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.)

 

[CopernicusLazarus]

So far AMD has currently 60% better efficency

No. Best case it's 10-15% at same power, worst case it's losing at lower power levels. That's exactly what computerbase also measured.

 

[Gideon]

Computerbase does these tests and intel is always the most efficient cpus on their testing. Here is the most recent testing they've done, the 13900k is the most efficient cpu they ever tested at 45w

....

While discussing CPU efficency is all fine and good, discussing multi-core power efficiency of CPUs with significantly different core counts and drawing conclusions of "architectural efficiency" from that, can result in meaninglessoversimplifications pretty quickly.


When discussing multi-core power efficency It's worth remembering:

1. More cores clocked lower always gives a better result (due to the nature of voltage/freq scaling)
   • a 16 core CPU is always significantly more efficient than its 12 or 8 counterparts
   • this is part of the reason why lots of efficency cores do really well in MT benchmarks
     • (so do zen4c cores on servers - as they would in theoretical 24c desktop CPUs)
2. Chiplets on organic substrate (AMD's cheap and dirty solution) really starts to take it's toll at low TDPs (say 45W) where the "uncore" power starts to dwarf the cores own power usage
   • ( In AMDs case It has to route all the memoryand I/O traffic between the two CCX chiplets and the I/O dies through the organic substrate package)

No. Best case it's 10-15% at same power, worst case it's losing at lower power levels. That's exactly what computerbase also measured.

It loses by 6% at 45W (reasons discussed above) and wins at every other power level (direct comparisons being 13%, 19% and 21%):

The verdict on computerbase (source article here):

1. Intel's new CPUs benefit from the high and even higher maximum consumptions much more significantly than the Ryzen 7000. There is no trace of efficiency from the factory, but the performance continues to rise.
2. Nevertheless, Intel cannot quite catch up on the Ryzen 9 7950X even with the open Core i9-13900K, which consumes much less.
3. Compared to Raptor Lake to Alder Lake, the jump in efficiency is clear. In each TDP class, the new CPUs are on the road much faster than the old ones.
4. Against Ryzen 7000, Raptor Lake is also still losing. Only the Core i9-13900K can achieve a respect for the assembled competition with 16 E-Cores at 45 watts. Otherwise, there is at least one competing product in each class that is more performant.

Overall I must say, that I'm quite impressed at the overall performance and efficiency Intel is able to churn out from the "formerly known as the 10nm+++" node, especially if they improve it with the refresh. Even if Arrow Lake has limited ST gains, the MT / efficiency gains should be quite large due to the process jump alone

 

[CopernicusLazarus]

While discussing CPU efficency is all fine and good, discussing multi-core power efficiency of CPUs with significantly different core counts and drawing conclusions of "architectural efficiency" from that, can result in meaninglessoversimplifications pretty quickly.


When discussing multi-core power efficency It's worth remembering:

1. More cores clocked lower always gives a better result (due to the nature of voltage/freq scaling)
   • a 16 core CPU is always significantly more efficient than its 12 or 8 counterparts
   • this is part of the reason why lots of efficency cores do really well in MT benchmarks
     • (so do zen4c cores on servers - as they would in theoretical 24c desktop CPUs)
2. Chiplets on organic substrate (AMD's cheap and dirty solution) really starts to take it's toll at low TDPs (say 45W) where the "uncore" power starts to dwarf the cores own power usage
   • ( In AMDs case It has to route all the memoryand I/O traffic between the two CCX chiplets and the I/O dies through the organic substrate package)

It loses by 6% at 45W (reasons discussed above) and wins at every other power level (direct comparisons being 13%, 19% and 21%):

The verdict on computerbase (source article here):


Overall I must say, that I'm quite impressed at the overall performance and efficiency Intel is able to churn out from the "formerly known as the 10nm+++" node, especially if they improve it with the refresh. Even if Arrow Lake has limited ST gains, the MT / efficiency gains should be quite large due to the process jump alone

I think im in agreement with everything you said. The ultimate efficiency goes to the 13900k, but in higher power levels amd can keep up or even surpass it, but just as you said, not a huge difference either way.

The problem is lighter workloads where because of the chiplets, amd becomes a power hog. Consumes more while being slower at the same time.

