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.)

 

[AMDK11]

Anyway, I'm of the opinion that RedwoodCove is RedwoodCove, regardless of whether it's the one from GraniteRapids or MeteorLake.What would be the point of designing two different microarchitectures and calling them the same thing?If RedwoodCove in Meteorlake doesn't have the improvements mentioned in GraniteRapids, what's the point of naming them the same in Meteorlake and emphasizing that it's a new P-core?I believe that tests at this stage, where there is no certainty about the microcode or operation at a given frequency, do not provide an image of IPC.I will wait for more details and appropriate tests or the official IPC curve from Intel.

 

[SiliconFly]

Is there a chance that GraniteRapids uses ArrowLake cores?

It's hard to say when we still don't know whether Arrow Lake uses RWC+ or the rumored Lion Cove cores.

Assuming Arrow Lake uses RWC+, it's highly unlikely Granite Rapids also uses RWC+ as Intel usually takes quite some time to introduce new µarch in their server lineup. Fyi, the release dates for both Granite Rapids & Arrow Lake are pretty close.

 

[cebri1]

WILSON: ...That 2x integrated graphics capability...

And not just the design head, I've seen two other videos where Intel people make similar claims. But like i mentioned, we have to take it with a pinch of salt. I think what they really mean is, that the iGPU's processing power has doubled. But that doesn't mean it'll directly translate to double the frame rates due to other obvious limitations mentioned previously.

Driver support has also gotten way better. If they are at 780M level it’s already a significant step forward. I imagine that it also allows them for better margins as OEMs won’t need to spend money on dGPUs

 

[Kepler_L2]

Driver support has also gotten way better. If they are at 780M level it’s already a significant step forward. I imagine that it also allows them for better margins as OEMs won’t need to spend money on dGPUs

Slightly faster than 780M on synthetics at least.

 

[SiliconFly]

Slightly faster than 780M on synthetics at least.

Thats nice. It's a good start. Actually a terrific start I'd say. But is that performance enough?

Driver support has also gotten way better. If they are at 780M level it’s already a significant step forward. I imagine that it also allows them for better margins as OEMs won’t need to spend money on dGPUs

I tried this once before. Lets try again. Having an excellent iGPU is a very good thing for most people. Most browsers nowadays use gpu acceleration for performance. Many other apps use gpu too these days. And it's excellent for light and casual gaming too.

But actual gamers tend to buy laptops with dGPU for performance reasons. And the performance provided by MTL's iGPU (which is around the performance of Arc A370M) just isn't sufficient enough for regular gaming. Today, we need at least a rtx 4050 or 4060 (or amd equivalent) for decent gaming.

In other words, MTL iGPUs won't be replacing dGPUs anytime now I believe. Performance gap is too high.

 

[Mopetar]

Are you aware that the desktop chips only features 32 EUs compared to the mobile chips with 96 EUs? Also AMD didn't release a new desktop lineup yet with RDNA graphics expect for the 2CU version in Ryzen 7000 which is even slower than Intels GT1 Xe LP. It matters much more for mobile.

On paper Xe LPG is twice as fast as Xe LP due to a much higher clock speed and 33% more shader cores.

Nope. I had just assumed it was the same for the mobile chips since I really only buy MacBooks and only get Desktop PCs.

Based on the higher core count and newer technology they shouldn't have as much of a problem as I had thought.

 

[cebri1]

Thats nice. It's a good start. Actually a terrific start I'd say. But is that performance enough?


I tried this once before. Lets try again. Having an excellent iGPU is a very good thing for most people. Most browsers nowadays use gpu acceleration for performance. Many other apps use gpu too these days. And it's excellent for light and casual gaming too.

But actual gamers tend to buy laptops with dGPU for performance reasons. And the performance provided by MTL's iGPU (which is around the performance of Arc A370M) just isn't sufficient enough for regular gaming. Today, we need at least a rtx 4050 or 4060 (or amd equivalent) for decent gaming.

In other words, MTL iGPUs won't be replacing dGPUs anytime now I believe. Performance gap is too high.

Better iGPUs remove the need for dGPUs at the bottom of the stack. Intel iGPUs were very basic, now they can do light gaming. Not saying they are going to replace a 4050 but anything below seems now redundant, and we used to have the MX series and 1030s as dGPUs on mobile. Not arguing with the rest of your point.

 

[SiliconFly]

Tom's Hardware has published a new article about MTL-S. In that, it says: "Both (MTL & ARL) will likely feature graphics tiles made on TSMC's N3E technology."

How can that be? There were many mentions in AT forums that said MTL's tGPU tile is on N5. I thought so too. But Toms Hardware says otherwise!

Is alchemist going to span 3 nodes? N6, N5, N3

Apparently not. Looks like it's all N3E from here on.

 

[eek2121]

Hopefully the GPU performs somewhere close to the A370.

NVIDIA has a lot to lose moving forward. Beefy APUs will steel sales from NVIDIA/AMD GPUs in the future. Note that the APU doesn’t have to be the fastest, it just has ti be good enough to play recent games at okay-ish settings.

Actually, there's more to it. If we compare the leaked benchmarks of Core Ultra 7 with the equivalent i7-13700H (both operating at similar clocks), there's a slight performance regression. Possibly due to the introduction of the NOC interconnect fabric. Or maybe because the MTL samples are all ES.

Or maybe it is because you are comparing a 28W part to a 55W part? In fairness. we don’t know how much/if at all Intel overclocked MTL-P/H, but the die was designed to be a 28W chip.

Tom's Hardware has published a new article about MTL-S. In that, it says: "Both (MTL & ARL) will likely feature graphics tiles made on TSMC's N3E technology."

