[AT] Intel trying to release 7 nm in 2021

[beginner99]

Yes because Intel 7nm should still be competitive with TSMC 3nm which currently has no timeline for release, and Samsung “3nm” that was just announced is for 2021 is less dense than TSMCs 5nm but has a cool new GAA device architecture, also Sapphire Rapids is 10++, 2022 for 7nm Server CPUs (Granite Rapids) as per Navin

Reading this I just remembered that foundries had at least some issues/delays with finfets. Maybe, just maybe intel made some important learnings during their 10nm crisis which the foundries yet have to do? This is pure speculation but could mean intel could catch up again.
Also AMD should be focused and continue to deliver. I mean they know zen2 should actually be competitive to icelake at the minimum. With all the skylakes and 10nm issues Intel will at some point have a new uarch in production that will actually add usable IPC besides clocks. In late 2021 that could mean a huge jump in CPU performance for intel (7nm EUV + jumping several years of uarch improvements). could easily be >100% performance/watt.

 

[TheELF]

Reading this I just remembered that foundries had at least some issues/delays with finfets. Maybe, just maybe intel made some important learnings during their 10nm crisis which the foundries yet have to do? This is pure speculation but could mean intel could catch up again.

The problem is that we have no idea what the problem with 10nm is right now.
It could just be that they can't reach 4.5Ghz + right now so they can't release a gen that will match the previous one.
Can any other fab reach 4.5Ghz + right now on low nm on such a complicated arch?
Intel could very well still be leading in everything but not releasing 10nm because it doesn't make economic sense for them to do so,14nm still sells like crazy so they are not forced to do anything,they don't need the new gimmick of 10nm yet.

 

[DrMrLordX]

Intel is making the most advanced CPU with the highest performance right now

Apple is making the most advanced CPU right now.

how does that equate into having to get back into the fab game?

10nm is still on the ropes. Intel looks to be abandoning it in favor of 7nm EUV. If 7nm EUV has the same trouble as 10nm, then they are pretty much dead in the water at that point. TSMC wins.

Also what fab game?
Intel has their own fabs to make their own chips they don't need to have any fab game,they are their own customers,they just build a new fab because they needed more production.

Intel can't keep pimping 14nm forever. They have two cores on the sidelines right now (Cannonlake, Sunny Cove) and one coming up (Willow Cove) that they can't put into mass production because of 10nm failures. We are only going to see a trickle of IceLake products, with max core counts of 4c. Rocket Lake may bring us Sunny Cove or Willow Cove on 14nm, but that's mobile only for right now, and it's questionable as to whether that will be a good idea in the long run. Intel is struggling to release new products. Without node advancements, their CPU designs will have to go to other foundries.

Yes because Intel 7nm should still be competitive with TSMC 3nm which currently has no timeline for release, and Samsung “3nm” that was just announced is for 2021 is less dense than TSMCs 5nm but has a cool new GAA device architecture, also Sapphire Rapids is 10++, 2022 for 7nm Server CPUs (Granite Rapids) as per Navin

I am skeptical as to whether anything 10nm will be released in significant volume. If Intel can't launch 7nm in 2021 or 2022, then it won't be competitive with anything.

 

[Dayman1225]

I am skeptical as to whether anything 10nm will be released in significant volume. If Intel can't launch 7nm in 2021 or 2022, then it won't be competitive with anything.

The waiting game truly is a terrible one

 

[jpiniero]

The problem is that we have no idea what the problem with 10nm is right now.

It's yield. Pretty obvious that Intel had wanted Cannonlake to handle the i3 and lower U parts but since the yield is so bad they had to release the 8130U and friends. Which in turn played a big part in the shortage.

 

[TheELF]

Apple is making the most advanced CPU right now.

Are they?Run windows 10 on their CPU run far cry 5,tell us how well that goes.

10nm is still on the ropes. Intel looks to be abandoning it in favor of 7nm EUV. If 7nm EUV has the same trouble as 10nm, then they are pretty much dead in the water at that point. TSMC wins.

What does TSMC win? Intel's business?Market share?Profit?

