News Intel Announces Delay to 7nm Processors, Now One Year Behind Expectations

[watzupken]

Something is seriously wrong at Intel, they have a huge R&D budget and yet cannot get 10mm to work for the umpteenth time. This stinks of a leadership/management issue and I am not confident they can right the ship, specially with Apple going ARM and Nvidia could jump into the fray.

In my opinion, Intel is failing because of complacency. I cannot think of anything other than complacency that lead Intel to continue waiting for 10nm to work, while recycling the Skylake architecture for 5 generations. It seems like they are expecting AMD to continue with Bulldozer kind of performance, and no significant improvements from ARM SOCs.

They have deep pockets, which further adds to their complacency. The fact that they lost the fab advantage, means they lost half the advantage they used to have. Now they have to work even harder to deliver IPC gains to keep up with competition.

 

[Phaaze88]

I feel Intel has grown too big to be nimble.

A typical result.

Intel's a for profit company, and just like some others, they want to expand to multiple markets - because mo' money.
But then that means they move away from being a specialty company to a jack-of-all-master-of-none.

 

[InvalidError]

I feel Intel has grown too big to be nimble.

Intel bet on its own internal process R&D, ran into more complications than it bargained for and looks like it may have run into more unforeseen complications with 7nm. There isn't much that "nimbleness" could do about that.

I cannot think of anything other than complacency that lead Intel to continue waiting for 10nm to work, while recycling the Skylake architecture for 5 generations.

Intel isn't "waiting for 10nm to work", it scrapped most plans to scale up 10nm production in favor of fast-tracking 7nm and apparently ran into complications with that too. As for recycling Skylake, that is a byproduct of not having a sufficiently mature more advanced process to fab its more complex CPUs on. Development-wise, it is three architectures ahead of where desktop currently is, just lacking a suitable fab process to cost-effectively scale them up on.

 

[watzupken]

My prediction based on AMD p2e, margin growth and share price growth
By 2023
AMD makert cap equal Intel

NVIDIA starts to feel heat. R&D budget increaes significantly for RTG. They are trying to get ARM and cant clear regulatory hurtles as they would become a monopoly (nvidia likes that idea as they are ruthless). Possesing ARM would have clear benefits to forcing partners to incorporate NVIDIA eco system from servers to phones through licensing. Lawsuits from competitors will stop this. (Could threaten apple and google's plans)

I don't know if the numbers would work but nvidia may buy out excess fab production at tsmc just to slow down competitors is also a possibility. (Even if they don't use full cap.)

I agree that it is going to be a big problem if Nvidia manages to acquire ARM.

As for Nvidia buying out excess fab at TSMC to slow down competition, I think this is an interesting point, which may turn out to be valid.

 

[Giroro]

Called it.

In regards to weather a CEO should be from a technical field, I would just like tho provide a counter example in Boeing.
Dennis Muilenburg came out of engineering, yet his actions as CEO and company culture suggest he HATED Engineers. Short term thinking for fast profit, overworking, underpaying, high stress, no perks, layoffs.. They didn't even vacuum the tiny call-center style desks they pack their "best and brightest" into, at least back in the beforetimes. This resulted into major brain drain which further ruined the reputation of Boeing engineers who, when applying for real engineering companies, get comments like "if you're so smart, then why do you work at Boeing" - especially when they are in software.
They're going to be feeling that pain for a long time, with or without a pandemic.

Anyways, my point is sometimes people move into management because they don't actually like the job they were hired to do, or their coworkers, or they were just bad at it.

 

[InvalidError]

As for Nvidia buying out excess fab at TSMC to slow down competition, I think this is an interesting point, which may turn out to be valid.

Highly unlikely since AMD and other major clients that Nvidia may have axes to grind with book their wafer commitments a year or more in advance and no amount of wafers Nvidia may book after-the-fact will affect those. If AMD has a wafer supply agreement that sets the baseline at 180k wafers/month, that capacity is locked down and there is nothing Nvidia can do about it.

