[AT] Intel trying to release 7 nm in 2021

[DrMrLordX]

Yes exactly,they bought them because of the number of threads not because of smaller nm than intel or because of more IPC than intel.

AMD now has the opportunity to launch Zen2 dice on AM4 with up to twice the number of cores as Intel's best desktop CPU thanks to "smaller nm". That is why you need process improvements to stay competitive. That is why 14nm isn't "good enough" anymore. Even a 12c Zen2 chip will be trouble for the 9900k, and Comet Lake after it.

 

[PotatoWithEarsOnSide]

Is it still the case that you cannot run a 9900K at 95w unless you own a Maximus Hero XI? And even then you have to consciously switch off the default MCE...
The 9900K certainly is the most power efficient CPU within it's TDP limit, but its nigh on impossible to get one to run to its listed spec...not that anyone would ever recommend that you tried to.

 

[JDG1980]

Yes and wonder of wonders this post is about intel announcing that they will bring 10nm to servers first...

Yes. The question is, how soon can they do this, and how competitive will these 10nm products be with the 7nm server products that AMD will already have had out for a year or more ahead of time? There's a good chance that Intel comes in a day late and a dollar short. Intel's first-generation 10nm server chips might end up having to compete with Zen *3*, and that's going to be a tough fight to win.

Desktop is the moneymaker for intel,they make more from PC centric than from everything else combined.
https://www.intc.com/investor-relat...t-Quarter-2019-Financial-Results/default.aspx

Re-read the chart you posted. It says "PC-centric", not desktops. Almost certainly a majority of that volume consists of laptop chips. And on laptops, performance per watt is *much* more important than whether a chip maxes out at 4 GHz or 5 GHz (except in perhaps a handful of big desktop-replacement notebooks). Notebook manufacturers want to know who can provide the best performance at a target TDP - often 15W or less. Intel is going to have a very hard time beating AMD on this metric once AMD gets their laptop chips onto 7nm and Intel is stuck at 14nm. And even when Intel manages to ship something on a denser node, I wouldn't be so sure that a low-yielding, early-generation 10nm process is going to be superior to mature TSMC 7nm, either.

The fact is that once AMD gets onto 7nm, the *only* place Intel has a decent chance of being competitive while they're still stuck on 14nm is on gaming desktops, where they might still maintain a marginal edge in single-thread performance and/or memory latency. But even though this is what a lot of posters here care most about, it's really a small sideshow when it comes to the PC market as a whole. Make no mistake about it, the failure to execute on 10nm has been a tactical and strategic *disaster* for Intel, regardless of what happy face they're putting on for investors.

 

[TheELF]

Re-read the chart you posted. It says "PC-centric", not desktops. Almost certainly a majority of that volume consists of laptop chips. And on laptops, performance per watt is *much* more important than whether a chip maxes out at 4 GHz or 5 GHz (except in perhaps a handful of big desktop-replacement notebooks). Notebook manufacturers want to know who can provide the best performance at a target TDP - often 15W or less. Intel is going to have a very hard time beating AMD on this metric once AMD gets their laptop chips onto 7nm and Intel is stuck at 14nm. And even when Intel manages to ship something on a denser node, I wouldn't be so sure that a low-yielding, early-generation 10nm process is going to be superior to mature TSMC 7nm, either.

9 out of 10 laptop users will prefer having quick sync rather than a slightly faster or slightly more power efficient chip.
Now if AMD can get a rather big difference this will probably change but for a small difference people will prefer the chip with the most support and the most added value,like qsv.

 

[TheELF]

AMD now has the opportunity to launch Zen2 dice on AM4 with up to twice the number of cores as Intel's best desktop CPU thanks to "smaller nm". That is why you need process improvements to stay competitive. That is why 14nm isn't "good enough" anymore. Even a 12c Zen2 chip will be trouble for the 9900k, and Comet Lake after it.

How do you think that will look like?You think they will release a 16/32 CPU that will run even at the (not even, at all core) 4.2 they run at now?
Do you think AMD would price such a CPU at 9900k levels?
Not many people will care much about a chip that needs 50% more cores to beat a 9900k,unless it's for a very specific purpose.

 

[dmens]

Yes and wonder of wonders this post is about intel announcing that they will bring 10nm to servers first...

Pro-tip: they won't.

 

[DrMrLordX]

9 out of 10 laptop users will prefer having quick sync rather than a slightly faster or slightly more power efficient chip.

. . . what? Where did you conduct this survey?

How do you think that will look like?You think they will release a 16/32 CPU that will run even at the (not even, at all core) 4.2 they run at now?

Clocks will be higher than the 2950X, in all likelihood. And clocks for that chip are pretty good. My guess is that all-core will be somewhere around 3.7-3.8 GHz for the 16c chip. Unless you really load up those threads, it'll boost higher than that.

