Rumors of ARC's cancellation...

Yea well- this is going to take a Jobsian type turnaround. You can fanboi about your friend all you want- Im sure he's a great kisser. From my perspective: I've already worked for a company in this condition and understand intimately what needs to happen.
Again, totally don't disagree - but either Intel gives up and goes out of business, or they pick a ruthless engineering leader to try and fix it. Gelsinger is about the best of those I know - if he can't, I'd be surprised who could (Raja? LOL).

Also definitely not my friend - I don't float at those levels quite yet, but I worked on his teams and with his other hand-picked C-levels, and the guy has a history for being very good at what he does.
Intel is finished as an industry leader unless they can catch up on process. AMD is pumping out 5nm designs while knee deep in Zen 5 development.
That's what they said about Intel in the Athlon64 days, and what they said about AMD in the ... well, most of the last decade, and what they said about Apple, and so on. All things are cyclical in this industry, if you've got the cash to ride out the slow times. Intel definitely has the cash. I'm not an Intel fanboy by any means (most of my main systems are AMD) - but I can also look at the business fundamentals and see that they have ~time~ and ~resources~ to fix this, with the right leadership and (again) time to get there. Can't roll out new CPUs in 12 months, after all, any more than you can roll out a new GPU that fast. Those take years to develop. What happens when AMD hits a snag?
There's some slight of hand going on here:

The first "joke" is e-cores. Now in a laptop this is completely understandable (Apple's primary market). But not in high performance computing. The only reason Intel went that route is to stretch 10nm as far as possible. And they are hitting a wall.
Absolutely. I find the required scheduler and everything else... less than optimal by far. The funnier thing is that they're having to stretch 10nm after fighting that hard to even get there!
Unless they outsource their manufacturing right now- they're done. They don't even have a 7nm part. And Zen 4 is cruising at 5nm. A 2023 7nm release coming from an Intel fab isn't going to save them. AMD has legs with Zen 4 and 5. There's lots of headroom there.
Maybe. Or maybe they need to license the fab tech - because with the way TSMC and co have been tapped out recently, that's the only way to actually produce anything in quantity. Only so many 5nm and 7nm fabs around. And, FWIW, and to give intel a little credit, their transistor density at 10nm is FAR better than it has any rights to be - up there (IIRC) beating TSMC 7nm by a small margin.
https://www.digitimes.com/news/a202...features,and Samsung's 7nm nodes respectively.

And yes, TSMC has since managed to nudge out the lead again with several improved versions of 7nm, but Intel isn't THAT far behind.
How much you want to bet that e-cores go away in x86 land assuming Intel catches up in process? It's a lock.

They are, in short, pucked.
Oh they will. Technology wise they're behind.

Business wise? They've got a LONG runway to work with still - heck, they made 3.1B the quarter before last (operating income), and have 27B in cash on hand. You can ride that for a long time.
 
Does the new Intel fab being built in Ohio have a place I this discussion? Are they really calling it quits or just getting started in their Discrete GPU venture? This is the 1st of 2 plants they plan to build in the US by 2025. That's not far away and falls within the 2nd or 3rd gen Discrete GPU from them along with better CPU fabs hopefully benefitting consumers with more choice, competition and price levels.
 
Business wise? They've got a LONG runway to work with still - heck, they made 3.1B the quarter before last (operating income), and have 27B in cash on hand. You can ride that for a long time.

Well good we are having a discussion....

In my mind if Intel is going to claw back- they need Jim Keller as CEO. He said he left for personal reasons. The word on the street was that it was a dispute over fab strategy.

That is the issue. And that is the core of my pessimism. Gelsinger is the wrong guy.
 
Last edited:
Does the new Intel fab being built in Ohio have a place I this discussion? Are they really calling it quits or just getting started in their Discrete GPU venture? This is the 1st of 2 plants they plan to build in the US by 2025. That's not far away and falls within the 2nd or 3rd gen Discrete GPU from them along with better CPU fabs hopefully benefitting consumers with more choice, competition and price levels.

The problem is that the industry has changed. It's very difficult to be both the engineering firm and the production firm in semiconductors these days. Actual production of the ships is so demanding, that a company really has to specialize in it.

Intel is trying to be two things.

I believe Jim Keller knew this when he left Intel. And was opposed to the in-house investment in production.
 
Does the new Intel fab being built in Ohio have a place I this discussion? Are they really calling it quits or just getting started in their Discrete GPU venture? This is the 1st of 2 plants they plan to build in the US by 2025. That's not far away and falls within the 2nd or 3rd gen Discrete GPU from them along with better CPU fabs hopefully benefitting consumers with more choice, competition and price levels.
Yes and no. They're still building fabs, both because of the government, and because, well, we need more fabs - EUV fabrication is ... complex, and takes a LOT of space (and a @#$@ ton of water), so it's not like you can roll out a new one in a day to spin up manufacturing (see most of 2020/2021).
Well good we are having a discussion....

In my mind if Intel is going to claw back- they need Jim Keller as CEO. He said he left for personal reasons. The word on the street was that it was a dispute over fab strategy.
Now, this is a fascinating idea - the thing to me is, Jim hasn't had any experience at ~that level~, and the CEO (even at Intel) doesn't do nearly as much with the actual engineering as much as setting the strategy and hiring the right subordinates. Or, I should say, he shouldn't.

I'd 100% get behind Keller as president of product or engineering - hell, give him the whole stack to manage and lead - but I'm not sure if using his talents at CEO would be the right idea, or a waste. He wouldn't be concentrating on actual engineering anymore (where they DO need his skills and experience - even if just to call bullshit or point them in a different direction, or cut out the cultural horsecrap that is rumored to go on there), he'd be talking to fund managers and the SEC, and dealing with a lot of other things that the CEO has to. My fondness for Gelsinger is from cutting out crap products and non-core competencies, and then hiring the right CTO/CIO/CMO/President of Engineering to get the job done on the right products. Keller truly under Gelsinger, with the power to make change - could be very interesting.

That being said - Gelsinger does have an opinion on where to go, and it may even be biased by his history in the "good old days" - I'll grant that - which could cause conflict. Given when Keller left (prior to Gelsinger starting), it could have been just like you said - a conflict around fab strategy, given many of the other rumors around Intel these days (that the "chair sitters" and beancounters are refusing to let the real architects and engineers make the necessary changes) - basically getting shut down by a talking head. Keller was only an SVP there, and again didn't overlap, which means there was at LEAST one, if not two levels above him before you got to the C-suite, never mind to the CEO... He could have gotten shut down by almost anyone. It could very well be true that he and Gelsinger would work fine together and solve that problem, even if it meant licensing the fab technology or disposing of it (but given the number of other things that Intel runs fabs for, I'm still not sold that's the right idea - I'd have to dig into the 10k to see if that product line is broken out on the PNL).

In short - bogging Keller down with the shenanigans that it takes as a CEO might take away his best skills from where they're needed - building an actual world class engineering team to make the next generation of groundbreaking CPUs - and waste him on stuff that others are good at instead.

It's an interesting idea I'd have to think about.

That is the issue. And that is the core of my pessimism. Gelsinger is the wrong guy.
Pure engineering heads often don't work out - they're not built to deal with the "bullshit" that comes along with it, and that's a waste of what they're so supremely good at (especially ones like Keller that also make great leaders). Jim Keller is a once-in-a-generation engineer and engineering leader. Other than as the current talking-head for Intel, do you have any experience with Gelsinger's companies (honest question)? Given how he came up under Andrew Grove, and his experience (and history) at Intel, he seems like the logical choice - and Grove was the one who bloody built Intel (even over the other two founders, and certainly believed that sometimes it was time to exit a market, like fabs, as the company grows and evolves). The marketing idiot they had in Kraznich was definitely the wrong choice though.

The problem is that the industry has changed. It's very difficult to be both the engineering firm and the production firm in semiconductors these days. Actual production of the ships is so demanding, that a company really has to specialize in it.

Intel is trying to be two things.