 

[Giroro]

I agree that it is going to be a big problem if Nvidia manages to acquire ARM.

As for Nvidia buying out excess fab at TSMC to slow down competition, I think this is an interesting point, which may turn out to be valid.

Its going to be a big problem if basically anybody interested in manufacturing ARM processors or mobile devices buys ARM.

 

[Geef]

Its not like anyone was expecting anything else from them. Just add another + to 14nm and they are good for another year or two.

 

[Phaaze88]

Its not like anyone was expecting anything else from them. Just add another + to 14nm and they are good for another year or two.

I'm not sure if that's going to work out too well. From what I've seen and heard so far, the 10th gen K SKUs have even less overclocking headroom compared to their predecessors - especially the 10900K; it's the worst of the 4.

 

[PapaCrazy]

No, he (BK) shot himself in the foot when he let the process run off the track while he was CEO. The mess was created with him in charge, why would you let him hang around to fix a multi-billion dollar mess he either created or let happen.

Whether he was the worst is a toss up between him and Otelini. Both ran the company horribly.

Not to mention his first reaction to specter/meltdown was to sell his stock before news got out. Dude wasn't just complacent or negligent, but actively avoided accountability.

 

[vinay2070]

This gives AMD a nice chance to move ahead Though I was really waiting to see how highend Intel GPU would perform

 

[Jim90]

poor intel. slept for ten years then woke up with a hangover.

THIS!!

 

[nofanneeded]

By that time 3nm will be in the market ....

 

[Deleted member 2720853]

It's okay Intel, you'll figure it out... eventually.

 

[bit_user]

I think Intel shot themselves in the foot when they fired Brian Krzanich from the CEO position.

https://en.wikipedia.org/wiki/Brian_Krzanich

Brian Krzanich was a Chemistry Major and worked as a process engineer in Intel's Fab's.
He worked his way up to CEO position, but he's still a Chemistry guy at the core.

Are you serious?? Just who do you think got them into this mess?

Not to defend Swan, but he simply hasn't been in charge long enough to derail them this much. Their issues go back a lot further.

 

[bit_user]

In regards to weather a CEO should be from a technical field, I would just like tho provide a counter example in Boeing.

...

Anyways, my point is sometimes people move into management because they don't actually like the job they were hired to do, or their coworkers, or they were just bad at it.

THIS.

A lot of us have seen developers or engineers move into management (or marketing, I'd add), and they are often not the best at engineering. Either that's not where their gifts lie, or they're simply more interested in the benefits of climbing the corporate ladder.

The best engineers often stay in a technical role, though some are reluctantly drafted into management and a subset of those actually embrace the challenges of their new position (instead of trying to sneak back into the code, at the expense of their management duties).

 

[TerryLaze]

poor intel. slept for ten years then woke up with a hangover.

Intel did not sleep for 10 years,they kept their tech close enough to AMDs offerings to not completely wipe them out,because monopoly laws...just look at what happened to AT&T, and guess what,that's exactly what they keep doing right now,intel has doubled their net income in the last two years,a thing they could have done 10 years ago if they started to add cores back then which would have wiped out AMD completely.

 

[barryv88]

Problem with BK was that fratting around with the girls at the office seemed to have been a bigger priority, along with riding out a monopoly for as long possible since AMD couldn't compete well enough with their FX arch. The result? Roughly 5-7 years of stagnation (however you wanna look at it) and every new core gen giving us the consumers a paltry 3-5% IPC uplift and barely no core expansion (the 6C chips were a total rip off and unaffordable to the masses anyway).
The i7-7700K HIGH END CPU still had only 4 cores! The big fat middle finger that Intel has shoved in the world's face for years. 7 generations of high end quad core. The software eco system as a result was kept back and hamstrung over this low core count grip that is still echoing today. Even a 2600K gets the job done, whatever you throw at it. Yet people boast about how well their sandy bridge chips have lasted. Its actually, very sad if you think about it. Those chips should have been buried and forgotten by now if Intel had a better attitude towards innovation.
TSMC is already working on 3nm and planning 2nm at this stage. Good luck catching up Intel...
BK was as bad a CEO as it gets.

