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 (20A)	Arrow Lake (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	4P + 4E	4P + 8E
LLC	24 MB	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.)

 

[maddie]

More than 23%?


At 23%, the fun never stops!

Apple sells about 15-20% of worldwide unit sales. Profits are much higher. The opposite with Intel, server marketshare higher and profits lower when compared to the competition. Not great.

 

[maddie]

There have been a lot of recent members using this same line. Curious coincidence I'm sure.

"Intel has nothing to worry about."

 

[Markfw]

There have been a lot of recent members using this same line. Curious coincidence I'm sure.

"Intel has nothing to worry about."

He calls this nothing to worry about. Its getting creamed. One of my 9554's almost is equal to TWO 8592+'s !!!

Quote from the article number 1 (I will look for more

"At this point, a few things are reasonably clear. Intel does not have the core count to go toe-to-toe with AMD in this generation."

 

[poke01]

Apple sells about 15-20% of worldwide unit sales. Profits are much higher. The opposite with Intel, server marketshare higher and profits lower when compared to the competition. Not great.

Apple sells less phones, well it’s still a lot but less then others at around 150-200 million iPhones sold every year.

But since most phones they are in the $700-$1200 range they make a lot of profit.


Intel also sells the same amount of products in client.

 

[DavidC1]

View attachment 107000

He calls this nothing to worry about. Its getting creamed.

Yea, because they have a core count deficit. That disappears with Granite Rapids. Even if I don't believe they won't take #1, they are probably going to be only about 10% or so away, the best position in servers since Skylake.

If you let your bias influence your decisions, you'll constantly be wrong.

 

[Markfw]

Yea, because they have a core count deficit. That disappears with Granite Rapids. Even if I don't believe they won't take #1, they are probably going to be only about 10% or so away, the best position in servers since Skylake.

If you let your bias influence your decisions, you'll constantly be wrong.

And the 9554 equalling 2 8592's ???? Core count there ? And wattage ? I don't see any actual numbers there. But 2 8592s vs one 9554 ? I doubt they have less.

 

[DavidC1]

And the 9554 equalling 2 8592's ???? Core count there ? And wattage ? I don't see any actual numbers there. But 2 8592s vs one 9554 ? I doubt they have less.

According to that very chart, 60 core 8592 is beating 64 core 9554. EPYC 9554 is also 360W, similar to 8592. And they are both dual sockets.

Like I said, you let your bias influence your decisions, then you are going to be constantly wrong. You have to go solely by data.

 

[H433x0n]

View attachment 107000

He calls this nothing to worry about. Its getting creamed. One of my 9554's almost is equal to TWO 8592+'s !!!

Quote from the article number 1 (I will look for more

"At this point, a few things are reasonably clear. Intel does not have the core count to go toe-to-toe with AMD in this generation."

I think you may be misreading this.

It shows Dual 9554 EPYC (2x64C, 256 threads) vs Dual Platinum 8592+ (2x64C, 256 threads) being close in Linux compile times with the Xeon slightly edging ahead.

Unless you mistyped and meant 9654 EPYC but a single socket 9654 isn’t in this chart.

Edit: fixed core counts

 

[Hitman928]

I think you may be misreading this.

It shows Dual 9554 EPYC (2x64C, 256 threads) vs Dual Platinum 8592+ (2x60C, 240 threads) being close in Linux compile times with the Xeon slightly edging ahead.

Unless you mistyped and meant 9654 EPYC but a single socket 9654 isn’t in this chart.

8592+ is 64 cores.

 

[DavidC1]

8592+ is 64 cores.

I stand corrected.

It's still beating the 64 core EPYC. 128 core Granite Rapids will be very competitive.

Intel has good products coming, which markfw can't see. But that's not Intel's problem. The problem is they are basically a finance company and will fall flat in engineering again.

 

[Markfw]

According to that very chart, 60 core 8592 is beating 64 core 9554. EPYC 9554 is also 360W, similar to 8592. And they are both dual sockets.

Like I said, you let your bias influence your decisions, then you are going to be constantly wrong. You have to go solely by data.

I think you may be misreading this.

It shows Dual 9554 EPYC (2x64C, 256 threads) vs Dual Platinum 8592+ (2x64C, 256 threads) being close in Linux compile times with the Xeon slightly edging ahead.

Unless you mistyped and meant 9654 EPYC but a single socket 9654 isn’t in this chart.

Edit: fixed core counts

You are right I did not see dual both places. But I would like to see actual wattages. My 9554 @%100% is running 282 watts. What is the 8592 REAL actual ?

 

[Abwx]

According to that very chart, 60 core 8592 is beating 64 core 9554. EPYC 9554 is also 360W, similar to 8592. And they are both dual sockets.

