Question Alder Lake - Official Thread

[Hulk]

With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.

 

[TheELF]

That's an interesting notion. I would imagine that Intel has been embarrassed enough times by AMD to want to remove this thorn from its side completely. Are you saying that Intel is charitable enough to let the people at AMD keep their jobs by refraining from doing everything they can to destroy them, out of sheer goodness of heart?

Intel is forced by the FTCs of the world, but also they don't have any personal reason to want to destroy amd.
Before ryzen intel was making a lot of money and after ryzen they are making even more money so either way intel is making all the moneys so why would they care?

Intel had 6core with htt by the end of 2011 at $600 in the hedt market(i7-3930K) ,5 years before amd released the 8 core 1800x at $500...if they wanted to destroy amd they would have released that as a desktop part, even a few years later and still at the same price it would have caused amd sales back then to drop to zero.
intel also released the 7280x at the same time the 1800x came out, with the same amount of cores and at $100 more but also faster than the 1800x , if intel dropped that to the consumer platform ryzen would have been dead right there and then.

You can discard all that as intel being greedy but that comes out to the same thing in the end, intel prefers to make money instead of trying to kill amd.

 

[igor_kavinski]

You can discard all that as intel being greedy but that comes out to the same thing in the end, intel prefers to make money instead of trying to kill amd.

Good point, though it's a complacent strategy that led them to the desperation of releasing the power hogging 12900K, just to win some benchmarks. I hope Zen3D is not just Zen 3 with V-cache. I like Intel in desperate mode. Forces them to move their performance unlocking secrets from the research phase to production phase much quicker.

 

[deasd]

People keep crying for Intel to add more Gracemont. Until they fix the voltage rail problem it's just gonna bleed power.

YEP I don't know why people want Ecore inside, maybe they have lots of tasks in background?
Voltage rail is another story, but to me Ecore is quite underperforming when comes to realworld usage. (12600k = 5800x in most scene and 4 Ecore = 2C4T Zen3 is well below my expectation)
And just for these perf it screw up the thread scheduling which is a huge cost...

 

[Heartbreaker]

YEP I don't know why people want Ecore inside, maybe they have lots of tasks in background?

E-cores are both more power efficient, and more space efficient. So they will be the preferred way to expand MT performance.

The Knee-jerk, "Just give me 16 P-cores" response is well and good until you need to raise the prices 80% more, to make the same profit margin.

People on forums tend to just think what they want, instead of putting themselves in the shoes of a profit motivated business, that has to consider all the trade-offs.

 

[coercitiv]

People on forums tend to just think what they want

That's how it works in a free-market economy, the first company that manages to offer customers what they want wins the sale. People on forums tend to get that.

The 16-core Zen was always a niche proposition, always aimed at a very specific prosumer willing to pay for the silicon as long as it delivers performance, and the only reason the 12900K was sent on this weird mission to match it in productivity workloads is because Intel lacks a HEDT platform. If they had a Golden Cove based HEDT platform in place they would not need to push ADL-S beyond it's limits, and both of their consumer platforms would be working as intended in terms of efficiency, power delivery and scope.

Let's play game though, let's assume E-cores are absolutely the best choice for MT performance on the mainstream desktop. If that is the case, why is Intel selling E-cores only on the K SKUs? Why are they making 6+0 chips for the value segment instead of 4+8? How come they chose not to cash in on this critical competitive advantage when it came to most of their i5 lineup?

It's almosty as if E-cores also come with some significant downside, a trade-off that Intel chose to avoid for their desktop value products but fully embraced on their entire mobile platform. Mind boggling, isn't it?

 

[TheELF]

Good point, though it's a complacent strategy that led them to the desperation of releasing the power hogging 12900K, just to win some benchmarks. I hope Zen3D is not just Zen 3 with V-cache. I like Intel in desperate mode. Forces them to move their performance unlocking secrets from the research phase to production phase much quicker.

The i9-xx900 is the counterpart to the r9-x900, the 12900kf is even priced exactly as the 5900x barring any discounts.

If you run both at suggested TDP, meaning 125w for intel and ppt of 142w for ryzen, the 12900k uses less power than the 5900x ,it has lower temps than the 5900x because the big die helps a lot in transferring the heat to the cooler (less heat per surface) ,and in both efficiency as well as performance it is above the 5900x.
Even at 125w it is close enough to the 5950x especially if you consider how much cheaper it is.

Intel doesn't even want to counter the 5950x because they would have to spend twice the resources giving them half the number of cpus to sell if they would go with the 16 pcores model, but even every ecore cluster is a ecore cluster they wouldn't be able to use in another cpu, amd is the desperate one here because they are forced to spend that many cores to make sales, with vcache they will be forced to use that many cores and then so much cache on top of that just to keep making sales.

https://www.hardwareluxx.de/index.p...-desktop-cpus-alder-lake-im-test.html?start=8

 

[eek2121]

Huh?! Are you still talking about desktop here? Because the 12900k and all other alder lakes are one single CPU, they don't have cpu clusters, and if you call the quad-ecore-packages clusters, then the 12900k has 2 of those clusters plus the normal cores for a total of 3 clusters.