Do you think AMD would price such a CPU at 9900k levels?

Sure, why not? They sold the 1800x for $499 in 2017.

Not many people will care much about a chip that needs 50% more cores to beat a 9900k,unless it's for a very specific purpose.

. . . heh. Okay pal.

 

[extide]

Is it still the case that you cannot run a 9900K at 95w unless you own a Maximus Hero XI? And even then you have to consciously switch off the default MCE...
The 9900K certainly is the most power efficient CPU within it's TDP limit, but its nigh on impossible to get one to run to its listed spec...not that anyone would ever recommend that you tried to.

You should be able to do it on most motherboards right?

In any case the 9900K using >>95W is not really Intel's fault, it's the mobo makers fault. The fact is the stock settings of the CPU actually do run it at 95w but there is quite a bit more performance left on the table when doing so, and many mobos just enable it 'for free'. If that was possible with AMD chips then they would be doing it for them too. I mean the performance between Intel & AMD when limited to ~100w is very similar in reality. The difference is there is not much more you can squeeze out of the AMD chip which is not the case for the Intel chip.

 

[IntelUser2000]

Re-read the chart you posted. It says "PC-centric", not desktops. Almost certainly a majority of that volume consists of laptop chips.

Yup. Not only laptop chips are far greater volume, they also make more money per device. Probably 60% of volume and 70% revenue. It crossed the 50% mark nearly 2 decades ago.

 

[DrMrLordX]

In any case the 9900K using >>95W is not really Intel's fault, it's the mobo makers fault. The fact is the stock settings of the CPU actually do run it at 95w but there is quite a bit more performance left on the table when doing so, and many mobos just enable it 'for free'.

That's not 100% true. Intel has provided base, PL1, and PL2 modes for the chip. Almost no motherboard manufacturers run a 9900k within the strict 95W TDP limit by default (it is user-selectable in the UEFI). Most of them ship with the PL1 power limit enabled.. A few motherboard manufacturers opted for PL2 out-of-the-box which is why we saw those 210W+ reviews at 9900k launch.

Intel is at least complicit in creating the PL spec rather than forcing motherboard manufacturers to adhere to a 95W power limit. The "real" TDP of the 9900k is around 160W for most users. That is by design, per Intel's intentions.

 

[jpiniero]

I don't know about changing PL1... but you should be able to configure PL2 to stay enabled indefinitely which kind of has the same effect.

 

[coercitiv]

That's not 100% true. Intel has provided base, PL1, and PL2 modes for the chip. Almost no motherboard manufacturers run a 9900k within the strict 95W TDP limit by default (it is user-selectable in the UEFI). Most of them ship with the PL1 power limit enabled.. A few motherboard manufacturers opted for PL2 out-of-the-box which is why we saw those 210W+ reviews at 9900k launch.

There's nothing wrong with the default power limits of 9900K, and when I say "default" I'm referring to PL1 and PL2 as defined by Intel engineers in the 9th gen CPU spec sheets. There's everything wrong with how motherboard makers implemented power limits in 8th and 9th gen boards, in the sense that high end boards come with very high power limits or simply no power limits at all. It is very likely that Intel said one thing in private conversations and completely another in publicly disseminated spec sheets, and a partial proof of that is the weird PL2 values that were given to Anandtech by Intel PR when 9900K reviews were in and numbers were all over the place.



As you can see from the table above, the default values for 95W TDP 8 core CPUs are PL1 = 95W and PL2 ~= 120W.

I don't know about changing PL1... but you should be able to configure PL2 to stay enabled indefinitely which kind of has the same effect.

Both should and are user configurable.

 

[DrMrLordX]

As you can see from the table above, the default values for 95W TDP 8 core CPUs are PL1 = 95W and PL2 ~= 120W.

Weird, almost every Z390 board sets PL1 = 160W. I guess the only thing you can blame Intel for is not forcing mobo OEMs to follow their specs more closely. User-configurable also means OEM-configurable.

 

[coercitiv]

I guess the only thing you can blame Intel for is not forcing mobo OEMs to follow their specs more closely.

Plausible deniability at its best 😎

 

[extide]

Weird, almost every Z390 board sets PL1 = 160W. I guess the only thing you can blame Intel for is not forcing mobo OEMs to follow their specs more closely. User-configurable also means OEM-configurable.

Really, you are going to toe that line right now? I mean honestly if it was locked there would be people crying about that. It's definitely better that it is configurable. In pretty much every case enabling higher power use gets the user more performance for basically no downside. In the cases where it actually matters ... the options are there. You can bet that OEM's like Dell and such who are selling these chips will use the spec'd Intel numbers. I know Supermicro sticks to the limits exactly. It's honestly a non-issue.