Like I said, you let your bias influence your decisions, then you are going to be constantly wrong. You have to go solely by data.

Average power 2P 8592+ 556W.

Average power 2P 9554 377W.

And the 2P 9554 still 3.6% ahead in the perfs average, that s the data.

So that s RPL transfered to the server area.

Bias influence you said.?...

Intel Xeon Platinum 8592+ "Emerald Rapids" Linux Benchmarks Review - Phoronix

www.phoronix.com

 

[Hitman928]

You are right I did not see dual both places. But I would like to see actual wattages. My 9554 @%100% is running 282 watts. What is the 8592 REAL actual ?

Phoronix reviewed the CPU as well, here was the min and avg power consumption measurements across the full test suite:

	Min (W)	Avg (W)
Dual Epyc 9554	25.64	377.42
Dual Xeon 8592+	100.34	556.83

With GNR being on Intel 4, we should expect a significant drop in per core power consumption.

 

[Markfw]

Average power 2P 8592+ 556W.

Average power 2P 9554 377W.

And the 2P 9554 still 3.6% ahead in the perfs average, that s the data.

So that s RPL transfered to the server area.

Bias influence you said.?...

Intel Xeon Platinum 8592+ "Emerald Rapids" Linux Benchmarks Review - Phoronix

www.phoronix.com

I let this speak for me..... But my point still stands. Intel DOES still have something to worry about. And Turin is not out yet......

 

[DavidC1]

I let this speak for me..... But my point still stands. Intel DOES still have something to worry about. And Turin is not out yet......

No one said Intel will be a leader but it'll be much closer this time, within 10%. That's the whole point. SPR is their weakest point in servers.

 

[Markfw]

No one said Intel will be a leader but it'll be much closer this time, within 10%. That's the whole point. SPR is their weakest point in servers.

Now THAT I agree with. If power was competitive, then AMD WOULD have to worry, due to Intels big installed base lead. Power is still a problem, and in the data center, 10 times more than desktop.

 

[511]

You can't actually buy Sierra, even if you wanted to.

Yeah sure this is fake than

X14 Systems - Systems

store.supermicro.com

 

[AMDK11]

I wonder if in the next generation after LionCove we will see a 10-Wide, 8x ALU and 6x FP decoder, or maybe Intel will immediately switch to a 12-Wide, 10x ALU and 8x FP decoder.

 

[FlameTail]

More than 23%?


At 23%, the fun never stops!

In the global smartphone industry. Apple has about 15-20% marketshare in terms of volume shipped.

But in terms of profit, they take home 70-80% of all the profit in the industry. Android OEMs have to fight for the scraps.

 

[Hitman928]

Yeah sure this is fake than

X14 Systems - Systems

store.supermicro.com

Those were all not active when we were looking at them just a few hours ago, lol.

 

[sgs_x86]

Phoronix reviewed the CPU as well, here was the min and avg power consumption measurements across the full test suite:

	Min (W)	Avg (W)
Dual Epyc 9554	25.64	377.42
Dual Xeon 8592+	100.34	556.83

With GNR being on Intel 4, we should expect a significant drop in per core power consumption.

GNR is on Intel 3.

 

[511]

GNR is on Intel 3.

Yes also Intel 7 -> 4 -> 3 is approx 1.20*1.18 = 40% PPW ON A best cae even with no ipc it should improve by 40% ppc best case btw redwood cove on Intel 4 has 20% PPW vs 13700H p core

 

[cebri1]

People get real nervous when there are any chances of good competition between the two. I though that was good for consumers.

 

[DrMrLordX]

Let me try to explain again. After all the lithography steps are taken for a given EUV layer the wafer can be optically scanned for defects. The statement is that if 3 hours are spent performing that optical scan and no defects are found, that optical scan only provides confidence that there are no defects in 62-63% of the 1cm^2 die. If less time was spent on the optical scan the confidence rate would be lower. More time spent would give a higher confidence rate. It says nothing about actual Intel 3 defect density.

So then what was the value of this observation in the first place, that people would see fit to distribute the information? Do we have any reference data from other EUV nodes for comparison?

@Markfw

Phoronix has a pretty good review of Emerald Rapids you can use for reference if necessary:

Intel Xeon Platinum 8592+ "Emerald Rapids" Linux Benchmarks Review - Phoronix

www.phoronix.com

Draw your own conclusions from that.

@Hitman928

Indeed those products weren't even listed for sale a few hours ago. At least now one outlet is selling Sierra Forest!

 

[igor_kavinski]

No one said Intel will be a leader but it'll be much closer this time, within 10%. That's the whole point.

Within 10% in AVX-512 workloads too?

I think Intel saw that threat incoming and embarked on the AMX bandwagon. At least on that front, they have nothing to worry about.