Intel doesn't want to put more cores into their cpus right now because they are going to make all the moneys anyway and they need to keep enough room for expectations for the next gen.

Also they don't want to destroy amd, if they wanted that they wouldn't have held back for the last 10 years.

Intel uses a ring bus design for most of their desktop/mobile parts (Some Xeons and certain HEDT parts use a mesh type design) .

Each “cluster” of 4 “E” cores is a ring stop vs. a single “P” core. Intel, historically hasn’t been able to scale above 10 “P” cores with their ring bus design (probably because of bandwidth and latency, also power. I am not a modern CPU engineer here, so folks please feel free to chime in with additional thoughts. IIRC Intel utilizes 2 rings for their current designs), so therefore, they must operate within the confines of that design.

Intel does have mesh type designs, but a pure mesh design has many drawbacks.

Note that AMD also uses a ring bus based approach for the cores in their chiplets, that is why you will likely never see a CCX with more than 8 cores.

One final interesting note: Intel did have one design with 12 cores, but presumably they found issues with that design, as they moved to a mesh after for high core count parts.

Note that AnandTech has a very good article which covers this. If I had the page handy I would link it here.

 

[Heartbreaker]

That's how it works in a free-market economy, the first company that manages to offer customers what they want wins the sale. People on forums tend to get that.

What some random guy on a forum wants, is not the market though, and they likely aren't considering the trade-offs that make their wants sub-optimal. On forums you typically just get a bunch of individuals asking for their personal version of this:

While continuing to express their bafflement that the companies, keep missing this obvious great idea.

Let's play game though, let's assume E-cores are absolutely the best choice for MT performance on the mainstream desktop. If that is the case, why is Intel selling E-cores only on the K SKUs? Why are they making 6+0 chips for the value segment instead of 4+8? How come they chose not to cash in on this critical competitive advantage when it came to most of their i5 lineup?

There's no mystery. MT performance isn't the only consideration. If it was, there wouldn't be any P-Cores at all. That may be the case for some server chips, that will be all E-cores, since they are typically only concerned with MT performance.

But for the desktop, you are balancing multiple types of workloads, and you need to determine an optimal number of P-Cores for low thread applications. Intel will have run tests and simulations to determine the reasonable minimum number of performance cores they can have before it starts hurting more thread limited applications.

Given that they are going with 6 P-Cores even in their main mobile chip, then it seems obvious they determined this is an important minimum (unless severely power constrained).

So given that even on Mobile 6 P-cores is the minimum, and the 12600K has 6 P and 4 E cores, the next reasonable step below that is 6+0.

 

[jpiniero]

Intel will have run tests and simulations to determine the reasonable minimum number of performance cores they can have before it starts hurting more thread limited applications.

Given that they are going with 6 P-Cores even in their main mobile chip, then it seems obvious they determined this is an important minimum (unless severely power constrained).

So given the 12600K has 6 P and 4 E cores, the next reasonable step below that is to reduce the E cores.

It's more marketing driven than performance driven.

- Intel didn't want to regress the number of Big Cores on i5.
- Also based on the rumors, Intel was considering at one point to use Rocket Lake Refresh as well or instead of the Alder Lake smaller die. Presumably the Rocket Refresh i5 models would continue to have 6 (Big) cores so you can't have Alder Lake have less than that.
- Intel will release Xeons that would be strange to have small cores enabled on and they do go down the entire stack. AVX-512 is more important to those customers I think .

 

[AdamK47]

The thing is, the power level is fully under the user's control. If you don't like the 12900k power level, either set the tau to a short period or use a 125 W cooler. Either way and the chip will perform almost entirely at 125 W or below. Sure, you won't turbo all the time, but like you said it isn't horrible on efficiency then. The people running it over 125 W are doing so to push peak performance at the cost of efficiency.

That is the problem with almost all reviews. They put on the best cooling possible, push it to the max power, to get the best possible scores. Sure, that is fine for headline numbers. But that isn't how many users actually use it (once it hits OEMs) or want to use it (for people like you that need power efficiency).

That's the thing. It's a choice. Intel released a CPU that allowed me that choice.

Besides, I'm not using my 12900K to run Prime95 / Cinebench loads all day. That's only for stability testing. My regular usage is not going to constantly consume mass amounts of power and heat a room like a furnace. That gets blown out of proportion from those with different CPU brand alignments.

 

[Markfw]

That's the thing. It's a choice. Intel released a CPU that allowed me that choice.

Besides, I'm not using my 12900K to run Prime95 / Cinebench loads all day. That's only for stability testing. My regular usage is not going to constantly consume mass amounts of power and heat a room like a furnace. That gets blown out of proportion from those with different CPU brand alignments.

While that is true, you should not belittle me for my choice, since I DO run 100% 24/7/365. Your last reply was very condescending.

 

[Heartbreaker]

It's more marketing driven than performance driven.

Do you think 4+8 would beat 6+0 for all use cases?