 

[SiliconFly]

Computerbase does these tests and intel is always the most efficient cpus on their testing. Here is the most recent testing they've done, the 13900k is the most efficient cpu they ever tested at 45w

View attachment 87230

This test shows 7950X is less efficient than 13900K

 

[Geddagod]

Computerbase does these tests and intel is always the most efficient cpus on their testing. Here is the most recent testing they've done, the 13900k is the most efficient cpu they ever tested at 45w

View attachment 87230

...Because Intel doesn't have to deal with IO die issues at 45 watts? Compare this with any mobile AMD sku, which gets to be monolithic, and you will see them crush their competing Intel parts at 45 watts.
Or, just look at the better binned 7945HX, which still has to deal with IO die issues, that also prob better power allocation between core vs IO



So esentially, here's the story. For desktop, for normal consumer power draws, both under gaming, or application workloads, AMD is ahead in efficiency. In idle, it's esentially a tie.
In mobile, AMD is better there too. Has a great lead in battery life as well. Really the only place where Intel can stake a claim is battery life in the desktop replacement class, so 7945HX vs 13980HX, but that's also a pretty useless metric to compare because no one is going to be lugging around their 2 pound desktop replacement laptops for school. There's a reason the segment is dubbed "desktop replacement"...
Intel can win all the useless efficiency metrics they want ig, because that's exactly what looking at the 45 watt power consumption of a 7950x and 13900k is- useless.

 

[CopernicusLazarus]

...Because Intel doesn't have to deal with IO die issues at 45 watts?

And that's a problem they need to fix, how is that relevant? AMD doesn't have to deal with 10 nm but instead use TSMC's 5. That's an intel problem that they need to fix. Why would you even bring that up?

.So esentially, here's the story. For desktop, for normal consumer power draws, both under gaming, or application workloads, AMD is ahead in efficiency. In idle, it's esentially a tie.

Neither of those are true though. There are plenty (actually, the vast majority) of applications and games that don't draw over 40-50-60 watts. Even creativity applications like the whole adobe suite etc do not use that much power. 99% of day to day tasks Intel is both faster and more efficient.


Also, how can idle power be a draw? You just said AMD has the IO die that consumes power, that's why it doesn't scale properly at 45w. So how does it idle at the same levels as Intel? That doesn't make sense to me.

Anyways, everyone can have an opinion, nobody really cares. What matters are the facts. Unless computerbase's testing is flawed or they straight up lying, Intel has the most efficient desktop CPU's. According to their numbers of course.

 

[SiliconFly]

99% of day to day tasks Intel is both faster and more efficient.

Exactly what I tried to say too. 7950X3D & 13900K have pretty much similar power efficiency in real world workloads. And our people here are continuously quoting random synthetic benchmarks that doesn't make much of a difference in the real world.

Thats the same reason I also mentioned that MTL can be up to 2X power efficient compared to previous gen. Even though Intel 7 -> Intel 4 ppw gain is only 20%, MTL overall power-efficient SoC design might easily deliver what Intel claims. Meaning, 2X the efficiency in real world workloads.

 

[Gideon]

Exactly what I tried to say too. 7950X3D & 13900K have pretty much similar power efficiency in real world workloads. And our people here are continuously quoting random synthetic benchmarks that doesn't make much of a difference in the real world.

This is absolutely measurably not true:

AMD Ryzen 9 7950X3D Review - Best of Both Worlds

The Ryzen 9 7950X3D is the spearhead of the AMD Zen 4 X3D lineup. In our performance review we test AMD's new 16-core flagship with dozens of applications and 14 games at up to 4K, to see whether AMD can take the performance crown from Intel, how power consumption is affected, and what can be...

www.techpowerup.com

And while a number of these are synthetic benchmarks, many are not:

 

[FangBLade]

I see Pat Gelsinger has hired a few people to strengthen the marketing, even a new account has emerged, lol, and the amount of lies they've told speaks for itself. I have nothing more to comment on that. Have a pleasant rest of the day.