Intel can't keep pimping 14nm forever. They have two cores on the sidelines right now (Cannonlake, Sunny Cove) and one coming up (Willow Cove) that they can't put into mass production because of 10nm failures. We are only going to see a trickle of IceLake products, with max core counts of 4c. Rocket Lake may bring us Sunny Cove or Willow Cove on 14nm, but that's mobile only for right now, and it's questionable as to whether that will be a good idea in the long run.

They don't need to pimp it forever,10nm will mature in the background no matter how long it will take,intel will release them on server and laptop first at lower clocks to partially compensate and when it's ready they will release it on desktop.

Intel is struggling to release new products. Without node advancements, their CPU designs will have to go to other foundries.

Tell us one foundry that could produce their CPU designs on a better node... "other foundries" can't even produce the RYZEN 10nm design at more then 4-4.2Ghz intel's design is at least as complicated so how are they going to pull off higher clocks?

 

[TheELF]

It's yield. Pretty obvious that Intel had wanted Cannonlake to handle the i3 and lower U parts but since the yield is so bad they had to release the 8130U and friends. Which in turn played a big part in the shortage.

Yes it is yields,but what kind of yields? If they have bad yield at even 3Ghz than yes this is extremely bad,if they have bad yields at 5Ghz well that's to be expected since it's incredibly difficult,it's still not good but it's not that doom and gloom that everybody tries to make it out to be.

 

[TheGiant]

Are they?Run windows 10 on their CPU run far cry 5,tell us how well that goes.

What does TSMC win? Intel's business?Market share?Profit?


They don't need to pimp it forever,10nm will mature in the background no matter how long it will take,intel will release them on server and laptop first at lower clocks to partially compensate and when it's ready they will release it on desktop.

Tell us one foundry that could produce their CPU designs on a better node... "other foundries" can't even produce the RYZEN 10nm design at more then 4-4.2Ghz intel's design is at least as complicated so how are they going to pull off higher clocks?

from pure results, you are right
it doesn't matter what your node is, your architecture...whatever, what matters is performance,power, price, name and profit ..not not really PROFIT
despite all the haters, Intel manages well now
looking forward it can change

 

[maddie]

The problem is that we have no idea what the problem with 10nm is right now.
It could just be that they can't reach 4.5Ghz + right now so they can't release a gen that will match the previous one.
Can any other fab reach 4.5Ghz + right now on low nm on such a complicated arch?
Intel could very well still be leading in everything but not releasing 10nm because it doesn't make economic sense for them to do so,14nm still sells like crazy so they are not forced to do anything,they don't need the new gimmick of 10nm yet.

If the only problem was clocks the we would have all of the low powered CPUs lines on 10 nm. This pretty much refutes your assumption.

 

[jpiniero]

Yes it is yields,but what kind of yields?

Defects, the chips are literally busted.

 

[DrMrLordX]

Are they?Run windows 10 on their CPU run far cry 5,tell us how well that goes.

Cute. MS isn't going to port Windows to A12x or A13 since Apple won't allow anyone to run that software on their hardware anyway. Win10 and FarCry 5 aren't indicators of what is or isn't advanced. You know exactly why A12x is a headturner (and why A13 promises to be one).

What does TSMC win?

The foundry war. At that point it's TSMC vs Samsung, with Samsung playing second fiddle. Anyone who wants to produce high-performance desktop or server CPUs would likely have to bow to TSMC. Intel's only play would be to license their tech or spin off their fabs to TSMC (or Samsung). 7nm EUV is Intel's last chance.

10nm will mature in the background no matter how long it will take

Don't be so sure. If/when 7nm comes online, Intel is going to be in a race to move all their designs to that node. There is still no 10nm desktop CPU in 2020, and possibly not even in 2021.

Tell us one foundry that could produce their CPU designs on a better node...

TSMC

"other foundries" can't even produce the RYZEN 10nm design at more then 4-4.2Ghz

Ryzen 10nm? What are you talking about? GF 12nm already hits 4.3 GHz turbo. TSMC will exceed that on 7nm.

