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.

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.)

OneEng2 · May 13, 2025

Thunder 57 said:
It seems some people have a hard time grasping this. It's why I don't get why the Core Ultra is crapped on so badly. It pretty damn good in productivity. It's average at best in gaming though so it gets a bad rep.

In all fairness, it gets beaten handily in the majority of tasks. I think you have to cherry pick to get places where ARL is a good idea.

511 said:
the only issue with core ultra is regression in workloads outside gaming that's it otherwise it's fine.

Like I stated, it isn't fine for many tasks (the majority of them). Stuck between having a latency issue in the design, lacking AVX512 and SMT, it just struggles. Only in bandwidth heavy tasks like CB24 does it excel.

I do agree that DIY and Gaming is only a small minority of the market; however, latency bound applications and AVX512 supported applications abound in most markets.

Thunder 57 · May 13, 2025

OneEng2 said:
In all fairness, it gets beaten handily in the majority of tasks. I think you have to cherry pick to get places where ARL is a good idea.

Like I stated, it isn't fine for many tasks (the majority of them). Stuck between having a latency issue in the design, lacking AVX512 and SMT, it just struggles. Only in bandwidth heavy tasks like CB24 does it excel.

I do agree that DIY and Gaming is only a small minority of the market; however, latency bound applications and AVX512 supported applications abound in most markets.

To be clear I would never recommend it. It is unbalanced and on a dead end socket. Future derivatives might prove fruitful though.

511 · May 13, 2025

Thunder 57 said:
To be clear I would never recommend it. It is unbalanced and on a dead socket. Future derivatives maybe prove fruitful though.

Except for the 265K on a decent MB Either

Thunder 57 · May 13, 2025

511 said:
Except for the 265K on a decent MB Either

Still wouldn't bother. The price cut helps but if they do fix its issues with Nova Lake you wouldn't be able to drop in a CPU upgrade.

Hitman928 · May 13, 2025

eek2121 said:
Yep, you can’t say a benchmark is invalid because one of the vendors chose not to implement an instruction set. Otherwise, Apple Geekbench 6 results wouldn’t be valid, and AVX-512 is used in far more things.

I have seen games use 16-32 cores. Shoot, even 2D games like Factorio.

Benchmarks I’ve seen say otherwise,

eek2121 said:
Does every game do that? No. However, if you build it, they will come.

More importantly, PCs aren’t just about gaming. Productivity tasks are a thing.

I haven’t seen any games use that many cores outside of maybe during shader compilation or texture decompression operations.

DrMrLordX · May 13, 2025

Hitman928 said:
Benchmarks I’ve seen say otherwise,

Okay now do that for the 9950X3D vs 9800X3D.

Hitman928 · May 13, 2025

DrMrLordX said:
Okay now do that for the 9950X3D vs 9800X3D.

That won't change the conclusion about using 16+ cores but sure:

Still no benefit to the additional cores, they just get put on the right 8-core CCD now. . .

DrMrLordX · May 13, 2025

Hitman928 said:
Still no benefit to the additional cores, they just get put on the right 8-core CCD now. . .

On the contrary, it does. By quite a bit. The 9950X3D and 9800X3D are neck-and-neck, while the 7800X3D is significantly faster than the 7950X3D (72% faster). That tells me that the 7950X3D test was done with older firmware/chipset drivers and that game threads were suffering from running on the wrong ccd and/or being swapped between ccds where as the 9950X3D did not have that problem.

MS_AT · May 13, 2025

DrMrLordX said:
On the contrary, it does. By quite a bit. The 9950X3D and 9800X3D are neck-and-neck, while the 7800X3D is significantly faster than the 7950X3D (72% faster). That tells me that the 7950X3D test was done with older firmware/chipset drivers and that game threads were suffering from running on the wrong ccd and/or being swapped between ccds where as the 9950X3D did not have that problem.

You undestand that by your own admission it is not additional cores that help, but placing them on the right CCD? So the additional cores do not convey additional benefit?

LightningZ71 · May 13, 2025

Factorio still has a single thread that is the primary performance determinant. It leans heavily on memory access latency, which is why the X3D chips show such an improvement. Even when the 3d cache is present, the additional cores aren't typically holding the game back like the first thread is.

My point is that precious few games have more than one to maybe three threads that are the biggest factor in game performance. They may spawn a bunch of other threads for housekeeping, data movement and unpacking, and even extra "AI" work, but none of that will hold the game back unless the first couple of threads are obscenely fast and we'll fed. This is why single and two core boosting behavior is often such a big deal and all-core boost barely affects anything GAME related.

So why does Intel continue to make CCDs with 8 performance cores? A couple of reasons centered around benchmarking and a few non-game programs that scale out past 4 threads well.

alcoholbob · May 13, 2025

LightningZ71 said:
Factorio still has a single thread that is the primary performance determinant. It leans heavily on memory access latency, which is why the X3D chips show such an improvement. Even when the 3d cache is present, the additional cores aren't typically holding the game back like the first thread is.

My point is that precious few games have more than one to maybe three threads that are the biggest factor in game performance. They may spawn a bunch of other threads for housekeeping, data movement and unpacking, and even extra "AI" work, but none of that will hold the game back unless the first couple of threads are obscenely fast and we'll fed. This is why single and two core boosting behavior is often such a big deal and all-core boost barely affects anything GAME related.

