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.)

 

[SiliconFly]

Efficiency has never been Intel's primary goal, performance always has been.

I believe it changes with LNC, with efficiency being their primary goal.

 

[H433x0n]

Yeah, I saw It, but that needs to be verified. Even If It's true, It still loses at low TDPs.

What I want to know is If those ULV E-cores are actually working during benches or not.

If that’s the power curve then it’s fine as a product. It wouldn’t be perfect but it has a good argument going for it since you can’t really argue that Phoenix outperforms it or is noticeably more efficient. A better binned 165H would do slightly better also.

 

[Saylick]

Golden pig Upgrade posted a Cinebench power/efficiency curve with a new firmware. Depending on the firmware this might explain the differences between some tests. In another posting he says the latest Arc test driver improves low power gaming performance for the Arc tGPU, the driver is expected in 1 month.

You’ve got to be (expletive) me. I almost hope that’s not really the case. If that’s a legit power curve graph then I’d say it basically met expectations and it’s be totally irrelevant since they’ve already got roasted by poor day 1 performance.

Yikes. Looks like Pat ordered MTL to be taken out the oven a little undercooked so that Intel could meet their promise of MTL before 2024, but if it mars early benchmarks it might be penny wise, pound foolish. First impressions mean a lot to consumers, and having immature drivers and firmware is basically pulling an AMD.

 

[adroc_thurston]

You mean the waiting?

No.
Things in the mirror something something

LNC architecture is still under wraps

No it's not, ARL compute tile taped out some time ago.

Looks like Pat ordered MTL to be taken out the oven a little undercooked so that Intel could meet their promise of MTL before 2024

It's literally been delayed 3 times because CPU was breaking itself apart with weird deadlocks.
Compute tile is C0 step.
It's the most delayed intel client part in years.

 

[cebri1]

Ignoring AMD comparisons, which are still ahead at lower power limits. That would mean a pretty nice jump over RPL.

 

[Khato]

Heh, looks like most of the reviews were with something like half a year old pcode? Somewhat surprising, but does explain the results.

I'm still firmly in favor of day 0 reviews versus having to wait a few months after a paper launch though. Especially since the 'miraculous' improvement will just net them another news cycle.

 

[Geddagod]

View attachment 90488

Ignoring AMD comparisons, which are still ahead at lower power limits. That would mean a pretty nice jump over RPL.

If this is true, the jump here is pretty big. That's a ~10% jump.
Also, the 20% jump in PPW seems much more clear with this.

 

[Meteor Late]

View attachment 90488

Ignoring AMD comparisons, which are still ahead at lower power limits. That would mean a pretty nice jump over RPL.

I'm very skeptical of this, if this is the case, how come Intel showed 8% more performance at similar power in official slides (compared to 1370P)? do you think Intel would show the product in a worse light in their slides by running the tests with old pcode?
I'm going to call this "new pcode" BS for now, unless it's replicated by more reputable reviewers.

 

[cebri1]

If this is true, the jump here is pretty big. That's a ~10% jump.
Also, the 20% jump in PPW seems much more clear with this.

Isn't the 20% jump at ISO frequencies? I think could be due to improved IPC from new e-cores. Real PPW shouldn't be that high, I think we need more independent reviews.

edit: Sorry 40% was at ISO frequency. 20% is at ISO power. So with additional e-core IPC makes a lot more sense.

I'm very skeptical of this, if this is the case, how come Intel showed 8% more performance at similar power in official slides (compared to 1370P)? do you think Intel would show the product in a worse light in their slides by running the tests with old pcode?

Well, they were using SPEC not CB. But now seems too good, we need more in-depth reviews.

 

[Geddagod]

I'm very skeptical of this, if this is the case, how come Intel showed 8% more performance at similar power in official slides (compared to 1370P)? do you think Intel would show the product in a worse light in their slides by running the tests with old pcode?

A thing with laptop reviews people always have to be cautious abt is PL1 vs PL2 impacting scores.
But GoldenPig was actually the one with diff results than others, at 28W Jarods tech found MTL to be 14% ahead of RPL, Just Josh 16%, and this test makes the PPW gain 21%. That's still a surprisingly wide variance though.