 

[mitch074]

My prediction based on AMD p2e, margin growth and share price growth
By 2023
AMD makert cap equal Intel

NVIDIA starts to feel heat. R&D budget increaes significantly for RTG. They are trying to get ARM and cant clear regulatory hurtles as they would become a monopoly (nvidia likes that idea as they are ruthless). Possesing ARM would have clear benefits to forcing partners to incorporate NVIDIA eco system from servers to phones through licensing. Lawsuits from competitors will stop this. (Could threaten apple and google's plans)

I don't know if the numbers would work but nvidia may buy out excess fab production at tsmc just to slow down competitors is also a possibility. (Even if they don't use full cap.)

Not unreasonable, however, you're forgetting one thing : RISC-V is a thing now, and if Huawei goes forward with its plans we might see the first physical chips next year. Longer term, the dangers for US-based chip makers (I'll include the UK in that too, it's not much more than a puppet state now) if Loongson and Zhaoxin actually take off.

 

[Chung Leong]

I feel Intel has grown too big to be nimble.

It has grown too big to be paranoid. I thought it was a sign of trouble when Intel blew 15B on Mobileye.

 

[Spike666]

I worked at Intel from the 8088 introduction until just before the Pentium introduction. This would never have happened with Andy Grove at the helm.

 

[Spike666]

Whether he was the worst is a toss up between him and Otelini. Both ran the company horribly.

I never understood how Otellini got there. I worked in the 486 group when he was co-manager and he really didn't get it. He was a finance guy who was rebranded as a marketing guy but with no understanding of marketing. He also took the eye off the ball with mobile, what a huge mistake.

 

[usiname]

Just wait one more year and they will delay it again with 1 year just like their 10nm do every year.

 

[PCWarrior]

So Intel's 7nm production is delayed for 6-months for some parts (albeit not all as Ponte Vecchio is still on track). I don’t really see a problem. Because if they manage to finally get good yields on 10nm they should have no problem remaining competitive. Let’s not forget that Intel has yet to meaningfully launch anything on 10nm and they have thus three generations worth of architectural improvements ready to go when the process finally allows mass production of bigger dies. There is at least a 50-55% IPC increase on average (1.18x1.10x1.15=1.49x, 1.18x1.12x1.17=1.55x) between Skylake and Golden Cove. And in some applications like tile-based rendering (Blender, Cinema4d, 7Zip, etc) we are really looking for at least 70-75% improvement (1.25x1.15x1.20=1.725x)

Combining IPC with the 2.3x density improvements of 10nm over 14nm it means that they can possibly make a 10nm 8-channel 56-core die with 80 PCie5 lanes that has equal die area as a 14nm 28-core and that to have a 3x higher performance on average, and 3.5x higher performance in tile-based rendering. In tile-based rendering, the 64-core 3990X is only around 2-2.1x better than the W-3175X. AMD will probably reach 2.5x with their MCM 64-core zen3 top sku and then 3.5-4x with a 5nm MCM 96-core zen4 top sku.

So I see Intel staying pretty competitive throughout 2021 and 2022 and that is without even considering the greater adoption of AVX-512 or other specialised instructions that Intel is including in their new cpus for AI, DL, etc.

 

[jeremyj_83]

The best engineers often stay in a technical role, though some are reluctantly drafted into management and a subset of those actually embrace the challenges of their new position (instead of trying to sneak back into the code, at the expense of their management duties).

All depends on where the company has its headquarters. As I stated earlier, a lot of German companies have their CEOs come from R&D. Their idea to increase profit and market share is to make a better widget rather than spin out some terrible product over and over and say its the best. Dr. Lisa Su is another example of an engineer becoming a CEO.