So why does Intel continue to make CCDs with 8 performance cores? A couple of reasons centered around benchmarking and a few non-game programs that scale out past 4 threads well.

I assume its out of habit maybe now that the P and E cores have been disaggregated it does seem weird, but on the other hand, the P-core clocks are still linked, and they have foreground apps and games set up to populate on the P-cores in a certain way, so maybe they just don't want to mess with their clock boosting algorithm too much and it's easier to just keep things the way they are until they go to a unified core.

511 · May 13, 2025

On the point of Intel there is the bigLLC SKU

poke01 · May 13, 2025

Hitman928 said:
Benchmarks I’ve seen say otherwise,

View attachment 123717

I haven’t seen any games use that many cores outside of maybe during shader compilation or texture decompression operations.

and even accounting for non-3D modles. The reason why the i9 is ahead is not cause number of cores but clock speed advantage it has.

In the end 1t always reigns supreme for most client workloads.

DrMrLordX · May 14, 2025

MS_AT said:
You undestand that by your own admission it is not additional cores that help, but placing them on the right CCD? So the additional cores do not convey additional benefit?

The point you are missing, is that the additional cores appear to hurt gaming performance in factorio (according to the Zen4 graphs) yet the reality is, they do not.

MS_AT · May 14, 2025

DrMrLordX said:
The point you are missing, is that the additional cores appear to hurt gaming performance in factorio (according to the Zen4 graphs) yet the reality is, they do not.

I think the point I must have missed is anyone claiming that additional cores hurt gaming performance. I thought the point was they don't help, not that they actively hurt

Hitman928 · May 14, 2025

MS_AT said:
I think the point I must have missed is anyone claiming that additional cores hurt gaming performance. I thought the point was they don't help, not that they actively hurt

Correct, no one was claiming the extra cores hurt. If Factorio really could use 16+ cores, then getting put on the "wrong CCD" wouldn't matter because it would be using all of the available cores and the higher core count CPUs should have an advantage. I guess the Zen 5 results make it more obvious with the X3D chips not having the "wrong CCD" issue, but point was simply to show that Factorio cannot actually make use of that many cores. It really seems to be limited to effectively using only a handful of cores at most, probably less but I don't have results for less than 6-core CPUs. There's been a few other claims that other modern games can effectively use high core count CPUs as well, but there has only been claims, no real evidence has been posted supporting them or specific games named. We finally got a specified game as an example, Factorio, which made it possible to check whether it was true or not. Turns out, it's not true for that title.

511 · May 14, 2025

Well unless someone is running Crysis on CPU I don't see a point beyond 8 Fast Cores with tons of cache sure it would increase in future but it should be fine for now.

Also Intel's plan to sell Big Cache Version now people would definitely buy the 52 Core version

https://twitter.com/x/status/1922618865948782720

Joe NYC · May 14, 2025

511 said:
Well unless someone is running Crysis on CPU I don't see a point beyond 8 Fast Cores with tons of cache sure it would increase in future but it should be fine for now.

Also Intel's plan to sell Big Cache Version now people would definitely buy the 52 Core version

https://twitter.com/x/status/1922618865948782720
View attachment 123750

How much cache is the Big Cache version? If it is split between 2 chiplets, than it is effectively half of the total amount (for gaming).

511 · May 14, 2025

Joe NYC said:
How much cache is the Big Cache version? If it is split between 2 chiplets, than it is effectively half of the total amount (for gaming).

So the big Cache chiplet is 144MB the non cache one is 36 MB iirc.

msj10 · May 14, 2025

Joe NYC said:
How much cache is the Big Cache version? If it is split between 2 chiplets, than it is effectively half of the total amount (for gaming).

it's not split. one die will have access to the big cache and the other die doesn't just like the 7950X3D.

511 · May 14, 2025

Now some how this has become viable to sell.
If they do not screw it up and increase gaming perf by 15-20% vs 9800X3D than it would be a win.
🤣

gdansk · May 14, 2025

511 said:
increase gaming perf by 15-20% vs 9800X3D

That'd be nearly a 1.4x increase in gaming performance relative to the 285K. What is it with speculators and 40%

511 · May 14, 2025

gdansk said:
That'd be nearly a 1.4x increase in gaming performance relative to the 285K. What is it with speculators and 40%

what i didn't actually meant to say 40% 🤣 reminds me of @Kepler_L2.
Stock ARL is behind by 15% according to TPU

gdansk · May 14, 2025

511 said:
what i didn't actually meant to say 40% 🤣 reminds me of @Kepler_L2.
Stock ARL is behind by 15% according to TPU

From their latest review with an RTX 5090 and new ARL firmware:

221.7/172.5 = 9800X3D 128.5% ahead of the 285K in gaming performance in this test.

To be 1.15 to 1.2x above that would be 1.48x to 1.54x. Eclipsing even the great Zen 5 hype train, though unintentionally.

511 · May 14, 2025

Did ARL get the fixes in time? Edited my og post to clarify few things. My English is broken sometimes 🤣.

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

Senior member

Attachments

Senior member

Diamond Member

Platinum Member

Diamond Member

Diamond Member

Lifer

Diamond Member

Lifer

Senior member

Platinum Member

Diamond Member

Platinum Member

Diamond Member

Lifer

Senior member

Diamond Member

Platinum Member

Diamond Member

Platinum Member

Member

Platinum Member

Diamond Member

Platinum Member

Diamond Member

Platinum Member