 

[dullard]

While that is true, you should not belittle me for my choice, since I DO run 100% 24/7/365.

But, Alder Lake doesn't run full turbo 241 W 100% 24/7/365.

That is, unless you want it to look bad with power efficiency by setting it to max power with your motherboard settings and cooler selection.

I simply want people who run things 100% 24/7/365 to look at realistic long-term usage power levels, not peak power levels in their calculations.

 

[jpiniero]

Do you think 4+8 would beat 6+0 for all use cases?

No. I mentioned AVX-512 for instance.

While that is true, you should not belittle me for my choice, since I DO run 100% 24/7/365. Your last reply was very condescending.

It's not like you are forced to run it at the default specs.

 

[Markfw]

No. I mentioned AVX-512 for instance.



It's not like you are forced to run it at the default specs.

Question, aside from EPYC, what is the most powerful and efficient CPU for 100% 24/7/365 ? The 5950x, yes ? Thats why I run it, and don't want an Alder lake.

 

[dullard]

Question, aside from EPYC, what is the most powerful and efficient CPU for 100% 24/7/365 ? The 5950x, yes ? Thats why I run it, and don't want an Alder lake.

You are probably correct, but what data are you basing that on? Since I think you use Folding at Home, what Folding at Home Alder Lake data are you seeing? What is that data like at full turbo vs at 125 W?

 

[Zucker2k]

Also they don't want to destroy amd, if they wanted that they wouldn't have held back for the last 10 years.

That wasn't entirely at Intel's discretion.

 

[Markfw]

You are probably correct, but what data are you basing that on? Since I think you use Folding at Home, what Folding at Home Alder Lake data are you seeing? What is that data like at full turbo vs at 125 W?

I do F@H on GPU, Rosetta@home on CPU and WCG on CPU. All I know is that the 12900k only beats the 5950x in most benchmarks@241 watts. At 125w it does not do near as well, and I only use 142 watts all day, every day.

 

[Zucker2k]

Intel is forced by the FTCs of the world, but also they don't have any personal reason to want to destroy amd.
Before ryzen intel was making a lot of money and after ryzen they are making even more money so either way intel is making all the moneys so why would they care?

Intel had 6core with htt by the end of 2011 at $600 in the hedt market(i7-3930K) ,5 years before amd released the 8 core 1800x at $500...if they wanted to destroy amd they would have released that as a desktop part, even a few years later and still at the same price it would have caused amd sales back then to drop to zero.
intel also released the 7280x at the same time the 1800x came out, with the same amount of cores and at $100 more but also faster than the 1800x , if intel dropped that to the consumer platform ryzen would have been dead right there and then.

You can discard all that as intel being greedy but that comes out to the same thing in the end, intel prefers to make money instead of trying to kill amd.

Killing AMD = MOAR MONEY!

 

[Zucker2k]

I hope Zen3D is not just Zen 3 with V-cache. I like Intel in desperate mode. Forces them to move their performance unlocking secrets from the research phase to production phase much quicker.

The thing about Zen3D that a lot of people are overlooking is that AMD did those tests at 4GHz. Anyone care to bet that's the point of optimal returns in games? With Alder Lake clocking P cores to the high GHz (4.7GHz+) in games, I don't think this is any longer about extrapolating data from percentages by which Alder Lake trumps Zen 3 in gaming.

 

[Zucker2k]

Let's play game though, let's assume E-cores are absolutely the best choice for MT performance on the mainstream desktop. If that is the case, why is Intel selling E-cores only on the K SKUs? Why are they making 6+0 chips for the value segment instead of 4+8? How come they chose not to cash in on this critical competitive advantage when it came to most of their i5 lineup?

I think the answer is simple. A complement of E-cores are needed when you push the P-cores. When you're not pushing the P-cores, you don't need it. The E-cores also help in keeping the mid to upper tier chips more competitive in multi-threaded scenarios. See how 6+4 overshadows the 5800x in multithreaded workloads even though the P-cores are not being pushed as hard. So the e-cores has its uses, it's simply not for every segment because Intel strategically positioned the chips/performance according to price, or the other way around.

 

[Zucker2k]

It's more marketing driven than performance driven.

Sure. In a world where Zen 3 doesn't exist.

 

[Heartbreaker]

No. I mentioned AVX-512 for instance.

That's a cop out, what about non AVX-512 loads?

 

[dullard]

I do F@H on GPU, Rosetta@home on CPU and WCG on CPU. All I know is that the 12900k only beats the 5950x in most benchmarks@241 watts. At 125w it does not do near as well, and I only use 142 watts all day, every day.

So, you still are basing all of your decisions and posts on assumptions, not Rosetta@home or WCG benchmarks? Your assumptions may be correct, but if you want to know why people think you have a brand alignment, that is the reason. I just want real data on the actual applications that I use to base my decisions on.

 

[AdamK47]

While that is true, you should not belittle me for my choice, since I DO run 100% 24/7/365. Your last reply was very condescending.

I didn't intend for any direct belittlement. It was more of a general belittlement directed at those that constantly cling to edge cases to satiate their own preferences.