How can that be? There were many mentions in AT forums that said MTL's tGPU tile is on N5. I thought so too. But Toms Hardware says otherwise!


Apparently not. Looks like it's all N3E from here on.

I doubt it since N3E isn’t done cooking until later this year.

Then again,I suppose there is a possibility, and it would explain the December release and likely limited launch availability.

I just want to see some reviews.

 

[mikk]

Tom's Hardware has published a new article about MTL-S. In that, it says: "Both (MTL & ARL) will likely feature graphics tiles made on TSMC's N3E technology."

How can that be? There were many mentions in AT forums that said MTL's tGPU tile is on N5. I thought so too. But Toms Hardware says otherwise!


Apparently not. Looks like it's all N3E from here on.

Because it's wrong, GPU tile is on TSMC N5 for MTL. Anton Shilov is a clueless guy.

 

[SiliconFly]

Because it's wrong, GPU tile is on TSMC N5 for MTL.

We know N5 & N3E have very different design rules. Intel creating a N5 gpu tile for MTL & another N3E gpu tile for ARL is very very expensive and doesn't really make sense if they're going to ditch the N5 tile so soon. Just not worth it.

Like @eek2121 said, maybe its on N3E and thats what is causing the delay pushing it to a december release!

 

[igor_kavinski]

In other words, MTL iGPUs won't be replacing dGPUs anytime now I believe. Performance gap is too high.

They will still make great christmas presents once their prices are slashed to clear inventory for their next big mobile product push. For a lot of kids, they would make excellent gaming laptops. And since these will outnumber AMD APU laptops, they will reach a much wider audience. As someone's first introduction to PC gaming, they will, for the first time in Intel's history, not suck as much as their previous iGPU efforts. And it forces nGreedia to eat a loss as now they have to basically give away their cheap mobile GPUs so OEMs will consider adding them to their laptop designs so nGreedia can get the marketing they need with their label on the laptop chassis.

 

[Abwx]

Or maybe it is because you are comparing a 28W part to a 55W part? In fairness. we don’t know how much/if at all Intel overclocked MTL-P/H, but the die was designed to be a 28W chip.

So far existing "28W" SKUs boost to 64W , that s the current Intel way, stated as 28W and benched at 64W...

 

[H433x0n]

So far existing "28W" SKUs boost to 64W , that s the current Intel way, stated as 28W and benched at 64W...

That’s the specification listed by Intel for 28W SKUs.

Specs

 

[SiliconFly]

So far existing "28W" SKUs boost to 64W , that s the current Intel way, stated as 28W and benched at 64W...

Maybe. But beats the competition in perf-per-watt. I'd attribute it to MTL's superior design.

 

[Abwx]

Maybe. But beats the competition in perf-per-watt. I'd attribute it to MTL's superior design.

Dunno if they ll beat the competition since the gap is huge...

Numbers from NBC for the 1360P, wich is 4 + 8 and theorically comparable to a an AMD 8C/16T, point to 64W to barely match a 7840U@30W, so they have to improve perf/watt by roughly 2x at this perf level wich is about 12400 pts in CB R23.

 

[Geddagod]

Dunno if they ll beat the competition since the gap is huge...

Numbers from NBC for the 1360P, wich is 4 + 8 and theorically comparable to a an AMD 8C/16T, point to 64W to barely match a 7840@30W, so they have to improve perf/watt by roughly 2x at this perf level wich is about 12400 pts in CB R23.

4+8 should be comparable to a 6 big core cpu. 6+8 would be comparable to a 8 big core cpu

 

[Abwx]

4+8 should be comparable to a 6 big core cpu. 6+8 would be comparable to a 8 big core cpu

Two e cores should match a big core throughput wise, so 4P + 8e = 8P.

If they need more e cores to close the gap then it means that it s a bad concept even for area efficency.

 

[H433x0n]

Dunno if they ll beat the competition since the gap is huge...

Numbers from NBC for the 1360P, wich is 4 + 8 and theorically comparable to a an AMD 8C/16T, point to 64W to barely match a 7840U@30W, so they have to improve perf/watt by roughly 2x at this perf level wich is about 12400 pts in CB R23.

It reportedly does match 13700H performance at half the power consumption.

 

[Geddagod]

Two e cores should match a big core throughput wise,

At what frequency? At what power?
4 e cores take up around the same area as 1 P-core. If Intel gets more perf/area from using the E-core/ P-core strategy, that's just a design choice diff.

 

[Geddagod]

It reportedly does match 13700H performance at half the power consumption.

according to who?

 

[Abwx]

At what frequency?

At same frequency, wich is possible since we re talking of mobile and power at about 30W for the whole chip, at this power level the P cores cant be clocked high in MT, so e cores can be easily clocked at same frequency..

It reportedly does match 13700H performance at half the power consumption.

So the 4 + 8 wont beat the 7840U in perf/watt, the 6 + 8 eventually but not overhelmingly, and that s assuming that Intel are not tricking the numbers as they do since a few years...

 

[H433x0n]

according to who?

Michelle Johnson, she said it in a PCWorld interview iirc. I’m looking for it now.

 

[H433x0n]

Source

It’s around the 1:04 mark in the video. Some of the press materials also explicitly say same performance at half thr power consumption of RPL.

 

[Geddagod]

At same frequency, wich is possible since we re talking of mobile and power at about 30W for the whole chip, at this power level the P cores cant be clocked high in MT, so e cores can be easily clocked at same frequency..

If Intel had no E-cores, they a 6+8 chip would be a 8 big core chip. A 6+4 would be a 6 big core chip. Simply from the floorplan of the CPU itself.
Comparing a 6+4 CPU vs a 8 core CPU is just unfair to the Intel CPU, because it's "8 core competitor" is literally a 6+8 product.