CPU Intel 7nm process delayed by 6 months - not hitting retail until mid 2022

Valantar

Shrink Ray Wielder
Original poster
Jan 20, 2018
2,073
2,049
So Intel just finished their earnings call, and pretty much announced that the X86 market will belong to AMD through 2021, possibly 2022. They have identified a "defect mode" in their 7nm process, meaning yields are now 12 months behind projections. This has led to an official 6-month delay of the process being launched (guess their roadmap had some buffer built in?). Nonetheless, Intel 7nm CPUs, previously projected for "end of 2021" should then be expected no sooner than mid-2022 (though Tom's Hardware says late 2022 to early 2023). They are reportedly looking at outside foundries to alleviate this, probably mostly for their GPUs.

 

JSItems

Caliper Novice
Apr 28, 2019
25
21
This is somehow less surprising to me than AMD announcing they're still on schedule to release Zen 3 this year.

Does anyone expect to see Intel 7 nm before 2023?
 

AlexTSG

Master of Cramming
Jun 17, 2018
557
529
www.youtube.com
Fabrication looks like it's a big challenge. There were rumor's just a few days ago that NVIDIA may have also run into problems with finding 7nm capacity at TSMC (for the price they were willing to pay), and may be looking to Samsung's 8nm process for the initial release of it's next gen cards.

Overall, things are looking good for AMD on both the CPU and GPU sides of their business.
 

Valantar

Shrink Ray Wielder
Original poster
Jan 20, 2018
2,073
2,049
Fabrication looks like it's a big challenge. There were rumor's just a few days ago that NVIDIA may have also run into problems with finding 7nm capacity at TSMC (for the price they were willing to pay), and may be looking to Samsung's 8nm process for the initial release of it's next gen cards.

Overall, things are looking good for AMD on both the CPU and GPU sides of their business.
Haven't there been rumors of consumer Ampere being Samsung 8nm for at least a year?
 

Phuncz

Lord of the Boards
Editorial Staff
Moderator
Gold Supporter
May 9, 2015
5,433
4,704
I can't even...

It's hard for me not to blame Intel's strategy from 4th gen Core onwards that was geared towards respinning their same 14nm but offset it with more and more proprietary tech (like AVX-instructions and Thunderbolt) just to dig their monopoly trenches even deeper.

After Broadwell in 2014 Intel has been stuck on 14nm and they clearly weren't going to be introducing 10nm or smaller any time soon after 3 years of serious competition from AMD. I don't like this situation of now Intel not offering much competition towards AMD. It has been clear since Skylake that Intel was bearing down to make a lot of profit with a very small development and research budget. Now they're being mis-managed and with every AMD CPU release they fall behind more and more.
 
  • Like
Reactions: AlexTSG

AlexTSG

Master of Cramming
Jun 17, 2018
557
529
www.youtube.com
Haven't there been rumors of consumer Ampere being Samsung 8nm for at least a year?

From what I've read it seems that NVIDIA was hoping that TSMC would lower their pricing in response to leaked rumors that Samsung was going to get the lion's share of the business for next gen cards. When TSMC didn't budge, NVIDIA found that AMD had bought up the spare capacity.

It looks like the 3080Ti cards need to be made on 7nm to hit their performance targets, where cards lower down the order can use the cheaper, and less efficient, Samsung 8nm process.

Recently, in this Moore's Law is Dead video the rumors are that Nvidia may initially go with Samsung, and release a "Super" refresh within 6 months or so to get the performance to where they want it on TSMC 7nm.



As long as there's good competition, we, the consumers, win!
 
Last edited:

AlexTSG

Master of Cramming
Jun 17, 2018
557
529
www.youtube.com
I can't even...

It's hard for me not to blame Intel's strategy from 4th gen Core onwards that was geared towards respinning their same 14nm but offset it with more and more proprietary tech (like AVX-instructions and Thunderbolt) just to dig their monopoly trenches even deeper.

After Broadwell in 2014 Intel has been stuck on 14nm and they clearly weren't going to be introducing 10nm or smaller any time soon after 3 years of serious competition from AMD. I don't like this situation of now Intel not offering much competition towards AMD. It has been clear since Skylake that Intel was bearing down to make a lot of profit with a very small development and research budget. Now they're being mis-managed and with every AMD CPU release they fall behind more and more.

Personally, I'm glad that Intel's misstep has given AMD a chance. I'm not some AMD fanboy either, every system I own has an Intel CPU, but Intel can't just keep adding pluses to the end of their 14nm process.

Intel's financial position and brand will allow them to weather this (I think their revenue is still close to 10x AMDs), but I'm hoping they consider a move to outside fabs, sooner rather than later, so that there's some competition over the next few years.
 
Last edited:

AlexTSG

Master of Cramming
Jun 17, 2018
557
529
www.youtube.com
This 7nm delay has caused downgrading of Intel’s stock by analysts everywhere. Intel’s share price has dropped around 16% as I’m typing this post.
 

