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

Page 139 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
T

Tigerick

Senior member
Apr 1, 2022
782
750
106
PPT1.jpg
PPT2.jpg
PPT3.jpg



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.



LNL-MX.png

Intel Core Ultra 100 - Meteor Lake

INTEL-CORE-100-ULTRA-METEOR-LAKE-OFFCIAL-SLIDE-2.jpg

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.)



Clockspeed.png
 

Attachments

  • PantherLake.png
    PantherLake.png
    283.5 KB · Views: 24,025
  • LNL.png
    LNL.png
    881.8 KB · Views: 25,517
Last edited:
  • Like
Reactions: LaudaSur, jorden6655, john404 and 15 others
M

mikk

Diamond Member
May 15, 2012
4,292
2,382
136
  • Like
Reactions: Tlh97, Elfear and TESKATLIPOKA
dullard

dullard

Elite Member
May 21, 2001
25,932
4,522
126
tamz_msc said:
What's the point of the Ultra 9 185H?
Click to expand...
That table is woefully incomplete. For example, what type of IO, iGPU, or AI tiles do these have? That missing data would drastically impact the power used.

At the very least, the extra TDP would let the 185H be in turbo mode for far longer than the 165H. That would give faster and more consistent performance.
 
tamz_msc

tamz_msc

Diamond Member
Jan 5, 2017
3,865
3,730
136
But that doesn't explain why it supposedly has only 2.5 GHz base while the rest have 3.6-3.8 GHz base AND lower PL1.
 
SarahKerrigan

SarahKerrigan

Senior member
Oct 12, 2014
735
2,036
136
igor_kavinski said:
What is the possibility of Intel introducing a Core Ultra 9X Arrow Lake CPU with 256MB HBM3e on-package L4 cache? Would it be profitable for them at the $699 price point?
Click to expand...

puts on her NostaSeronx hat

Actually, the laptop gen after MTL (Wolf Lake) will use clustered GPU EU pairing with high-bandwidth low-latency "GfxRAM". Lunar Lake and Arrow Lake have both been canceled.

GfxRAM is like Hybrid Memory Cube, but it uses 25gbit serdes and QDR IV SRAM dice for both high bandwidth and low latency.

GfxRAM stack:

8GB
<90ns total latency, from issuing the load to getting data back
320GB/s

Wolf Lake can use up to four stacks. They can be allocated to the GPU alone or used as a CPU cache. The GfxRAM base die will be using GF 12nm FD-SOI.

After Wolf Lake comes Amber Lake. Amber Lake uses paired clustered CPU cores, sharing L1I, FPU, and integer multiply and divide units. It also adds next-generation GfxRAM.

Wolf Lake => MTL cores with GfxRAM and a new GPU uarch
Amber Lake => "Ruby Cove" paired cores, Intel 3
Crane Lake => "Opal Cove" paired cores with AVX-VL, Intel 20A, 64gbit serdes for GfxRAM
Tunnel Lake => ????

(TL;DR: No. Also, none of the above is true.)
 
  • Haha
  • Like
  • Wow
Reactions: Mopetar, Tlh97, Thibsie and 8 others
dullard

dullard

Elite Member
May 21, 2001
25,932
4,522
126
tamz_msc said:
But that doesn't explain why it supposedly has only 2.5 GHz base while the rest have 3.6-3.8 GHz base AND lower PL1.
Click to expand...
??? That exactly explains why. Suppose the 185H had a power hungry iGPU that uses up to 30 W. Then the CPU could only be guaranteed 15 W of power, and thus the base frequency must be lower than if it had a less power hungry iGPU. But if instead the 165H has in iGPU that uses up to 8 W of power, its CPU has a guaranteed 20 W and can have a higher base. That is exactly how power sharing works. When you put in a power hungry component with a fixed TDP, the base power the rest get is diminished.

Same thing if the 185H had power hungry AI tile, or onboard memory, or more IO, etc. The more power that those components can use, the less guaranteed power the CPU portion has and the lower the base clocks will be. If at any time those other components aren't used, the CPU portion gets the full 45 W and can turbo along happily.

Note: All power numbers were made up for this post as hypothetical as I don't actually know the power levels.
 
tamz_msc

tamz_msc

Diamond Member
Jan 5, 2017
3,865
3,730
136
dullard said:
??? That exactly explains why. Suppose the 185H had a power hungry iGPU that uses up to 30 W. Then the CPU could only be guaranteed 15 W of power, and thus the base frequency must be lower than if it had a less power hungry iGPU. But if instead the 165H has in iGPU that uses up to 8 W of power, its CPU has a guaranteed 20 W and can have a higher base. That is exactly how power sharing works. When you put in a power hungry component with a fixed TDP, the base power the rest get is diminished.

Same thing if the 185H had power hungry AI tile, or onboard memory, or more IO, etc. The more power that those components can use, the less guaranteed power the CPU portion has and the lower the base clocks will be. If at any time those other components aren't used, the CPU portion gets the full 45 W and can turbo along happily.

Note: All power numbers were made up for this post as hypothetical as I don't actually know the power levels.
Click to expand...
Theoretically, yes that could happen. But the last time there was a difference between the H and U parts in terms of iGPU was Tiger Lake.
 
dullard

dullard

Elite Member
May 21, 2001
25,932
4,522
126
tamz_msc said:
Theoretically, yes that could happen. But the last time there was a difference between the H and U parts in terms of iGPU was Tiger Lake.
Click to expand...
Intel has repeatedly stated for the last several years that their goal was to mix and match tiles. So why would anyone be surprised if one chip has different tiles? I have no information if Intel is starting this with the 185H, but it could explain the base frequency differences.