Intel manages well now
looking forward it can change

They've already reduced guidance. Keep your eye on their investor materials. Should be interesting, to say the least.

 

[jpiniero]

Intel doesn't have to spin off anything, especially given the move to chiplets. Use the existing fabs they have for what they can get away with on older nodes, and then if it comes to it, source from TSMC and/or Samsung for the main parts.

 

[PotatoWithEarsOnSide]

Did TSMC spawn infinite capacity out of their arses overnight?
Not entirely sure how anyone can expect them to be fabbing for Apple, Huawei, AMD (read Sony and Microsoft console volumes) and Intel all at the same time.

 

[sxr7171]

Right, but you still haven't proven the correlation on loss of sales to such a insignificant thing such as chip nm sizing and the general public's perception and how it affects overall sales. Example, take a look at all of the media attention and negativity that Volkswagen received. They were still #2 in overall car sales last year and actually UP 1% year over year in overall global sales. Mercedes was also up 1% YOY in global sales.

https://www.best-selling-cars.com/global/2018-full-year-international-global-top-car-brands/

Another example is Microsoft Vista. OMG, do you know how many people here trashed it and said MS was going down and everyone was going to move to Linux/Apple? They got so much crap, but they still have, what, 90% of the market?

But you guys actually think that the 9900k or difference between 14 and 10/7nm is going to make the mass majority of the market shift their purchase? IF you asked 1000 people out in the public what a nanometer is, how many do you actually think know what the hell it means?

The power of brand recognition, history and marketing is strong. Also, most software developers recommend Intel and so do most PC vendors. I'm not saying anything against AMD, as I currently and have used their chips in the past, but some people are not separating actual customer perception and purchasing history, with their own personal feelings towards the subject.

Are they?Run windows 10 on their CPU run far cry 5,tell us how well that goes.

What does TSMC win? Intel's business?Market share?Profit?


They don't need to pimp it forever,10nm will mature in the background no matter how long it will take,intel will release them on server and laptop first at lower clocks to partially compensate and when it's ready they will release it on desktop.

Tell us one foundry that could produce their CPU designs on a better node... "other foundries" can't even produce the RYZEN 10nm design at more then 4-4.2Ghz intel's design is at least as complicated so how are they going to pull off higher clocks?

Can’t wait for Apple to release ARM based MacBooks. I don’t even like MacBooks.

Then I’d like to see Lenovo release an ARM X1 carbon.

Then hubristic monopolies can go choke on a rod.

That’s kind of dreaming. I really would like some competition in this space. Some choices. It’s getting to a point that even as security matter you can’t bet on one horse.

Intel will probably be back up if and then this happens. But going forward they will have to stay on their toes. Not stagnate like this.

 

[Ajay]

Did TSMC spawn infinite capacity out of their arses overnight?
Not entirely sure how anyone can expect them to be fabbing for Apple, Huawei, AMD (read Sony and Microsoft console volumes) and Intel all at the same time.

^This. I haven’t read anything about TSMC or Samsung building up a couple of high capacity fabs for some unnamed top tier semiconductor designer. It’s just ridiculous rumor mongering at this point.

 

[sxr7171]

from pure results, you are right
it doesn't matter what your node is, your architecture...whatever, what matters is performance,power, price, name and profit ..not not really PROFIT
despite all the haters, Intel manages well now
looking forward it can change

They do. But at some point some corporate user will go to IT guy asking for a new machine.

IT guy will say you can have one now and have the same performance, battery life etc.

Or because Intel screwed up you can wait a year and get some improvement. Maybe.

 

[DrMrLordX]

Intel doesn't have to spin off anything, especially given the move to chiplets. Use the existing fabs they have for what they can get away with on older nodes, and then if it comes to it, source from TSMC and/or Samsung for the main parts.

Anything's possible with EMIB/Foveros. In the long run, they're not going to want to use 14nm for anything, though. 10nm for small, less-critical parts (basically SoC functions).

Did TSMC spawn infinite capacity out of their arses overnight?

If Intel wants TSMC to take over for them, TSMC needs more fab capacity. That means Intel either has to sell them fabs directly, spin them off into a separate unit that can sell off fabs as it chooses, or start cross-licensing tech to fix the problems via cooperation that they can't solve through competition. Intel has the fabs, they just don't have nodes. This all assumes 7nm doesn't work out. If it does then the game is afoot.

 

[TheGiant]

They do. But at some point some corporate user will go to IT guy asking for a new machine.

IT guy will say you can have one now and have the same performance, battery life etc.

Or because Intel screwed up you can wait a year and get some improvement. Maybe.

I am sitting on my cable to the brain, sorry I don't understand

especially the part of same performance and battery life...

 

[JDG1980]

The problem is that we have no idea what the problem with 10nm is right now.
It could just be that they can't reach 4.5Ghz + right now so they can't release a gen that will match the previous one.

If that was the case, then Intel would have already released laptop and server CPUs on 10nm. As I have said before, CPUs with clock rates above 4 GHz are in fact a fairly small niche. The bulk of production consists of laptop and server chips with max speeds well below that. Single-thread performance is not the only metric that matters, and not the most important one.

Can any other fab reach 4.5Ghz + right now on low nm on such a complicated arch?
Intel could very well still be leading in everything but not releasing 10nm because it doesn't make economic sense for them to do so,14nm still sells like crazy so they are not forced to do anything,they don't need the new gimmick of 10nm yet.

Do you think Intel wants to give up the server performance crown to EPYC? That's what is going to happen in the next couple of months. If Intel could put out a viable 10nm server product, they would have. And most server chips don't need to turbo faster than 3.0-3.5 GHz.

 

[sxr7171]

I am sitting on my cable to the brain, sorry I don't understand

especially the part of same performance and battery life...

Well let’s say you currently have a machine with a 7th gen Intel CPU. You bought it 2-3 years ago. Is 8th or 9th really going to improve battery life or performance by any substantial amount?

Rather your IT Dept supplied said 7th gen laptop 2-3 years ago. Would they be willing to upgrade your machine if it is not going to offer any significant upgrade? The 14nm+++ versions are just loosely packed more power hungry versions of 7th gen CPUs. So I don’t know if they really offer much improved performance.

When 10nm laptops come out it seems they will have lower clocks in the beginning. So it probably won’t be until 2020-2021 we see real improvements in performance. I suppose that may come with 10nm+? Or what they will call 10nm++ or whatever new name they give it.

We are already keeping our machines for longer these days. Even enthusiasts who like to upgrade often. I remember a time as an enthusiast when annual upgrades brought real improvements. Now we keep our machines for 3 years on the same CPU. The general public which always kept their machines for 3 years or more might be justified keeping them for longer now.

Of course they don’t really know what makes these machines tick so they just buy based on the actual laptop designs and little improvements like battery size, smaller bezels, keyboard improvements, looks, weight, size etc. But if their IT guy or techie friend says all that is well and good but it won’t really be any faster then they might hold off on upgrades.

The school season purchases I suppose will always be around the same.

On the server side I don’t know much but it seems like they can’t stay ahead there. But those are ordered by IT managers who don’t care about branding. They just care about price/performance. So they will just order AMD.

Still I have to admit Intel branding, name recognition and marketing is second to none. So these technical issues don’t affect them as much we enthusiasts like to claim. Yet still they know and we know that a lot of purchases are made on the recommendation of whichever techie guy they have in their circle. Which is why Intel caters to enthusiasts. They know we help them sell product.

 

[TheGiant]

Well let’s say you currently have a machine with a 7th gen Intel CPU. You bought it 2-3 years ago. Is 8th or 9th really going to improve battery life or performance by any substantial amount?

Rather your IT Dept supplied said 7th gen laptop 2-3 years ago. Would they be willing to upgrade your machine if it is not going to offer any significant upgrade? The 14nm+++ versions are just loosely packed more power hungry versions of 7th gen CPUs. So I don’t know if they really offer much improved performance.

When 10nm laptops come out it seems they will have lower clocks in the beginning. So it probably won’t be until 2020-2021 we see real improvements in performance. I suppose that may come with 10nm+? Or what they will call 10nm++ or whatever new name they give it.