Valantar

Shrink Ray Wielder
Original poster
Jan 20, 2018
2,073
2,049
"AMD IS DOOMED"
... said nobody since ca. 2018?
From what I've read it seems that NVIDIA was hoping that TSMC would lower their pricing in response to leaked rumors that Samsung was going to get the lion's share of the business for next gen cards. When TSMC didn't budge, NVIDIA found that AMD had bought up the spare capacity.

It looks like the 3080Ti cards need to be made on 7nm to hit their performance targets, where cards lower down the order can use the cheaper, and less efficient, Samsung 8nm process.

Recently, in this Moore's Law is Dead video the rumors are that Nvidia may initially go with Samsung, and release a "Super" refresh within 6 months or so to get the performance to where they want it on TSMC 7nm.



As long as there's good competition, we, the consumers, win!
I don't get why people keep listening, to that channel. So much nonsense, so many unsubstantiated rumors repeated as if true, so many errors uncorrected. I see zero reason to trust anything said there even if some of it inevitably turns out to be accurate - there is no way of knowing what is and what isn't.
I can't even...

It's hard for me not to blame Intel's strategy from 4th gen Core onwards that was geared towards respinning their same 14nm but offset it with more and more proprietary tech (like AVX-instructions and Thunderbolt) just to dig their monopoly trenches even deeper.

After Broadwell in 2014 Intel has been stuck on 14nm and they clearly weren't going to be introducing 10nm or smaller any time soon after 3 years of serious competition from AMD. I don't like this situation of now Intel not offering much competition towards AMD. It has been clear since Skylake that Intel was bearing down to make a lot of profit with a very small development and research budget. Now they're being mis-managed and with every AMD CPU release they fall behind more and more.
You've got a point, but we need to remember that 10nm was originally planned to hit the market in... 2016, IIRC? One really shouldn't discount just how late that node has been, nor how problematic. That isn't due to mismanagement per se, but rather due to Intel setting extremely ambitious goals for it (likely due to 22nm and 14nm being such successes for them) and a significant number of serious issues arising both due to the level of ambition and unrelated to this. They should of course have been far, far more flexible in their handling of this - refusing to backport a new architecture until nearly five years later (Rocket Lake) is downright lunacy, and their continued insistence on "10nm will arrive soon, let's just add a + to 14nm in the meantime" now seems to have been continued far too long. Ice Lake should have been backported to 14nm and released on the desktop last year if not earlier. But while these last points are due to bad management, it is bad handling of an already bad situation. More R&D money would likely not have made 10nm arrive any faster in high volumes.

Edit: damn autocorrect
 
Last edited:

Phuncz

Lord of the Boards
Editorial Staff
Moderator
Gold Supporter
May 9, 2015
5,433
4,704
You've got a point, but we need to remember that 10nm was originally planned to hit the market in... 2016, IIRC? One really shouldn't discount just how late that node has been, nor how problematic. That isn't due to mismanagement per se, but rather due to Intel setting extremely ambitious goals for it (likely due to 22nm and 14nm being such successes for them) and a significant number of serious issues arising both due to the level of ambition and unrelated to this. They should of course have been far, far more flexible in their handling of this - refusing to backport a new architecture until nearly five years later (Rocket Lake) is downright lunacy, and their continued insistence on "10nm will arrive soon, let's just add a + to 14nm in the meantime" now seems to have been continued far too long. Ice Lake should have been backported to 14nm and released on the desktop last year if not earlier. But while these kat points are due to bad management, it is bad handling of an already bad situation. More R&D money would likely not have made 10nm arrive any faster in high volumes.
That's the problem: Intel's 10nm and 7nm were on different teams being developed, so failure of 10nm shouldn't have impacted 7nm much, if at all. Or it could, if it would have been mismanaged by limiting or delaying development of 7nm because of the delays of 10nm, which seems a deliberate choice in a monopoly mindset (considering Intel's massive financial power) when expecting being able to extend 14nm for years without any competition.

There has been a lot of management restructuring since 2016 when the bad news during financial calls started becoming a regular thing, I contemplate those job replacements was to appease the shareholders. This often ends up in not appointing people with the real solutions for the problem but the people who can make the boldest claims believable. Companies are often ruined by the shareholders because they often don't understand the market their company is in.

Lets not forget how Intel released one single low end SKU with 10nm in extremely limited quantities two-three years after it was supposed to be released. One can assume it wouldn't have released at all if AMD didn't put the release of competing 7nm CPUs in 2018-2019 on the roadmap in 2017.

Today this insight has been given:

We can't blame ambition for this, as AMD's ambition has been much larger in my opinion, with a much smaller company. If anything, Intel has the financial capacity and talent to achieve that ambition. Intel has the financial capacity for research, development and talent acquiring. That AMD leap-frogged Intel can only be explained by having become a mismanaged company that has lost the goal they are selling products from their sights.
 