This plan is over 3 years old:
 
Last edited:
StinkyPinky

StinkyPinky

Diamond Member
Jul 6, 2002
6,956
1,268
126
tamz_msc said:
In terms of hardware - yes, don't know about Samsung's BIOS. They are pricey though. The best BIOS support for laptops with Intel CPUs is given by Dell.
Click to expand...

Samsung release the bios via Windows update, I've had a few of them come through.

I don't think they are as frequent as Dell, but certainly at a decent enough speed.
 
M

mikk

Diamond Member
May 15, 2012
4,292
2,382
136
dullard said:
Intel has repeatedly stated for the last several years that their goal was to mix and match tiles. So why would anyone be surprised if one chip has different tiles? I have no information if Intel is starting this with the 185H, but it could explain the base frequency differences.

This plan is over 3 years old:
Click to expand...


This not not the case because Intels uses two chips for Meteor Lake mobile, one chip for MTL-M/U and one chip for MTL-H. MTL-H gets a GPU tile with 128EUs and MTL-M/U a GPU tile with 64EUs. Ultra 7 155H gets the fully enabled 128EU version: https://videocardz.com/newz/intel-a...n-tested-in-opencl-benchmark-faster-arc-a350m

According to golden pig upgrade Ultra 5 125H uses a partly disabled version with 112 EUs and Ultra 5 135H gets the fully enabled version with 128EUs, means 135H and above should have 128 EUs. 185H can't get more than that, 128EU is the maximum Intel offers.
 
dullard

dullard

Elite Member
May 21, 2001
25,932
4,522
126
mikk said:
This not not the case because Intels uses two chips for Meteor Lake mobile, one chip for MTL-M/U and one chip for MTL-H. MTL-H gets a GPU tile with 128EUs and MTL-M/U a GPU tile with 64EUs. Ultra 7 155H gets the fully enabled 128EU version: https://videocardz.com/newz/intel-a...n-tested-in-opencl-benchmark-faster-arc-a350m

According to golden pig upgrade Ultra 5 125H uses a partly disabled version with 112 EUs and Ultra 5 135H gets the fully enabled version with 128EUs, means 135H and above should have 128 EUs. 185H can't get more than that, 128EU is the maximum Intel offers.
Click to expand...
Same frequency for the GPUs? Same supported features like ray tracing? And what of the rest of the power uses: AI, IO, on-board memory, etc? All the same throughout the entire stack?

I honestly don't know what Intel plans. But any one of those could explain lower base clocks.
 
M

mikk

Diamond Member
May 15, 2012
4,292
2,382
136
dullard said:
Same frequency for the GPUs? Same supported features like ray tracing? And what of the rest of the power uses: AI, IO, on-board memory, etc? All the same throughout the entire stack?

I honestly don't know what Intel plans. But any one of those could explain lower base clocks.
Click to expand...


Frequency could differ of course. It's the same chip with same features at the end of the day. Base clock isn't lower on 185H. Not sure if 3.8 Ghz is accurate btw, it's really high for 28W, maybe without GPU load. 2.5 Ghz is nonsense however considering that Raptor Lake 13900 already runs at 2.6 Ghz base. You are talking like we have confirmed base clock speeds which is not true.
 
dullard

dullard

Elite Member
May 21, 2001
25,932
4,522
126
mikk said:
You are talking like we have confirmed base clock speeds which is not true.
Click to expand...
I've said that the table is woefully inadequate, that I don't know Intel's plans, and I specifically mentioned that power numbers in a post are not real but used to explain a concept. I don't see how that adds up to me talking like we have confirmed speeds.

I'm simply stating that lower base clock speeds are quite normal and could have very reasonable explanations. They may be fake numbers, but we don't have any real numbers for the next 4 weeks. Might as well talk about rumors by assuming both sides: that the rumor is false and that the rumor is true. All people here want to do is to talk about rumors they like as being true and rumors they dislike as being false. I'm here to provide the counter-balance.
 
I

igor_kavinski

Lifer
Jul 27, 2020
26,456
18,188
146
  • Haha
Reactions: Mopetar, Tlh97, Saylick and 3 others
tamz_msc

tamz_msc

Diamond Member
Jan 5, 2017
3,865
3,730
136
So I did some digging on these results.

1700115694638.png

This matches closely with my own experience when it comes to gaming. Assuming that you are using GDDR6 and 2GB VRAM is enough for the game and settings, AMD's best iGPU is on par with the MX450.

Now searching for Iris Xe in the Geekbench database is a bit tricky as there is only one entry for both 80 EU and 96 EU configs. But from eyballing various results, I believe your typical 96 EU Iris Xe is going to score around 15000 points.

So this new iGPU is already shaping up to be nearly 1.8x faster in compute.

Now coming to synthetic graphics tests. Notebookcheck reveals that the 780M scores ~2600 in Time Spy while the 96 EU Iris Xe scores ~1500, so the 780M is 1.73x faster.

Finally coming to games, I used Hubwood YT channel's videos to calculate that the median performance difference between the 680M and Iris Xe 96 EU is 1.36x, i.e. the 680M is 36% faster than Intel, while the 780M is just 9% faster than the 680M. So 1.36*1.09 = 1.48. So scaling in games is 1.48x.

Assuming that scaling remains the same, we might see the lead that AMD currently enjoys to vanish. All depends now on driver support.
 
  • Like
Reactions: Tlh97, Saylick and Elfear
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom