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.





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.)

 

[511]

7800X3D gets spanked by 14600K for workload except gaming we can make the same logic there 🤣
N4P is slightly expensive than N5 so approx 14-15K considering N5 is 13-14K

 

[Saylick]

Videocardz has the leaked slide deck in English. Looks like the slides were legit:

https://videocardz.com/newz/intel-core-ultra-200s-arrow-lake-s-desktop-processors-announced-lion-cove-skymont-xe-lpg-npu-and-lga-1851

 

[coercitiv]

Videocardz has the leaked slide deck in English. Looks like the slides were legit:

Came here to complain about not getting some leaked slides yet, lol!

 

[mikk]

If that's not in Integer, rather in combined Int/FP, that's also a significant reduction from original expectations.

32% integer and 72% floating point from the full slide deck on Videocardz.

 

[Abwx]

32% integer and 72% floating point from the full slide deck on Videocardz.

ICC compiled, FTR ADL P core had 43% better IPC than its e core in CB wich is FP, this would be way faster than the previous and current P, wich is not the case.

 

[Henry swagger]

Videocardz has the leaked slide deck in English. Looks like the slides were legit:

https://videocardz.com/newz/intel-core-ultra-200s-arrow-lake-s-desktop-processors-announced-lion-cove-skymont-xe-lpg-npu-and-lga-1851

View attachment 109076
View attachment 109077
View attachment 109081
View attachment 109082
View attachment 109078
View attachment 109079
View attachment 109080

I want to see how arrow lake performs in gaming . Hardware unboxed techpowerup

 

[SteinFG]

reviews are droping on release day (oct 24), so 2 weeks from now

 

[majord]

Calling their direct competition a fellow "Titan" is certainly one of the strangest pieces of marketing in AMD / Intel history surely.

 

[cannedlake240]

32% integer and 72% floating point from the full slide deck on Videocardz.

Those values can't be true, for it to be 1.72x GMT in SpecFP iso freq, Skymont would have to have IPC much higher than a M4 P core

 

[511]

Those values can't be true, for it to be 1.72x GMT in SpecFP iso freq, Skymont would have to have IPC much higher than a M4 P core

Not really Gracemont was far behind

 

[GTracing]

According to anandtech's alder lake review, golden cove is nearly 2x gracemont in spec 2017 fp.

The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity

www.anandtech.com

 

[cannedlake240]

According to anandtech's alder lake review, golden cove is nearly 2x gracemont in spec 2017 fp.

View attachment 109095

The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity

www.anandtech.com

Clocks are different, gracemont 3.9Ghz much lower

 

[GTracing]

Clocks are different, gracemont 3.9Ghz much lower

Yup, and the compiler is different, and we don't know what clock speed Intel ran the "iso-clock" benchmarks at, and the benchmark is "simulated" (not run on the actual hardware?), and they claim a 10% margin for error, and they have a multithread estimate of 55%. Suffice to say, I don't think we'll see +72% in any real world program. Independent reviewers will probably have a hard time replicating Intel's numbers.

But I do think that it's within the realm of possibility that skymont gets within spitting distance of the p core in IPC.

 

[Josh128]

Less.

Cant be much less. Huang is making a million or so per wafer on Blackwell.

 

[naukkis]

Skymont is first x86 fp core being able to do 3 128 bit loads per cycle. Combined with plenty of 128 bit executing units Skymont will have best fp IPC amont x86 cores in some workloads.

 

[coercitiv]

Calling their direct competition a fellow "Titan" is certainly one of the strangest pieces of marketing in AMD / Intel history surely.

Looking at what the non-x86 competition is doing, Titan sounds like an euphemism for dinosaur.

 

[511]

Skymont is first x86 fp core being able to do 3 128 bit loads per cycle. Combined with plenty of 128 bit executing units Skymont will have best fp IPC amont x86 cores in some workloads.

Skymont and N3B are carrying hard

 

[Magio]

Skymont and N3B are carrying hard

Is it? N3B for LNL seemed like it allowed for a significant improvement at low power for Intel vs MTL on Intel 4, but on ARL at high power it's not clear where that improvement is compared to RPL on Intel 7...

Paying 3nm wafer prices for this (as opposed to N3E which actually seems like a serious step forward) has turnout out to be a mistake from Intel in my opinion. And 2 years out from the next desktop refresh, it may be the last mistake they get to make in that segment.

 

[dullard]

Calling their direct competition a fellow "Titan" is certainly one of the strangest pieces of marketing in AMD / Intel history surely.

Where did they call ARM Titan? AMD and Intel aren't competitors. They are on the same x86 team. I'll accept "friendly rival" maybe.

 

[511]

Where did they call ARM Titan? AMD and Intel aren't competitors. They are on the same x86 team.

ARM for foundry buisness

 

[511]

Is it? N3B for LNL seemed like it allowed for a significant improvement at low power for Intel vs MTL on Intel 4, but on ARL at high power it's not clear where that improvement is compared to RPL on Intel 7...

N3B will provide low to mid power efficiency

Paying 3nm wafer prices for this (as opposed to N3E which actually seems like a serious step forward) has turnout out to be a mistake from Intel in my opinion. And 2 years out from the next desktop refresh, it may be the last mistake they get to make in that segment.

They did not make mistake for 3nm it was TSMCs lol they bought the first TSMC 3nm Wafer turned out N3 was scrapped N3B was the first one kind of mid for N3 (improved SRAM by few% and slightly dense vs N3E) flavour N3E is the better one in PPA

 

[Heartbreaker]

Videocardz has the leaked slide deck in English. Looks like the slides were legit:

https://videocardz.com/newz/intel-core-ultra-200s-arrow-lake-s-desktop-processors-announced-lion-cove-skymont-xe-lpg-npu-and-lga-1851

View attachment 109076
View attachment 109077
View attachment 109081
View attachment 109082
View attachment 109078
View attachment 109079
View attachment 109080

I think you left out the most important slide. Gaming vs 14900K. Note under Title: Same Frame Rates.

This is ZERO percent improvement in gaming performance, according to Intel themselves:

 

[igor_kavinski]

This is ZERO percent improvement in gaming performance, according to Intel themselves

So incorrect. This is massive savings on power bills for 24/7 gamers who will be able to buy Nova Lake with the electricity savings from Arrow Lake.

Be positive!

 

[9949asd]

The single and multi core performance uplifts are ok, just the gaming performance not great.
I guess the intel’s new architecture and packaging also need the extra large cache to boost the gaming performance.

 

[dullard]

So incorrect. This is massive savings on power bills for 24/7 gamers who will be able to buy Nova Lake with the electricity savings from Arrow Lake.

Be positive!

I realize that was a joke post, but I'll run with it as I wait 2 more hours for the official embargo to end. I know we focus a lot on small differences in power here. Less power is better, better for costs, and better for the environment. But for the typical gamer, it really isn't as important as you point out.

• The leaked slide shows a geomean average of 73 W less power.
• I'll go with the ridiculous assumption of 24/7 gaming with no breaks for sleep.
• Lets use the average US energy cost of $0.166 / kWhr.
• Lets use the longest delay estimate for Nova Lake that I've seen, Oct 2026. Even though with the cancellation of other CPUs, there are rumors of this date being moved sooner.

0.073 kW * 24 hr use/day * 365 days / year * 2 years * $0.166 / kWhr = $212.31 in those 2 years.

That likely would not buy even the cheapest Nova Lake i5 chip, and certainly a 24/7 gamer would want something better than an i5. And a much more realistic 8 hr gaming per day would be $70.77 over two years.