-Cores disabled to simulate a 14700K:

View attachment 107207

So, with V-Ray 265K would be barely faster (1-2%) than 14700K. More than a poor improvement, to say the least

And we can assume the 14900K with 4 e-cores disabled would score at least a few percent higher.

 

[MarkPost]

Choosing the median makes the 14700K get a little closer to the KF, but that doesn’t solve the potential scenario you mentioned there.
View attachment 107208

And we can assume the 14900K with 4 e-cores disabled would score at least a few percent higher.

why? 13900K and 14700K are just the same thing except 13900K has 4 more E-cores.

 

[igor_kavinski]

why? 13900K and 14700K are just the same thing except 13900K has 4 more E-cores.

Different stepping maybe.

 

[jdubs03]

why? 13900K and 14700K are just the same thing except 13900K has 4 more E-cores.

I was thinking because they clock a bit higher right. The p-core turbos higher by 200MHz and the e-core turbos higher by 100MHz. Maybe they sustain at the same frequency and turbo here and there.

But looking at:

Maybe it’s barely a smidge higher.

 

[cannedlake240]

one with SMT for servers and one without?

Nope rumor is that smt is gone from IA P cores completely. Also Xeon 7 doesn't use Lion. Timeline matches panther cove

 

[Hulk]

I DEMAND to know why Intel couldn't cook up Dynamic SMT: https://www.ibm.com/docs/el/zvm/7.2?topic=architectures-simultaneous-multithreading-smt

Skymont

 

[Det0x]

Yeesh well this isn’t a great result for the 265KF.

https://twitter.com/x/status/1833452761436041545

33153 samples for the 265K.
34500 samples for the 14700K.

Not looking too hot indeed

From the old 9950X ES numbers from ~2 months ago:

Think 60k should not be too hard with more mature agesa and optimized settings

 

[jdubs03]

Not looking too hot indeed

From the old 9950X ES numbers from ~2 months ago:
View attachment 107212
Think 60k should not be too hard with more mature agesa and optimized settings

Yeah that was humming along. The average/median for the 9950X is 48525/49508.

 

[cebri1]

Not looking too hot indeed

From the old 9950X ES numbers from ~2 months ago:
View attachment 107212
Think 60k should not be too hard with more mature agesa and optimized settings

vray scales very well with threads.

 

[MarkPost]

Not looking too hot indeed

From the old 9950X ES numbers from ~2 months ago:
View attachment 107212
Think 60k should not be too hard with more mature agesa and optimized settings

yeah Zen 5 is really strong in rendering tasks such as vray, corona, blender and so on. Except Cinebench, where for some reason doesnt perform so amazingly well as with the aforementioned.

Lets see Arrow Lake, but for the leaks seems to be on the contrary side...

 

[MarkPost]

I was thinking because they clock a bit higher right. The p-core turbos higher by 200MHz and the e-core turbos higher by 100MHz. Maybe they sustain at the same frequency and turbo here and there.

But looking at:
View attachment 107209
Maybe it’s barely a smidge higher.

yeah thats basically margin of error

 

[Josh128]

yeah Zen 5 is really strong in rendering tasks such as vray, corona, blender and so on. Except Cinebench, where for some reason doesnt perform so amazingly well as with the aforementioned.

Lets see Arrow Lake, but for the leaks seems to be on the contrary side...

Arrow Lake will apparently have more of the Cinebench Intel secret sauce, not so much Blender and VRAY. I think it will come in ~>45K in R23.

 

[511]

i think in blender it will close the gap Vray/corona nope

 

[Anhiel]

The V-Ray leak just create more questions than answers.

The results for Zen 5s aren't consistence gen over gen, while Arrow Lake supposedly having 2x more resources is nowhere near.
Arrow Lake should be a beast here. It could mean some future microcode could fix this.

But it does look like things are set for Arrow Lake regardless. And this points to Arrow Lake winning in CB R23 for both ST and MT.
For Core Ultra 9 285K that would be ST ~2450 @5.7GHz and MT ~47.5k (can't be more but could be 2% less due to error of margin from the leak)
Zen 5 will lose mainly due to power limitation resulting in significant all-core clock speed regression. It doesn't look like 3D -cache could change that. But since it seems AMD realized that and promising new features things might still be close.