Isn't the 20% jump at ISO frequencies? I think could be due to improved IPC from new e-cores. Real PPW shouldn't be that high, I think we need more independent reviews.



Well, they were using SPEC not CB. But now seems too good, we need more in-depth reviews.

Nodes help with perf iso power, archs help with perf iso frequency (and also iso power, depending on how goated you are lol).
I agree, I'm still waiting for more reviews. MTL's PPW gains are all over the place.

 

[Hitman928]

I'm very skeptical of this, if this is the case, how come Intel showed 8% more performance at similar power in official slides (compared to 1370P)? do you think Intel would show the product in a worse light in their slides by running the tests with old pcode?
I'm going to call this "new pcode" BS for now, unless it's replicated by more reputable reviewers.

If someone can review an actual retail unit, that should confirm whether or not this new firmware uplift is real. If it is, then it's definitely puts MTL in a better light. Still no where near the hype/marketing, but better than initial reviews. Unfortunately it looks like we'll probably have to wait until January to get those reviews.

 

[H433x0n]

If someone can review an actual retail unit, that should confirm whether or not this new firmware uplift is real. If it is, then it's definitely puts MTL in a better light. Still no where near the hype/marketing, but better than initial reviews. Unfortunately it looks like we'll probably have to wait until January to get those reviews.

We need to know the microcode that GoldenPig is showing and then compare it to the retail laptop. Unfortunately I can't speak Mandarin so I can't ask him.

 

[Hitman928]

A thing with laptop reviews people always have to be cautious abt is PL1 vs PL2 impacting scores.
But GoldenPig was actually the one with diff results than others, at 28W Jarods tech found MTL to be 14% ahead of RPL, Just Josh 16%, and this test makes the PPW gain 21%. That's still a surprisingly wide variance though.

Nodes help with perf iso power, archs help with perf iso frequency (and also iso power, depending on how goated you are lol).
I agree, I'm still waiting for more reviews. MTL's PPW gains are all over the place.

I don't know about the reviews you are quoting, but I know at least some of the reviews I saw were using just using the PL1 as the power number and not taking actual power measurements so then the efficiency difference could look much better or worse than it really is.

 

[H433x0n]

I'm very skeptical of this, if this is the case, how come Intel showed 8% more performance at similar power in official slides (compared to 1370P)? do you think Intel would show the product in a worse light in their slides by running the tests with old pcode?
I'm going to call this "new pcode" BS for now, unless it's replicated by more reputable reviewers.

Wild that people hate Intel so much that they create dedicated troll accounts.

 

[trivik12]

I wont say Meteor Lake is a paper launch. You can buy laptops today at big online retailers. Amazon/Best Buy/Costco/BH photo video etc. Someone posted early impressions getting an MSI Prestige 16 from Microcenter and Newegg also is selling them.

Its not anywhere near what we get for Intel based laptops but it will take few months for this one to dominate shelfs. But at least you can buy one if you choose to. I dont see any Hawk Point Laptop available despite being "announced" a week ago.

 

[mikk]

View attachment 90488

Ignoring AMD comparisons, which are still ahead at lower power limits. That would mean a pretty nice jump over RPL.

Acer Swift Go 14 from Computerbase

155H 35W= 12833
155H 45W= 13936

Difference is big, this needs verification. Golden Pig Upgrade has access to something the usual tester won't have, good chance it's true.

 

[H433x0n]

Acer Swift Go 14 from Computerbase

155H 35W= 12833
155H 45W= 13936

Difference is big, this needs verification. Golden Pig Upgrade has access to something the usual tester won't have, good chance it's true.

The delta at 35W is 5%, delta at 45W is 13%. It's a big difference but I don't find it implausible.

 

[Hulk]

They are 20-30% behind AMD in the sub 30 watt range.
Funnily enough that's also how much power each core would get in a server environment as well.

I thought when you consider the time spent only on the SoC in many use cases the sub 30W usage was closer. But I will defer to your knowledge.

Intel "almost" caught up but are still behind in the sub 30W range.