 

[CopernicusLazarus]

This is absolutely measurably not true:

AMD Ryzen 9 7950X3D Review - Best of Both Worlds

The Ryzen 9 7950X3D is the spearhead of the AMD Zen 4 X3D lineup. In our performance review we test AMD's new 16-core flagship with dozens of applications and 14 games at up to 4K, to see whether AMD can take the performance crown from Intel, how power consumption is affected, and what can be...

www.techpowerup.com

And while a number of these are synthetic benchmarks, many are not:

Any comment on this? It's from technotice, he tested photoshop premiere etc, there is a huge discrepency in efficiency between amd and intel. Not only is the 13900k faster than the 7950x, it also consumed much less power.

 

[CopernicusLazarus]

I see Pat Gelsinger has hired a few people to strengthen the marketing, even a new account has emerged, lol, and the amount of lies they've told speaks for itself. I have nothing more to comment on that. Have a pleasant rest of the day.

This is a very important truth that people need to understand. Indeed, Intel supposed fans are just on a payroll from Intel. AMD fans on the other hand spread misinformation for free.

 

[FangBLade]

This is a very important truth that people need to understand. Indeed, Intel supposed fans are just on a payroll from Intel. AMD fans on the other hand spread misinformation for free.

I recognized your writing style immediately, you started in your own style from the very first message.

 

[CopernicusLazarus]

I recognized your writing style immediately, you started in your own style from the very first message.

Please, stay on topic man. If you have something to say about the graphs posted, go ahead so we can discuss them. Im leaving open the possibility that computerbase is intel biased, but until that's shown it's just a fact that intel has the more efficient cpus.

Also, you are this balanced and impartial individual called "dance on intels grave" right?

Spoiler

 

[Abwx]

No. Best case it's 10-15% at same power, worst case it's losing at lower power levels. That's exactly what computerbase also measured.

But efficency should be measured at same throughput, by increasing the work for the most efficient one you reduce its efficiency.
Efficency comparison is consumed power for a same task accomplished in the same time.

For instance in Geddagod post you can see that a 13950HX need 115W to do the work in the same time as a 7945HX@60W, in this case the latter is almost 2x more efficient at this perf level.

 

[CopernicusLazarus]

But efficency should be measured at same throughput, by increasing the work for the most efficient one you reduce its efficiency.

I disagree, but still, even with that in mind, I posted a graph up there, the 13900k was at the same time doing more work and consuming less power. In the graph above it's twice as efficient as the 7950x.

I don't know why you are bringing up mobile parts though, we are talking about the desktop parts here. Yes, mobile amd has the edge in efficiency, but not on desktops.

 

[Abwx]

I disagree, but still, even with that in mind, I posted a graph up there, the 13900k was at the same time doing more work and consuming less power. In the graph above it's twice as efficient as the 7950x.

I don't know why you are bringing up mobile parts though, we are talking about the desktop parts here. Yes, mobile amd has the edge in efficiency, but not on desktops.

You can disagree as much as you want you ll keep being wrong.

Imagine that you re measuring efficency of two cars and one is required to travel a same distance in 20% lower time than the other, that s basically your flawed vision of what is efficency.

To do so you ll have to increase its speed (throughput for a CPU) by 25% comparatively to the slower one, and its comsumption by unit of distance will increase by 60%, that s the trick you re using to twist the numbers.

Edit : For DT parts Computerbase number show the 13900K@142W being 3% faster than the 7950X@88W, so the latter need about 95W to match the 13900K@142W, that s 50% better efficency at this same perf level.

Intel Core i9-13900KS mit 6,0 GHz im Test

Der Core i9-13900KS ist der erste Intel-Prozessor mit bis zu 6,0 GHz ab Werk. Die dritte Special Edition in Anwendungen und Spielen im Test.

www.computerbase.de

 

[FangBLade]

Please, stay on topic man. If you have something to say about the graphs posted, go ahead so we can discuss them. Im leaving open the possibility that computerbase is intel biased, but until that's shown it's just a fact that intel has the more efficient cpus.

Also, you are this balanced and impartial individual called "dance on intels grave" right?

Spoiler

View attachment 87237

On every forum, with every account, you spin the same story, "Intel is the king of efficiency", that's why they banned you everywhere.

 

[CopernicusLazarus]

You can disagree as much as you want you ll keep being wrong.

Imagine that you re measuring efficency of two cars and one is required to travel a same distance in 20% lower time than the other, that s basically your flawed vision of what is efficency.

To do so you ll have to increase its speed (throughput for a CPU) by 25% comparatively to the other, and its comsumption by unit of distance will increase by 60%, that s the trick you re using to twist the numbers.

