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.)

511 · Nov 8, 2024

46% PPC difference between Intel/AMD both have around 27% frequency advantage at 5.7Ghz though

poke01 · Nov 8, 2024

511 said:
It will be around 20% if you factor the OS

i can assure you that Cinebench doesn’t increase the score based on OS. I can test Cinebench 2024 on my Intel MacBook tomorrow with macOS and Windows tomorrow.

poke01 · Nov 8, 2024

511 said:
View attachment 11125946% PPC difference between Intel/AMD both have around 27% frequency advantage at 5.7Ghz though

I wouldn’t put much stock into Geekerwan SPEC tests. Wait for David Huang.

Hitman928 · Nov 8, 2024

511 said:
there may be some edge cases otherwise we would not have options to disable SMT on SKUs

SMT enabled instances are used a lot by cloud providers, most of the instances for AMD that I've seen have it enabled. SMT disabled instances are offered as well for those that want the highest performance per CPU (typically for licensing costs) or who have security concerns.

igor_kavinski · Nov 8, 2024

cannedlake240 said:
Their next chance from rumors is the Atom based uarch that's targeting beyond 2028 timeline

If this goes on, they will be playing catch-up forever...

511 · Nov 8, 2024

Hitman928 said:
SMT enabled instances are used a lot by cloud providers, most of the instances for AMD that I've seen have it enabled. SMT disabled instances are offered as well for those that want the highest performance per CPU (typically for licensing costs) or who have security concerns.

Nice to know

511 · Nov 8, 2024

igor_kavinski said:
If this goes on, they will be playing catch-up forever...

Unless it has magical IPC increase yes

511 · Nov 8, 2024

Duplicate

MoistOintment · Nov 8, 2024

poke01 said:
so what is Intels answer to this?
Nova lake?

Even if Intel didn't improve IPC from ARL -> NVL (which would be a disappointment if they didn't, and isn't the case) the fact that NVL will be adopting LNL's design, putting Compute, iGPU, and Memory Controller all on the same tile, will be a huge improvement in desktop. Not to mention to multiple rumors about increased L3. Just those two things alone would be a good improvement - just a shame it's a 2 year gap instead of next year. Add in IPC improvements, and I think NVL will be a good gen. Compared to Zen 6? Idk - way too early to tell.

That being said, I'm not gonna hop on any hype bandwagons even though they'll inevitably form again. ARL and Zen 5 should be a lesson to all that increasing IPC doesn't necessarily mean increasing gaming performance.

Gideon · Nov 8, 2024

511 said:
It will be around 20% if you factor the OS but Nova Lake P core should be at least 20-25% ST Improvement considering they will fix their IMC/Mem controller along with APX and additional changes.Also they are doing a Tick and Tock so it is to be expected otherwise I don't know what would happen to x86

IMC / Mem controller will help in lots of thing, but i doubt it will make a dent in the SPEC scores (improvingL2 / L3 cache though might). APX same story. Useful but won't move the needle in SPEC.

poke01 said:
I wouldn’t put much stock into Geekerwan SPEC tests. Wait for David Huang.

Yeah, makes sense. Weren't they the ones to compile SPEC FP with AVX off?

cannedlake240 · Nov 8, 2024

igor_kavinski said:
If this goes on, they will be playing catch-up forever...

If they can get 30-40% ipc over skymont at >5Ghz by late 2028, then no... Two 15-20% jumps in 4-5 years isn't completely outlandish, NovaLake E core is probably halfway there already. Apple has been quite stagnant in iso clock perf, M4 is what 15% better than M1 at best? Intel doesn't have to beat apple either, just be competitive vs Oryon, Cortex and rumored Nvidia custom IP

511 · Nov 8, 2024

cannedlake240 said:
If they can get 30-40% ipc over skymont at >5Ghz by late 2028, then no... Two 15-20% jumps in 4-5 years isn't completely outlandish, NovaLake E core is probably halfway there already. Apple has been quite stagnant in iso clock perf, M4 is what 15% better than M1 at best? Intel doesn't have to beat apple either, just be competitive vs Oryon, Cortex and rumored Nvidia custom IP

M4 is what like 7-8% IPC Increase everything else is just Clocks and Clock scaling is pretty good in everyday applications

511 · Nov 8, 2024

Gideon said:
IMC / Mem controller will help in lots of thing, but i doubt it will make a dent in the SPEC scores (improvingL2 / L3 cache though might). APX same story. Useful but won't move the needle in SPEC.