I believe Jim Keller knew this when he left Intel. And was opposed to the in-house investment in production.
Or it could have been a personal conflict with other leadership up above - ones refusing to change/move/fire/hire/get out of the fab business/any of it. Sadly, we may never know, as I'm sure he's under an iron-clad NDA over the topic. :(

Intel is trying to do two things - history will tell us if that's the right idea or not... I'd love to see Keller back there innovating again though.
 
Raja said on twitter the rumors are false. He may just be playing public PR hoping to turn things around. The rumors in the end may just just be Intel execs telling the graphics team to shape up or risk cancellation and the leaks being misinterpreted.
 
If it was truly canceled, wouldn't Raja be gone/coincidentally moved on by now?
Depending if he's found another spot yet. He is kind of a lame duck. Barely got Polaris out the door at AMD and now is heading up a lame duck at Intel. Something tells me Nvidia doesn't want him.
 
Given when Keller left (prior to Gelsinger starting), it could have been just like you said - a conflict around fab strategy, given many of the other rumors around Intel these days (that the "chair sitters" and beancounters are refusing to let the real architects and engineers make the necessary changes) - basically getting shut down by a talking head. Keller was only an SVP there, and again didn't overlap, which means there was at LEAST one, if not two levels above him before you got to the C-suite

Snipped so neither of us revisit the mulberry bush....

I hate to say "my sources".... because I'm not one of "those guys".... so I won't.

I think, as you posited, that is exactly the issue- as far as function. And "I think" Keller wanted to outsource production either as a supplement or permanent business model- as AMD has successfully done. And "I think" Intel, in some powerful departments, said "we have to produce our own stuff".

"I think" Venkata Renduchintala was onboard with outsourcing as well. We know Renduchintala left. We know he was presumed to be a future CEO at some point.

This looks like a corporate housecleaning after an internal coup. It really does. Allowing Keller and Renduchintala to leave is a serious issue. No one, that actually knows anything about producing shit in this industry, looks at those exits and thinks "Oh... things will get better".

So.. yes... this may be a case where the wrong people have been handed power.

If those people chose Gelsinger there's only two possible and mutually exclusive motivations:

1. The powers that be wanted a solid Intel person at the helm who would maintain the status quo.

2. The powers that be had a moment of clarity and decided to make corrective course changes.

I fear scenario 1. I hope for scenario 2.

As a corollary:

General Motors is now essentially an engineer run company. The CEO is an engineer. And the rest of the managerial structure is chock full of engineers: The are producing a Corvette that rivals cars that are twice the price.

Intel needs to make a similar move. But you don't bring on the guy who designed the Chevy Nova to build a corvette. That's how I view Gelsinger. Wrong person in the wrong seat. It's like trying to get a security administrator to write a video editing application.

Do you keep Gelsinger at a high level in the company?... absolutely. But the CEO and top management (which should include Gelsinger) needs to be fully packed with engineers with track records.

Gelsinger might do that- let the engineers create some miracles.

5nm is a lot different than 1 µm, however.

In closing: I understand this guy is a friend of yours. And I'm glad we can have this discussion. I'm not bagging on Gelsinger: He's a credit to the industry.
 
Snipped so neither of us revisit the mulberry bush....

I hate to say "my sources".... because I'm not one of "those guys".... so I won't.

I think, as you posited, that is exactly the issue- as far as function. And "I think" Keller wanted to outsource production either as a supplement or permanent business model- as AMD has successfully done. And "I think" Intel, in some powerful departments, said "we have to produce our own stuff".

"I think" Venkata Renduchintala was onboard with outsourcing as well. We know Renduchintala left. We know he was presumed to be a future CEO at some point.

This looks like a corporate housecleaning after an internal coup. It really does. Allowing Keller and Renduchintala to leave is a serious issue. No one, that actually knows anything about producing shit in this industry, looks at those exits and thinks "Oh... things will get better".

So.. yes... this may be a case where the wrong people have been handed power.