But in this case it doesn't even matter that I disagree, since Intel seems to be both faster and more efficient at the same time. Why don't you actually take a look at the graph posted eg in premiere pro? The 7950x consumed 55% more power than the 13900k while being slower. If we follow your methodology, in order for the 7950 to match the speed of the 13900k it will end up consuming twice the power.

By the way, performance per watt means literally that, how much performance you get per every single watt. I think it's your understanding of efficiency that's wrong.

 

[Abwx]

But in this case it doesn't even matter that I disagree, since Intel seems to be both faster and more efficient at the same time. Why don't you actually take a look at the graph posted eg in premiere pro? The 7950x consumed 55% more power than the 13900k while being slower. If we follow your methodology, in order for the 7950 to match the speed of the 13900k it will end up consuming twice the power.

If you re talking of Premiere Pro it s not the 13900K CPU computing that is more efficient, its due to Premiere Pro using GPU encoding and Quicksync with Intel CPUs, i m sure you perfectly know it and are blatantly trying to mislead the general public.

Beside i edited my post with Computerbase measurements of power comsumption over a large panel of MT softwares, check the numbers rather than serving us intel sales force marketing arguments, this is a technical forum and people are not that easy to fool.

 

[CopernicusLazarus]

If you re talking of Premiere Pro it s not the 13900K CPU computing that is more efficient, its due to Premiere Pro using GPU encoding and Quicksync with Intel CPUs, i m sure you perfectly know it and are blatantly trying to mislead the general public.

Beside i edited my post with Computerbase measurements of power comsumption over a large panel of MT softwares, check the numbers rather than serving us intel sales force marketing arguments, this is a technical forum and people are not that easy to fool.

And if you are talking about gaming it's not the 7950x 3d CPU computing that is more efficient, it's the cache. So? The end result is the same. But nevermind premiere pro as well. Computerbase has the 13900k as the most efficient CPU at 45w. How does that happen?

 

[Abwx]

By the way, performance per watt means literally that, how much performance you get per every single watt. I think it's your understanding of efficiency that's wrong.

For every single watt at a same troughput for that watt.

And if you are talking about gaming it's not the 7950x 3d CPU computing that is more efficient, it's the cache. So? The end result is the same. But nevermind premiere pro as well. Computerbase has the 13900k as the most efficient CPU at 45w. How does that happen?

I m taking of power comsumption in multicore, click on +142 Eintrage on the graph and 142 other bar charts will be added to this graph with the 7950X and 13900K measured at differents power levels.

 

[CopernicusLazarus]

For every single watt at a same troughput for that watt.

English is not my first language but even if it was I would still not understand what you just said. What does that even mean? It's an inherently incoherent sentence.

 

[CopernicusLazarus]

I m taking of power comsumption in multicore, click on +142 Eintrage on the graph and 142 other bar charts will be added to this graph with the 7950X and 13900K measured at differents power levels.

Yes, I've seen those results from computerbase, they have done the tests 15 times by now. In all of those graphs they have the 13900k as the most efficient CPU though. Every single one. At 45w it tops the performance chart.

 

[Glo.]

And if you are talking about gaming it's not the 7950x 3d CPU computing that is more efficient, it's the cache. So? The end result is the same. But nevermind premiere pro as well. Computerbase has the 13900k as the most efficient CPU at 45w. How does that happen?

And what about 65W TDP, 88W TDP, 142W TDP, and the fact that 7950X achieves higher performance in 230W than 13900K in 253W TDP? Your table, itself proves wrong your agenda. You are the one who is spreading misinformation, based on one, extreme case scenario.

Are we going to judge efficiency by one extreme case scenario, or averages, that will appear in real world?

 

[CopernicusLazarus]

And what about 65W TDP, 88W TDP, 142W TDP, and the fact that 7950X achieves higher performance in 230W than 13900K in 253W TDP? Your table, itself proves wrong your agenda. You are the one who is spreading misinformation, based on one, extreme case scenario.

Are we going to judge efficiency by one extreme case scenario, or averages, that will appear in real world?

I've already said in my very first post that in higher power levels the 7950x edges ahead. But when we are talking about absolute efficiency first and foremost, nothing beats the 13900k. That is a fact, isn't it?