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 (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.)

 

[SiliconFly]

So if I remember right, Lion Cove was worked on by Jim Keller?

It was mentioned that LNC was just one part of his "responsibilities". Not exactly sure what it actually means. It was definitely developed under his supervision, but how much he directly contributed to the μarch is not clear.

 

[SiliconFly]

...Lower clock speeds mean lower power consumption, but also a sign that the core had to be significantly expanded...

Maybe thats one of the reasons why we may not see 8 or 10 P core models in the future. Just speculating.

 

[SiliconFly]

...the CPU portion of MTL as meeting expectations...

You mean performance or efficiency or both? I think RWC is not gonna be any better than previous gen in performance even with microcode updates. Lower single digit performance gains, if any, should be treated more like margin of error if you ask me.

 

[controlflow]

You mean performance or efficiency or both? I think RWC is not gonna be any better than previous gen in performance even with microcode updates. Lower single digit performance gains, if any, should be treated more like margin of error if you ask me.

Efficiency.

GoldenPig claims to have a P code update that fixes an alleged scheduling issue that increased energy consumption on early MTL review units. He posted the updated power efficiency curve which looks quite a bit better. Using Google translate on his post:

"
Regarding energy consumption, there is a scheduling problem due to the early measurement. After updating the pcode, there will be a significant improvement in low power consumption and energy efficiency, the measured U7 155H can have about 10%, compared with the 7840HS, the inflection point has been reduced from 50W to about 35W, which is a significant improvement, but the difference is not much under high power consumption, only 3%. However, due to the early shipment of the first evaluation model, this pcode was basically not updated.
"

I will wait until this update can be confirmed by other sources but it does somewhat line up with the wildly inconsistent results we've seen across different reviews. Hardware Canucks also noted massively inconsistent performance and the seemingly unfinished nature of the SW/FW/Drivers.

I am getting the impression that Intel really rushed to get this out before the end of the year and there are clearly somethings that don't seem fully baked.

 

[Khato]

The mediocre English speaking tech Youtubers should learn how to do a proper HW review from these guys

I can't understand it but the data goes way more in depth than the other reviews I've seen.

Quite nice indeed by comparison. The bit about the current models being part of an early enablement program (EEP) in particular explains the old pcode. Hopefully retail models will have been shipped with the latest, but wouldn't surprise me too much if that's not the case. The CPU package power graph for video playback was also nice to see. Looking forward to seeing how much better MTL gets with further fixes.

 

[Abwx]

The mediocre English speaking tech Youtubers should learn how to do a proper HW review from these guys

I can't understand it but the data goes way more in depth than the other reviews I've seen.

Would be nice if they can redo some of the power and battery tests with the new microcode update. If this data really is accurate, MTL looks pretty good especially when you consider the GPU and gaming benchmarks they also showed.

The efficency curve for Cinebench he display is bs, performance increasing by 23-24% and power only by 42% doesnt exist with the current laws of physics for semiconductors...

That s just PL2 time that is increased to boost the scores, also RAM speed is 6400 for his AMD system and 7457 for the Intel one...

 

[cebri1]

Using that guy's graph combined. Done with paint.net, so not 100% accurate.

 

[SiliconFly]

Efficiency.

GoldenPig claims to have a P code update that fixes an alleged scheduling issue that increased energy consumption on early MTL review units. He posted the updated power efficiency curve which looks quite a bit better. Using Google translate on his post:

"
Regarding energy consumption, there is a scheduling problem due to the early measurement. After updating the pcode, there will be a significant improvement in low power consumption and energy efficiency, the measured U7 155H can have about 10%, compared with the 7840HS, the inflection point has been reduced from 50W to about 35W, which is a significant improvement, but the difference is not much under high power consumption, only 3%. However, due to the early shipment of the first evaluation model, this pcode was basically not updated.
"

I will wait until this update can be confirmed by other sources but it does somewhat line up with the wildly inconsistent results we've seen across different reviews. Hardware Canucks also noted massively inconsistent performance and the seemingly unfinished nature of the SW/FW/Drivers.

I am getting the impression that Intel really rushed to get this out before the end of the year and there are clearly somethings that don't seem fully baked.

Thanks for the clarification. Now it all makes sense.

I was scratching my head on why there was so much variation in efficiency in many of the results.

 

[SiliconFly]

Using that guy's graph combined. Done with paint.net, so not 100% accurate.

View attachment 90518

Wow!

 

[SiliconFly]

The efficency curve for Cinebench he display is bs, performance increasing by 23-24% and power only by 42% doesnt exist with the current laws of physics for semiconductors...

That s just PL2 time that is increased to boost the scores, also RAM speed is 6400 for his AMD system and 7457 for the Intel one...

It's not a miracle. Buggy microcode tend to give wrong results. Both the updated results and the buggy pcode results are not directly related and are not relative to one another. It's a bug fix. Results may vary according to the bugs fixed.

 

[FlameTail]

W
Chinese reviewers

 

[SiliconFly]

Lets hope we get more power efficiency reviews with the new pcode update and see whether it really works as claimed.

 

[Abwx]

It's not a miracle. Buggy microcode tend to give wrong results. Both the updated results and the buggy pcode results are not directly related and are not relative to one another. It's a bug fix. Results may vary according to the bugs fixed.

A bug whose correction improve perf/watt by 20%..?..
And you think that Intel engineers were unaware of the thing till launch day..?.

 

[João Bortolace]

155h faster than the 780m with slower memory (5600 vs 6400) impressive

 

[SiliconFly]

A bug whose correction improve perf/watt by 20%..?..
And you think that Intel engineers were unaware of the thing till launch day..?.