code compiled with Intel APX contains 10% fewer loads and more than 20% fewer stores than the same code compiled for an Intel® 64 baseline.
It should increase ipc definitely
Source

Introducing Intel® Advanced Performance Extensions (Intel® APX)

These extensions expand the x86 instruction set with access to registers and features that improve performance.

www.intel.com

OneEng2 · Nov 8, 2024

MoistOintment said:
Even if Intel didn't improve IPC from ARL -> NVL (which would be a disappointment if they didn't, and isn't the case) the fact that NVL will be adopting LNL's design, putting Compute, iGPU, and Memory Controller all on the same tile, will be a huge improvement in desktop. Not to mention to multiple rumors about increased L3. Just those two things alone would be a good improvement - just a shame it's a 2 year gap instead of next year. Add in IPC improvements, and I think NVL will be a good gen. Compared to Zen 6? Idk - way too early to tell.

That being said, I'm not gonna hop on any hype bandwagons even though they'll inevitably form again. ARL and Zen 5 should be a lesson to all that increasing IPC doesn't necessarily mean increasing gaming performance.

Yea, but that tile will be as big as a pie plate. Not sure how cost effective this step back will be. Intel can't just keep throwing money at the problem. I think that is a bad habit they have from decades of huge positive cash flow.

When you are in a hole, quit digging!

OneEng2 · Nov 8, 2024

511 said:
code compiled with Intel APX contains 10% fewer loads and more than 20% fewer stores than the same code compiled for an Intel® 64 baseline.
It should increase ipc definitely
Source

Introducing Intel® Advanced Performance Extensions (Intel® APX)

These extensions expand the x86 instruction set with access to registers and features that improve performance.

www.intel.com

I think this is how everyone got so disappointed with Arrow Lake. Adding up all the little improvements into a big number ..... that doesn't actually materialize in any real world test.

I think they need to figure out how to get AVX512 back in their design before their DC market share drops further. Seems like their big problem with it is heat .... and maybe die size as well. AMD seems to have found a way to make it run fast .... and cool, so it seems it is not a technical impossibility. Seems like APX is a patch. Still, not a horrible idea. Every little bit helps.

Meteor Late · Nov 8, 2024

cannedlake240 said:
If they can get 30-40% ipc over skymont at >5Ghz by late 2028, then no... Two 15-20% jumps in 4-5 years isn't completely outlandish, NovaLake E core is probably halfway there already. Apple has been quite stagnant in iso clock perf, M4 is what 15% better than M1 at best? Intel doesn't have to beat apple either, just be competitive vs Oryon, Cortex and rumored Nvidia custom IP

Apple has been using part of their transistor budget into clocking the CPU higher, not just on increasing IPC, so it doesn't make much sense to only talk about IPC, what matters is performance, including clock speed.
Now that Apple is approaching x86 frequency, it will be interesting to see what happens, if Apple indeed has hit somewhat of a wall in terms of IPC AND frequency, or if they can still get solid gains, but as it still hasn't happened, one cannot say there is a wall already.
You are talking about 30-40% IPC improvement at >5GHz by late 2028, so 4 years from now. You can check the performance difference between M1 and M4, exactly 4 years timeframe, and you will see that it's higher than 40%.

511 · Nov 8, 2024

OneEng2 said:
I think this is how everyone got so disappointed with Arrow Lake. Adding up all the little improvements into a big number ..... that doesn't actually materialize in any real world test.

I think they need to figure out how to get AVX512 back in their design before their DC market share drops further. Seems like their big problem with it is heat .... and maybe die size as well. AMD seems to have found a way to make it run fast .... and cool, so it seems it is not a technical impossibility. Seems like APX is a patch. Still, not a horrible idea. Every little bit helps.

Nice way to take the conversation off topic lol.
AVX-512 is in P core design it's E core that lacks it lol it's AVX10 that's AVX-512 Patch and multiple improvements and it doesn't happen now you are talking about Ice lake now.
Does anyone in this form has proof of AVX-512 Hammering Clocks in Intel design from EMR?

511 · Nov 8, 2024

Meteor Late said:
Apple has been using part of their transistor budget into clocking the CPU higher, not just on increasing IPC, so it doesn't make much sense to only talk about IPC, what matters is performance, including clock speed.
Now that Apple is approaching x86 frequency, it will be interesting to see what happens, if Apple indeed has hit somewhat of a wall in terms of IPC AND frequency, or if they can still get solid gains.

For frequency they have been relying on Nodes more than their Arch

Meteor Late · Nov 8, 2024

511 said:
For frequency they have been relying on Nodes more than their Arch

That's only part of the equation, though, for example, M4 P cores use higher performance libraries, less dense, so that's transistor budget that could've been used to increase IPC instead if dense libraries would've been used.

naukkis · Nov 8, 2024

OneEng2 said:
I think this is how everyone got so disappointed with Arrow Lake. Adding up all the little improvements into a big number ..... that doesn't actually materialize in any real world test.

I think they need to figure out how to get AVX512 back in their design before their DC market share drops further. Seems like their big problem with it is heat .... and maybe die size as well. AMD seems to have found a way to make it run fast .... and cool, so it seems it is not a technical impossibility. Seems like APX is a patch. Still, not a horrible idea. Every little bit helps.

Intel server cpus have AVX512, only those E-core server versions lack it. APX instead is a new version of x86 trying to make x86 hardware faster. Of course APX won't help a bit old legacy software which won't support it but it's there to give x86 possibility to stay on performance race with ARM and RV. But as it's Intel extension there's also that possibility that their support for it ain't written in stone so it's still a bit uncertain feature.

DavidC1 · Nov 8, 2024

Robert Hallock is saying Arrowlake's performance issues are because it needs tuning and optimization and in some cases showing 180ns latency is related to it as well, although he says the poor performance isn't due to the latency nor the tiles. And that the optimizations are coming.

Lol, sounds a bit like Meteorlake again.

When did we switch to CPUs needing driver updates and Fine Wine?

DavidC1 · Nov 8, 2024

poke01 said:
I wouldn’t put much stock into Geekerwan SPEC tests. Wait for David Huang.

David Huang isn't doing tests himself though. Lot of them he gathers from elsewhere like X.

cannedlake240 said:
If they can get 30-40% ipc over skymont at >5Ghz by late 2028, then no... Two 15-20% jumps in 4-5 years isn't completely outlandish,

I think they can manage 30% with Arctic Wolf alone.

alcoholbob · Nov 8, 2024

DavidC1 said:
Robert Hallock is saying Arrowlake's performance issues are because it needs tuning and optimization and in some cases showing 180ns latency is related to it as well, although he says the poor performance isn't due to the latency nor the tiles. And that the optimizations are coming.

Lol, sounds a bit like Meteorlake again.

When did we switch to CPUs needing driver updates and Fine Wine?

He's saying the ring doesn't make much of a difference, but Der8auer found about 5% improvement OCing the ring from 3.7 to 4.2. I mean there's only a 6-7% gap between 14900K and 285K in gaming, and that one thing alone closes the majority of the gap.

He says focus on memory overclocking, which sure has a big effect on gaming, the problem is while it may be more effective than ring overclocking, RAM overclocking doesn't by itself do anything about performance gaps with the 14900K because why wouldn't you run the same fast memory on the 14900K?

MS_AT · Nov 8, 2024

OneEng2 said:
I think they need to figure out how to get AVX512 back in their design before their DC market share drops further. Seems like their big problem with it is heat .... and maybe die size as well. AMD seems to have found a way to make it run fast .... and cool, so it seems it is not a technical impossibility. Seems like APX is a patch. Still, not a horrible idea. Every little bit helps.

It will give x64 parity with ARM when it comes to number of architectural GPRs (32 to 32, right now x64 has 16 of them). I am looking forward to it as in some algos you can simply run out of GPRs for AVX512 addressing

DavidC1 · Nov 8, 2024

An hour long video? I'm not watching that. I liked the days when it was written and could read it all or skim the important parts.

alcoholbob said:
He's saying the ring doesn't make much of a difference, but Der8auer found about 5% improvement OCing the ring from 3.7 to 4.2. I mean there's only a 6-7% gap between 14900K and 285K in gaming, and that one thing alone closes the majority of the gap.

Geekerwan overclocked all the fabric on-die. Ring, the D2D and NOC or something and it wasn't enough. He concluded that there wasn't a single magical way of fixing it.

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

Senior member

Attachments

Golden Member

Platinum Member

Platinum Member

Diamond Member

Lifer

Golden Member

Golden Member

Golden Member

Member

Golden Member

Senior member

Golden Member

Golden Member

Senior member

Senior member

Member

Golden Member

Golden Member

Member

Senior member

Golden Member

Golden Member

Diamond Member

Senior member

Golden Member