 

[Hulk]

Wake me up when Intel catches upto Apple in efficiency

Apple has some remarkable products and my kids have iPhones and iPads but they are not for me. I love my PC's. You know, different strokes and all that. Apples GUI is absolutely confounding to me. My first experience with computers was programming Basic and Assembly. I need more insight into the machine than Apple provides. And yes, I'm a small minority of the population. I do admit that I was NOT sad to see DOS go like my older brother was. And yes, I know DOS is still there, but now we rarely have to access it.

 

[Hulk]

I believe it changes with LNC, with efficiency being their primary goal.

I think efficiency was Intel's primary goal and until efficiency couldn't be their primary goal. Remember how they used to say any increase in power use must bring a larger percentage in performance. With Raptor Lake they loosened that up bit. It's more like ten times more power only need bring 1 times more performance. At least that's how it is with Raptor from about 5.4GHz on up.

 

[Geddagod]

I don't know about the reviews you are quoting, but I know at least some of the reviews I saw were using just using the PL1 as the power number and not taking actual power measurements so then the efficiency difference could look much better or worse than it really is.

Just Josh measured both initial and sustained watts

He also tested sustained vs initial scores, with sustained taking 10 mins

You can see for chips that have a large difference between initial and sustained watts, there's also a large difference in initial vs sustained multicore performance. For example, the Inspiron 13700h has a 16% difference between the two measurements, because it's initial power is nearly 2x as much as the sustained power.

Lastly, something interesting I want to point out too from GoldenPig's review is using the 7840HS vs 7840U. I suspect the 7840U is much better at lower wattages than the HS.

I thought when you consider the time spent only on the SoC in many use cases the sub 30W usage was closer. But I will defer to your knowledge.

Intel "almost" caught up but are still behind in the sub 30W range.

If the Golden Pig data is true from the new RWC P-code, I would be wrong. I would wait for more testing, perhaps many of us were too quick with our disappointment with MTL.

I think efficiency was Intel's primary goal and until efficiency couldn't be their primary goal. Remember how they used to say any increase in power use must bring a larger percentage in performance. With Raptor Lake they loosened that up bit. It's more like ten times more power only need bring 1 times more performance. At least that's how it is with Raptor from about 5.4GHz on up.

Honestly it may have started much earlier. I think it started with SNC to be honest. SKL might have been the last arch that followed careful deliberation of perf/power increases for archs- iso node, along with impact of area. The last truly "good" Intel P-core arch might be even further behind SKL.

 

[SiliconFly]

Apple has some remarkable products and my kids have iPhones and iPads but they are not for me. I love my PC's. You know, different strokes and all that. Apples GUI is absolutely confounding to me. My first experience with computers was programming Basic and Assembly. I need more insight into the machine than Apple provides. And yes, I'm a small minority of the population. I do admit that I was NOT sad to see DOS go like my older brother was. And yes, I know DOS is still there, but now we rarely have to access it.

I like DOS too. But sadly, it can be run only on VMs in UEFI systems these days.

 

[SiliconFly]

Just Josh measured both initial and sustained watts
View attachment 90492
He also tested sustained vs initial scores, with sustained taking 10 mins
View attachment 90491
You can see for chips that have a large difference between initial and sustained watts, there's also a large difference in initial vs sustained multicore performance. For example, the Inspiron 13700h has a 16% difference between the two measurements, because it's initial power is nearly 2x as much as the sustained power.

Lastly, something interesting I want to point out too from GoldenPig's review is using the 7840HS vs 7840U. I suspect the 7840U is much better at lower wattages than the HS.

If the Golden Pig data is true from the new RWC P-code, I would be wrong. I would wait for more testing, perhaps many of us were too quick with our disappointment with MTL.

Honestly it may have started much earlier. I think it started with SNC to be honest. SKL might have been the last arch that followed careful deliberation of perf/power increases for archs- iso node, along with impact of area. The last truly "good" Intel P-core arch might be even further behind SKL.

If possible, can you kindly elaborate on why there's so much of a difference for the same cpu?

 

[AMDK11]

Delete

 

[igor_kavinski]

View attachment 90491

Thank you so much for posting that. Very interesting.



I know there are many factors to choosing a laptop but for performance reasons, I would find it difficult not to be swayed towards AMD by that comparison. And remember, AMD has working AVX-512.