 

[Elfear]

The V-Ray leak just create more questions than answers.

The results for Zen 5s aren't consistence gen over gen, while Arrow Lake supposedly having 2x more resources is nowhere near.
Arrow Lake should be a beast here. It could mean some future microcode could fix this.

But it does look like things are set for Arrow Lake regardless. And this points to Arrow Lake winning in CB R23 for both ST and MT.
For Core Ultra 9 285K that would be ST ~2450 @5.7GHz and MT ~47.5k (can't be more but could be 2% less due to error of margin from the leak)
Zen 5 will lose mainly due to power limitation resulting in significant all-core clock speed regression. It doesn't look like 3D -cache could change that. But since it seems AMD realized that and promising new features things might still be close.

I'm curious how gaming pans out between the two architectures. Looks like things will be close but ARL may take the lead with Zen 5's poor INT uplift. Either way, it's good to have tight competition!

 

[inf64]

I'm curious how gaming pans out between the two architectures. Looks like things will be close but ARL may take the lead with Zen 5's poor INT uplift. Either way, it's good to have tight competition!

INT uplift is not poor at around 11% for GR, it's just that memory subsystem is limiting Zen 5. ARL will first have to beat 7800X3D and 7950X3D which will be a tall order with its tiled architecture and clock regression vs RPL. Plus, AMD will have X3D on Zen 5 mere months after ARL launches which should be a good 10-15% faster than 7800X3D. So beating AMD in gaming is highly unlikely, but let's see.

 

[igor_kavinski]

Looks like things will be close but ARL may take the lead with Zen 5's poor INT uplift.

Memory latency may negate whatever IPC uplift ARL has. It may need DDR5-8400 or higher to realize decent gains over 14900KS.

CUDIMMs may save ARL-S, if DDR5-9200 becomes available before year end and actually possible on 285K.

 

[Hulk]

I predict that ARL will be be between 42,000 and 44,000 in CB R23 MT when running 5.4/4.6.

 

[Elfear]

INT uplift is not poor at around 11% for GR, it's just that memory subsystem is limiting Zen 5. ARL will first have to beat 7800X3D and 7950X3D which will be a tall order with its tiled architecture and clock regression vs RPL. Plus, AMD will have X3D on Zen 5 mere months after ARL launches which should be a good 10-15% faster than 7800X3D. So beating AMD in gaming is highly unlikely, but let's see.

Good point. I should say "lower than expected INT increase". I was fully onboard the hype train (right up front tugging on the whistle) so my expectations were rather unrealistic.

 

[Philste]

AMD will have X3D on Zen 5 mere months after ARL launches which should be a good 10-15% faster than 7800X3D.

In which world will ZEN5 X3D be 10-15% faster than ZEN4 X3D? ZEN5 is about 5% faster than ZEN4 in Gaming with same RAM. All Tests that show more is 5200MT vs 5600MT. So 5% over 7800X3D is what we should expect.

 

[inf64]

In which world will ZEN5 X3D be 10-15% faster than ZEN4 X3D? ZEN5 is about 5% faster than ZEN4 in Gaming with same RAM. All Tests that show more is 5200MT vs 5600MT. So 5% over 7800X3D is what we should expect.

In a world where 9800X3D is not limited to 5Ghz boost clock. 5.4Ghz should be easy to achieve and there you go, 13% faster than 7800X3D.

 

[Hulk]

In a world where 9800X3D is not limited to 5Ghz boost clock. 5.4Ghz should be easy to achieve and there you go, 13% faster than 7800X3D.

Besides performance I'm interested to see if ARL will be able to sustain 5.4 all-core under a moderately heavy load with air cooling and reasonable power. Peak clocks for Raptor have always been sort of a gaslight in that you need serious cooling to sustain them as well as good silicon. "Out-of-the-box" just working at sane power and thermals would be nice. I'm there with my 14900K but it took knowledge to set it up correctly. I'd rather not have to learn how to make my CPU run at stock, you know, like the old days.