If those people chose Gelsinger there's only two possible and mutually exclusive motivations:

1. The powers that be wanted a solid Intel person at the helm who would maintain the status quo.

2. The powers that be had a moment of clarity and decided to make corrective course changes.

I fear scenario 1. I hope for scenario 2.
BINGO. And oh hell yes to your posited scenario, although I'll add that the "powers that be" may have been the board, and thus the removal of Kraznich ("oh you fucked it up royally now losing them, you're out!")... praying for 2. A monopoly in the CPU industry is good for no one. I LIKE competition. Northwood vs AMD64 was excellent. 10th Gen vs Zen2 was excellent. I WANT that constantly. I've got TRX4 and x299. I want options! Real ones!
As a corollary:

General Motors is now essentially an engineer run company. The CEO is an engineer. And the rest of the managerial structure is chock full of engineers: The are producing a Corvette that rivals cars that are twice the price.

Intel needs to make a similar move. But you don't bring on the guy who designed the Chevy Nova to build a corvette. That's how I view Gelsinger. Wrong person in the wrong seat. It's like trying to get a security administrator to write a video editing application.
Fair. The 486 was good though!!! :D
Do you keep Gelsinger at a high level in the company?... absolutely. But the CEO and top management (which should include Gelsinger) needs to be fully packed with engineers with track records.

Gelsinger might do that- let the engineers create some miracles.

5nm is a lot different than 1 µm, however.

In closing: I understand this guy is a friend of yours. And I'm glad we can have this discussion. I'm not bagging on Gelsinger: He's a credit to the industry.
Hoping he doesn't let the ego get in the way - and lets be honest, if he's a CEO, he's got an ego with a capital E - and does what I've seen him do. Hire the right people. Tell them the priority. Let them execute, and get out of the way. If they tell you that it's time to get out of fabs - get out of fabs.
 
Raja said on twitter the rumors are false. He may just be playing public PR hoping to turn things around.
I'm fairly certain that this sort of information is extremely sensitive for a publicly traded company; ie Raja would be in a huge heap of legal trouble if he claims the rumors are false but in the end they turn out to be true.
 
I'm fairly certain that this sort of information is extremely sensitive for a publicly traded company; ie Raja would be in a huge heap of legal trouble if he claims the rumors are false but in the end they turn out to be true.

Or he hasn't yet been informed!
 
when you're recreating the complexity of a modern high-end supercomputer in a slot, it become nearly impossible to make any money when yo already have two well-established competitors you have to displace from the market - its the same reason most broadband companies wont invest in an already crowded market

oh,. and reason 2: if the etherium going proof-of-stake permanent kills GPU mining centers, we will be returning back to thew days of falling discrete card sales;its a lot easier to make i integrated graphics pay-off

View attachment 508891

Don't forget that Apple proved with their new "M" ARM processors that Integrated graphics don't necessarily have to suck. The days of needing a dedicated GPU for a mid-range gaming PC might be numbered.
 
Raja Koduri is not the person who's words you should trust. If Pat Gelsinger said it, sure. If literally anyone else said it, sure. But Raja has a history of outright lying. Desperate, panicked and uninformed lying, but lying nonetheless.


Also remember, they've technically Launched Arc in Q2™© by selling one boxed card off the back of a truck in Whuan by June 30th.

So when they say New Product lines by end of 2022™© I expect exactly 2 new products selling only in asian territories at extremely limited volume in mid-to-late december.
And got promoted for the successful ARC launch… 🫣
 
when you're recreating the complexity of a modern high-end supercomputer in a slot, it become nearly impossible to make any money when yo already have two well-established competitors you have to displace from the market - its the same reason most broadband companies wont invest in an already crowded market

oh,. and reason 2: if the etherium going proof-of-stake permanent kills GPU mining centers, we will be returning back to thew days of falling discrete card sales;its a lot easier to make i integrated graphics pay-off

View attachment 508891
This chart actually speaks more to the shift away from desktop to laptop than it does about desktop discrete to desktop integrated.
 
This chart actually speaks more to the shift away from desktop to laptop than it does about desktop discrete to desktop integrated.


well, if so, , i just wish their reports were less vague: they only ever report desktop shipments (so I assumed that covered them all)
 
Back
Top