We are already keeping our machines for longer these days. Even enthusiasts who like to upgrade often. I remember a time as an enthusiast when annual upgrades brought real improvements. Now we keep our machines for 3 years on the same CPU. The general public which always kept their machines for 3 years or more might be justified keeping them for longer now.

Of course they don’t really know what makes these machines tick so they just buy based on the actual laptop designs and little improvements like battery size, smaller bezels, keyboard improvements, looks, weight, size etc. But if their IT guy or techie friend says all that is well and good but it won’t really be any faster then they might hold off on upgrades.

The school season purchases I suppose will always be around the same.

On the server side I don’t know much but it seems like they can’t stay ahead there. But those are ordered by IT managers who don’t care about branding. They just care about price/performance. So they will just order AMD.

Still I have to admit Intel branding, name recognition and marketing is second to none. So these technical issues don’t affect them as much we enthusiasts like to claim. Yet still they know and we know that a lot of purchases are made on the recommendation of whichever techie guy they have in their circle. Which is why Intel caters to enthusiasts. They know we help them sell product.

actually you hit the spot right

I have a surface pro 4 with i5 6300U and a whiskey lake looks hell of an improvement over my machine atm with same form factor and lower price and higher battery life
ofc 10nm is a brake to the improvement but saying none is a real b..s..t

same as been said... the node doesnt matter, result matters

in fact, AMD ( and we too) should be very happy Intel messed 10nm, otherwise ...continuous monopoly times

competition wins

waiting to buy ryzen 7nm desktop (replacement for i5-6600k 4.4GHz undervolted) and 7nm laptop (replacement for i5-6300U SP4) but neither desktop (+10%+ first 4C with lower power) or laptop look that way
same as Intels offerings- really wanna see the Icelake laptop parts

 

[beginner99]

The problem is that we have no idea what the problem with 10nm is right now.
It could just be that they can't reach 4.5Ghz + right now so they can't release a gen that will match the previous one.
Can any other fab reach 4.5Ghz + right now on low nm on such a complicated arch?
Intel could very well still be leading in everything but not releasing 10nm because it doesn't make economic sense for them to do so,14nm still sells like crazy so they are not forced to do anything,they don't need the new gimmick of 10nm yet.

It's functionally yields. Ghz wouldn't matter for mobile. the cannonlake part without gpu was telling. Also that icelake will now come with a ton of variants especially gpu-side. If it would yield well but with "low" clocks", they would have moved at least Y/U parts and server parts to 10nm 1-2 years ago. Would only be an issue for desktop. 10 nm is imho dead. we will see couple niche parts and then intel will move to 7nm EUV in 2021 and 2022, hopefully.
Sad thing is, even if they are still on 14nm in 2022, what else can oems/people do? foundries wouldn't have enough capacity to fab so many AMD parts to suit whole x86 market so intel will still sell a lot.

 

[Topweasel]

^This. I haven’t read anything about TSMC or Samsung building up a couple of high capacity fabs for some unnamed top tier semiconductor designer. It’s just ridiculous rumor mongering at this point.

Well Actually they kind of did. Look at their 7nm and 5nm fabs. They threw nearly 10x into them to make them 3-4x larger than other fabs. TSMC by the time Intel moves to 7nm will have about 8 fabs worth of 7nm and 5nm production.

 

[Ajay]

Well Actually they kind of did. Look at their 7nm and 5nm fabs. They threw nearly 10x into them to make them 3-4x larger than other fabs. TSMC by the time Intel moves to 7nm will have about 8 fabs worth of 7nm and 5nm production.

8(!) Fabs @ 5-7N Do you have a link? Haven’t read this (read about one new fab @ TSMC and expansions @ Samsung).

 

[Dayman1225]

8(!) Fabs @ 5-7N Do you have a link? Haven’t read this (read about one new fab @ TSMC and expansions @ Samsung).

Considering TSMC currently has 5 12 Inch Wafer "Giga-Fabs" that isn't happening.