Valantar

Shrink Ray Wielder
Original poster
Jan 20, 2018
2,073
2,049
That's the problem: Intel's 10nm and 7nm were on different teams being developed, so failure of 10nm shouldn't have impacted 7nm much, if at all. Or it could, if it would have been mismanaged by limiting or delaying development of 7nm because of the delays of 10nm, which seems a deliberate choice in a monopoly mindset (considering Intel's massive financial power) when expecting being able to extend 14nm for years without any competition.

There has been a lot of management restructuring since 2016 when the bad news during financial calls started becoming a regular thing, I contemplate those job replacements was to appease the shareholders. This often ends up in not appointing people with the real solutions for the problem but the people who can make the boldest claims believable. Companies are often ruined by the shareholders because they often don't understand the market their company is in.

Lets not forget how Intel released one single low end SKU with 10nm in extremely limited quantities two-three years after it was supposed to be released. One can assume it wouldn't have released at all if AMD didn't put the release of competing 7nm CPUs in 2018-2019 on the roadmap in 2017.

Today this insight has been given:

We can't blame ambition for this, as AMD's ambition has been much larger in my opinion, with a much smaller company. If anything, Intel has the financial capacity and talent to achieve that ambition. Intel has the financial capacity for research, development and talent acquiring. That AMD leap-frogged Intel can only be explained by having become a mismanaged company that has lost the goal they are selling products from their sights.
You're mixing up a few things here. The 7nm delay is - at least as far as anyone outside of Intel knows - entirely unrelated to 10nm being late. Intel 7nm has never been delayed before this - probably partly due to not having public dates given for it, which is an obvious thing to avoid until the preceding node is at least near ready, after all. But regardless of this, the current 7nm delay is entirely unrelated to the lateness of 10nm.

There is of course still a relation between the two: developing a new node is a complex research and development process, which inevitably builds on previous research and development. The 7nm team thus needs knowledge, techniques, experience and equipment developed for and by the 10nm team, but also needs to build on this and improve on pretty much everything. If not, one could assume that Intel could just skip 10nm entirely and move on to 7nm, which given the projected 2016 launch of 10nm should then have begun development around 2015 or even earlier. Of course on this level 7nm has been delayed, as by that time nothing at all was ready, but those delays are long gone and happened before 7nm had any publicized timeframe. Of course the vast majority of this has been ready for years, as the 10nm delays of recent years haven't been due to equipment not being ready or anything like that - and likely the 7nm team has learnt a lot precisely from the reasons behind 10nm being delayed for so long.

I entirely agree that shareholders often have no clue whatsoever what is best for the companies they own, and that a lot of stupid decisions are made to appease them - but Intel's poor strategy and management can't directly be attributed to that. Rather, it seems to have been built on what is essentially the modus operandi for most contemporary corporations: prioritizing easy profits and cutting expenses wherever possible. (This of course disregards the decades-long experience that this approach inevitably drives companies into the ground as they shed everything that once made them leaders in their fields - it's a classic tactic of so-called vulture capitalism.) Intel saw themselves with a 30-50% IPC advantage, 5-year node advantage and massive efficiency advantage and thought "hey, let's just take it easy and rake in cash for a while." for a lot of their business - hence the years upon years of quad-cores and <10% IPC improvements. The thing is, this is not why 10nm failed. Intel 10nm was a massively ambitious node, which in itself is a goal that goes fundamentally against this thinking. Nor was it underfunded. It simply had a lot of issues due to trying to make a huge leap rather than an incremental development. 10nm was very clearly Intel's plan to cement their lead for another 5 years or more, and in no way a low priority. It just turned out to be much harder to achieve than they initially thought.

As for that lone 10nm SKU, it had absolutely nothing to do with AMD. It was launched purely to avoid a shareholder lawsuit, as Intel had previously promised that 10nm would "ship for revenue" in whatever year it appeared. This is down to US finance law, which allows shareholders to sue companies for making misleading promises about future products or performance. If Intel hadn't launched that single garbage SKU, they would have opened themselves up to potentially billions of dollars in payouts to shareholders due to making promises they couldn't keep.

As for AdoredTV, that channel is so full of garbage that I refuse to watch a single video from them, no matter its subject.

And lastly, you entirely misunderstood my point: I didn't say that the reason for Intel's woes is them being generally ambitious (which is where your AMD comparison comes in), but very specifically due to the 10nm node being a very ambitious design. Intel's strategies outside of this have been conservative to the extreme - as much of a cause for their current woes as the node delays (a less conservative approach would have seen Ice Lake backported to 14nm long ago, which would have put them in a much better position). But you are arguing that given enough money and enough talent, anything is possible, which is a very naive stance. Intel's 10nm woes are sufficient proof of this, as there is no evidence of node development being underfunded or lacking in talent. Time is also a necessity in the equation you're sketching up there: Given enough money, talent and time (pretty much) anything is possible. Intel had the first two, but hit a series of major bumps in the road. Throwing more money at the problem would very likely not have changed anything.