Depends on the errata. This is a new product launch. Show stopper bugs take precedence rather than performance or efficiency bugs. I'm sure they still have their hands full and are working at a feverish pace at the moment clearing out the backlog.

Good news is, if they've come this far along fixing efficiency bugs, then it means they've fixed more serious bugs already and the product is almost ready for mass market.

 

[SiliconFly]

155h faster than the 780m with slower memory (5600 vs 6400) impressive

Ya. Saw that. And it's inline with older predictions. MTL's iGPU performs more like GTX 1650 with the addition of RTX & XeSS (thats well below a RTX 3050).

Perfect for entry level gaming (not suitable for real gamers).

 

[PJVol]

Lets hope we get more power efficiency reviews with the new pcode update and see whether it really works as claimed.

I found it hilarious how could Intel waste time and resources to squeeze every possible bit of PPW (let alone the usual snake oil things) to make a new product look better at launch, when it turns out there's simply new firmware (magically appearing the next day) to improve PPW as much as shown in drawings.

 

[SiliconFly]

I found it hilarious how could Intel waste time and resources to squeeze every possible bit of PPW (let alone the usual snake oil things) to make a new product look better at launch, when it turns out there's simply new firmware (magically appearing the next day) to improve PPW as much as shown in drawings.

Intel should have used a little bit of snake oil themselves.

Actually, Intel ARC graphics drivers turned out to be the biggest joke of the year when launched as it took more than a year for the drivers to mature. Funny thing is, their drivers are still not top-notch and they're still improving it. God only knows when it'll be on par with competition.

 

[DrMrLordX]

This is MTL right now:

Expensive packaging technology: Check
New cores: Check
New node: Check
New branding: Check
New graphics tiles: Check

Also: has no presence on desktop. None, zip, zilch, nada. Forcing Intel to refresh a 10nm product instead (again!). This thing is really disappointing.

Insane lol

Ironic.

I found it hilarious how could Intel waste time and resources to squeeze every possible bit of PPW (let alone the usual snake oil things) to make a new product look better at launch, when it turns out there's simply new firmware (magically appearing the next day) to improve PPW as much as shown in drawings.

The incredible variation in review results is troublesome. It's hard to know whom to trust when it comes to the real performance of Meteor Lake. Will the real Meteor Lake please stand up?

 

[SiliconFly]

i9-15900K is gonna have a pretty tough time trying to beat the upcoming i9-14900KS. I doubt that it will go higher than 5.5 GHz so it's gonna need some crazy IPC to level the playing field and then also need to have a pretty strong IMC to get the edge in MT workloads with higher speed DDR5. Forget beating AMD. Intel has enough of a challenge beating themselves.

Lets forget 5.5 GHz for a while. Considering the development timeline of 20A, it's be a wonder if 20A can comfortably sustain even 4.5 GHz imho. Intel never had the time to optimize 20A for efficiency at higher frequencies considering the short development time.

For comparison, Intel 10nm (and Intel 7) has had a 6 year run (including RPL-R). Intel 4 was well into development even before Pat came into the picture. 20A is relatively new with a host of significant upgrades like GAA & PowerVia coupled with a short development time. Just won't be hitting higher frequencies like 5 GHz any time now (at least during its first iteration). And it's not exactly a surprise. Intel would have factored it in already.

Thats one of the key reasons i believe LNC is gonna be more of a 4 GHz part like Apple SoC, rather than a 5+ GHz part like RWC.

 

[AMDK11]

where's the L1 cache sizes from?
L2 is supposedly 3MB- but its split into two- prob a 2.5MB L2 and 0.5MB L 1.5. Apparently LNC adds another level of cache.

There was a leak once with a screenshot of, if I remember correctly, HWiNFO, a sample of supposedly ArrowLake. The screenshot showed L1 consistent with the one in RedwoodCove, but there were also new instructions. I'm looking now but I can't find it. If I find it, I'll post it.

Plus, the previous generation for Intel has always been a good indication of cache, and there's usually no less in the next generation.

 

[FlameTail]

Behold this masterpiece!

And put together by none other than the legend Ian Cutress!

 

[TESKATLIPOKA]

155h faster than the 780m with slower memory (5600 vs 6400) impressive

Difference in using faster memory is only 4%, that's good news in case Intel will later increase the EU count.

155h with 7467Mhz is 9% faster than 7840HS 6400MHz, not sure how much improvement we would see If AMD used the same memory.

My question is at what power Intel IGP is faster? 35W, 45W or more?

 

[Meteor Late]

The efficency curve for Cinebench he display is bs, performance increasing by 23-24% and power only by 42% doesnt exist with the current laws of physics for semiconductors...

That s just PL2 time that is increased to boost the scores, also RAM speed is 6400 for his AMD system and 7457 for the Intel one...

Yeah, what if this "pcode" update (if it's even real) all it does is boost PL2 time or power? I don't even know if the reviews we've seen that test power disable PL2 or not? because it's pointless to compare at 28W or at 45W if PL2 is much higher and affects scores significantly. I've a 12500H and I know these 13500H scores of 14200 at "65W" are absolute BS, it barely gets above 12000 points when 65W enforced properly.
So that chinese youtube review is not enforcing power limits properly, I've seen Alder Lake and Raptor Lake have similar efficiency as demonstrated by Jarrod Tech when comparing 12700H vs 13700H, so 13500H is just not getting like 15%-18% higher points at same power than my 12500H.

 

[Hulk]

Maybe thats one of the reasons why we may not see 8 or 10 P core models in the future. Just speculating.

Intel seems dedicated to the hybrid approach. If anything P core count will go down I think.