 

[Topweasel]

Really, you are going to toe that line right now? I mean honestly if it was locked there would be people crying about that. It's definitely better that it is configurable. In pretty much every case enabling higher power use gets the user more performance for basically no downside. In the cases where it actually matters ... the options are there. You can bet that OEM's like Dell and such who are selling these chips will use the spec'd Intel numbers. I know Supermicro sticks to the limits exactly. It's honestly a non-issue.

It's not about being locked. Even Anandtech covered during their review that Intel was specifying to mobo manufacturers and reviewers to leave the PL2 set to 210w. This also isn't the case of one board manufacturer setting to 210w and the rest quickly following. They all launched at that. It's also not the default PL2 for all CPU's only the 9700k and 9900k. I think there was a big early push to put this on the Mobo manufacturers but if the Official PL2 setting for the 9900k isn't the 210w. Its pretty apparent the unofficial setting from Intel is.

No one is saying the mobo guys should lock the setting or Intel. But there is a huge difference between it being available as a setting and it being requested to be the default when that CPU is installed.

 

[Zucker2k]

It's not about being locked. Even Anandtech covered during their review that Intel was specifying to mobo manufacturers and reviewers to leave the PL2 set to 210w. This also isn't the case of one board manufacturer setting to 210w and the rest quickly following. They all launched at that. It's also not the default PL2 for all CPU's only the 9700k and 9900k. I think there was a big early push to put this on the Mobo manufacturers but if the Official PL2 setting for the 9900k isn't the 210w. Its pretty apparent the unofficial setting from Intel is.

No one is saying the mobo guys should lock the setting or Intel. But there is a huge difference between it being available as a setting and it being requested to be the default when that CPU is installed.

That's not the case, iirc. The reason why this came to the fore was because of the significant discrepancy in some of the results between reputable sites. Also, on the 210w PL2 recommendation from Intel to reviewers, it's clear Intel came up with that figure because the 2700x consumes that much, stock (you can check this out from the Techspot redo of the 9900k @ 95w, among others).

 

[Topweasel]

That's not the case, iirc. The reason why this came to the fore was because of the significant discrepancy in some of the results between reputable sites. Also, on the 210w PL2 recommendation from Intel to reviewers, it's clear Intel came up with that figure because the 2700x consumes that much, stock (you can check this out from the Techspot redo of the 9900k @ 95w, among others).

It was because of cooling choices not because of different settings to motherboard, it doesn't matter why Intel came up with it. Now I guess in theory for a moment or two I could see the 2700x using up that much power but I have never heard of a stock setting 2700x using that much power for anything sustained.

Edit: Found the article and it has nothing about the 210w. The power usage there is the system power and it uses between 10-25W more power then a hardlocked 95w CPU system wise, knowing the 370x/470x is less efficient. Ehich is a much more reasonable take on the idea of cooling allow for limited and controlled increases in clockspeed. Intel choose 210w because it basically allowed the CPU to run at max speed all of the time not because 210 is some super number that makes it beat AMD. In actually it caps out at about 140w-180w depending on silicon (which is another reason to hate this practice). The worst is that this practice makes cooler choosing that much more complicated if you don't know better.

 

[Markfw]

That's not the case, iirc. The reason why this came to the fore was because of the significant discrepancy in some of the results between reputable sites. Also, on the 210w PL2 recommendation from Intel to reviewers, it's clear Intel came up with that figure because the 2700x consumes that much, stock (you can check this out from the Techspot redo of the 9900k @ 95w, among others).

WRONG: I just looked it up. the 95 watt version takes 181 watts, and the unlimited takes 249 watts, and the 2700x takes 205 in that review, so its probably total system power, but it takes way less than the unlimited 9900k.

HERE: https://www.techspot.com/review/1744-core-i9-9900k-round-two/

Edit: that was the blender gooseberry, the other power graph was about 5 watts lower, but the three were about the same as the other one.

And why is it that every time you post about AMD its tainted or wrong ?

 

[DrMrLordX]

it's clear Intel came up with that figure because the 2700x consumes that much, stock (you can check this out from the Techspot redo of the 9900k @ 95w, among others).

Uh.

https://www.tomshardware.com/reviews/amd-ryzen-7-2700x-review,5571-12.html

I found basically the same result from AnandTech:

https://www.anandtech.com/show/12625/amd-second-generation-ryzen-7-2700x-2700-ryzen-5-2600x-2600/8

I have heard that 2700x chips can hit maybe 140W sometimes if you cool them pretty well. But 210W? Not even close.

Interestingly enough, the initial AnandTech review of the 9900k showed the 9900k hitting 210W on a Z370 board thanks to that board defaulting to PL2. The re-review on a Z390 board showed it configured by default to hit a PL1 of ~160W. The re-review:

https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/21

 

[LikeLinus]

Honestly, this is a stupid argument and you're just running in circles. All you guys are doing is picking on the 9900K because you can. Guess what? The people buying that chip really don't give a flying funk about how much wattage is consumed. It's really meant for being over-the-top gaming chip to please those what want the fastest at all cost. It's like saying "OMG my Ferrari gets 12 miles to gets 12 miles to the gallon, but the Mustang gets 20!!!!!". The 9900K serves it's purpose. No one is picking up this chip to browse the web or use Office, lol.e

Not to mention, if you're getting a 9900k CPU, are you also getting the 1030 GPU or lower from because the 1080 TI uses too much power? Are you using the fastest ram possible? Keep things in perspective and think realistically people.

 

[Zucker2k]

It was because of cooling choices not because of different settings to motherboard, it doesn't matter why Intel came up with it. Now I guess in theory for a moment or two I could see the 2700x using up that much power but I have never heard of a stock setting 2700x using that much power for anything sustained.

Edit: Found the article and it has nothing about the 210w. The power usage there is the system power and it uses between 10-25W more power then a hardlocked 95w CPU system wise, (1)knowing the 370x/470x is less efficient. Ehich is a much more reasonable take on the idea of cooling allow for limited and controlled increases in clockspeed. (2)Intel choose 210w because it basically allowed the CPU to run at max speed all of the time not because 210 is some super number that makes it beat AMD. In actually it caps out at about 140w-180w depending on silicon (which is another reason to hate this practice). The worst is that this practice makes cooler choosing that much more complicated if you don't know better.

1. Please. So the bells and whistles clad $600 MSI MEG Z390 Godlike flagship motherboard is the more energy-efficient motherboard?



2. Except that PL2 is Turbo and should not be sustainable indefinitely. PL2=Unlimited is not an Intel design. The motherboard makers actually did more harm than good in this regard because the 9900k was hitting high voltages under avx load and consuming more power to sustain the multi-core-enhanced turbos in a vicious cycle of power-thermal impasse.

The 210w PL2 recommendation is just a ceiling, and just like the 2700x, the 9900k doesn't get anywhere near that number running non-avx code. Techspot used to publish idle numbers but in this review and beyond, that insightful variable vanished from their charts. The delta between idle and load should be telling for both chips.

WRONG: I just looked it up. the 95 watt version takes 181 watts, and the unlimited takes 249 watts, and the 2700x takes 205 in that review, so its probably total system power, but it takes way less than the unlimited 9900k.

HERE: https://www.techspot.com/review/1744-core-i9-9900k-round-two/

Edit: that was the blender gooseberry, the other power graph was about 5 watts lower, but the three were about the same as the other one.

And why is it that every time you post about AMD its tainted or wrong ?

I meant system. You have been wrong also. Case in point, the power consumption number in the Handbrake test is actually higher; it's 209w. If you subtract idle, (fans, ssd, etc) you should be ballpark 160w - 180w. Meanwhile, overclocked to a paltry 4.2GHz, the picture is worse.



So the toasty 210w 9900k monster system draws 255w @ 4.7GHz all core turbo, while the 105w 2700x Cinderella system draws 209w stock, and 264w overclocked to only 4.2GHz.

 

[Markfw]

You are twisting the numbers again. stock vs stock, the 2700x is between the TWO different stock settings for the 9900k. And the 105w CPU in the full system uses 209, not just the cpu, just like in your example the 9900k uses 176 or 255.

If I turn my vcore down, I can get it to run way less wattagre, the same as the 990k@95 watt settings. So against your own examples, and the two of DrMrlordx, its all proves you WRONG.

 

[coercitiv]

Also, on the 210w PL2 recommendation from Intel to reviewers, it's clear Intel came up with that figure because the 2700x consumes that much, stock

I meant system.

The 210w PL2 recommendation is just a ceiling

So the 210W recommendation for the 9900K came from Intel based on the 210W power consumption of an entire Ryzen system? Fascinating, are you making this up as you go along?

So the toasty 210w 9900k monster system draws 255w @ 4.7GHz all core turbo, while the 105w 2700x Cinderella system draws 209w stock

Nah, let's go back to how Intel calculated PL1 and PL2 for their 9th gen premium desktop parts, you're not getting out of this one by pointing fingers in the distance.

 

[PotatoWithEarsOnSide]

It isn't the PL2 being set to 210w that is the problem.
The problem is the Tau value being set to the maximum instead of the 24-28s as per Intel spec; Turbo is not meant to be indefinite, rather for a defined maximum period.