News [intel] Jim Keller resigns from Intel

[soresu]

It looks like they have some big/little plans for something? Would be interesting if Jim came back to AMD to help them out.

Interesting certainly, perhaps they have a secret Zen inspired Jaguar successor uArch in stealth mode - otherwise that would be a bit odd.

 

[Nereus77]

Yep. Nuvia came to my mind immediately when the news of him resigning from Intel broke.
That would be a hell of a dream team there.

I was told by a source that despite his personal family issue, Mr. Keller has been hired by a company already. I'm lukewarm to this rumor because I'm not 100% positive what I should make of it given the severity of this family issue juxtaposed to his abrupt exit from Intel. Nuvia does seem likely.

Nuvia would be his sort of thing, I'd imagine.

 

[Gideon]

Nuvia would be his sort of thing, I'd imagine.

That was my immediate guess as well (when I first heard the news of departure). The question is if they find him useful, considering they have a working team of crazy talent already.

 

[Nereus77]

That was my immediate guess as well (when I first heard the news of departure). The question is if they find him useful, considering they have a working team of crazy talent already.

That's easy.
Ask him what his dream project would be.
Then give him a blank page, his own team, deep pockets and tell him to make it happen.

 

[RasCas99]

The amount of worshiping that goes on in this thread is beyond ridiculous .......I cant even fathom how you guys come up with all of this ...... jeeez

 

[DAPUNISHER]

The amount of worshiping that goes on in this thread is beyond ridiculous .......I cant even fathom how you guys come up with all of this ...... jeeez

Are you going to limit yourself to ridicule, or do you have some valuable input to contribute to the discussion?

 

[RasCas99]

Are you going to limit yourself to ridicule, or do you have some valuable input to contribute to the discussion?

Well my input was that folks needs to get a grip of what they believe Jim Keller is capable of doing , I tried to avoid replying directly to ppl as it might offend them , sorry if it came rude.

Jim Keller is a big time old school manager , might be a good one even (I left Intel before he arrived) , from my friends small interactions with him @Intel, he was rude and vocal , but in the end he didnt change anything substantial of how they do things , he was not the chief Arch for Zen as so many folks here give him credit for (that was Michael Clark) , although he was the top boss in the division.

Working for over 15 years as a chip designer , i can tell you that , yes brilliant architects are extremely important for your chip success , maybe more then any other position in the org , BUT Uarch is almost always a big team and the innovations and progress they make are spread around , sure we have some folks that are a tier above the rest if you try to give them a score , but they rarely overlap in their expertise , and i would say that Jim Keller is so far from actual design of the chips now days that its not realistic to think about him as the guy who "dreams up" chips , In intel he was the guy that was supposed to put the right ppl in the right places and take out the trash management , which he did to some extent as i know some of the folks he promoted and some that he personally fired , but from a product line ? he never was the guy who had anything to do with defining and designing a new CPU for Intel , could he say "you suck , go get me a better design" after looking at projected performance/power/die size numbers , yes he could say it ,and he did , but thats not what some folks around here seems to think he is doing when he jumps around the companies , they believe he is the one defining those Archs , and just goes to a different company once his project is done or in execution.

The next company he will go to will hire him for his managerial skills and his experience in running a design team and not his technical skills which were not used in years to create anything , is technical skills are used to manage a technical organization , which is important (as i said , he can identify the smart and talented ppl and give them credit and control) , he can use his technical skills to ask the good questions and make the big calls , for example , do we TO in time or we push TO due to late bugs trends not converging , do we go to mass production in our current post silicon state , or we delay and wait for things to converge , do we approve 250 features for the next project , or do my team wont be able to execute in time and a ton more decisions he needs to make on a constant basis.

TLDR - he is NOT the Uarch engineer that has is Visio/Graffle opened and "dreams up" chips , he is the guy who runs the dreamers/execution guys organization.
The old schools guys that created this industry , and sure Mr Keller is one of them , are capable and smart and they are directors in a lot of orgs in the business , for him to stick around this long means he knows the game really well , but again , he is the MANAGER , not the dreamer , and for folks who think he can just waltz around and create amazing CPU`s as he zigzag every few years , I can only say from my experience , its far from reality.

Well Punisher , did you get to the end of my wall of text alive , that was a long one.

 

[darkswordsman17]

Well my input was that folks needs to get a grip of what they believe Jim Keller is capable of doing , I tried to avoid replying directly to ppl as it might offend them , sorry if it came rude.

Jim Keller is a big time old school manager , might be a good one even (I left Intel before he arrived) , from my friends small interactions with him @Intel, he was rude and vocal , but in the end he didnt change anything substantial of how they do things , he was not the chief Arch for Zen as so many folks here give him credit for (that was Michael Clark) , although he was the top boss in the division.

Working for over 15 years as a chip designer , i can tell you that , yes brilliant architects are extremely important for your chip success , maybe more then any other position in the org , BUT Uarch is almost always a big team and the innovations and progress they make are spread around , sure we have some folks that are a tier above the rest if you try to give them a score , but they rarely overlap in their expertise , and i would say that Jim Keller is so far from actual design of the chips now days that its not realistic to think about him as the guy who "dreams up" chips , In intel he was the guy that was supposed to put the right ppl in the right places and take out the trash management , which he did to some extent as i know some of the folks he promoted and some that he personally fired , but from a product line ? he never was the guy who had anything to do with defining and designing a new CPU for Intel , could he say "you suck , go get me a better design" after looking at projected performance/power/die size numbers , yes he could say it ,and he did , but thats not what some folks around here seems to think he is doing when he jumps around the companies , they believe he is the one defining those Archs , and just goes to a different company once his project is done or in execution.

The next company he will go to will hire him for his managerial skills and his experience in running a design team and not his technical skills which were not used in years to create anything , is technical skills are used to manage a technical organization , which is important (as i said , he can identify the smart and talented ppl and give them credit and control) , he can use his technical skills to ask the good questions and make the big calls , for example , do we TO in time or we push TO due to late bugs trends not converging , do we go to mass production in our current post silicon state , or we delay and wait for things to converge , do we approve 250 features for the next project , or do my team wont be able to execute in time and a ton more decisions he needs to make on a constant basis.

TLDR - he is NOT the Uarch engineer that has is Visio/Graffle opened and "dreams up" chips , he is the guy who runs the dreamers/execution guys organization.
The old schools guys that created this industry , and sure Mr Keller is one of them , are capable and smart and they are directors in a lot of orgs in the business , for him to stick around this long means he knows the game really well , but again , he is the MANAGER , not the dreamer , and for folks who think he can just waltz around and create amazing CPU`s as he zigzag every few years , I can only say from my experience , its far from reality.

Well Punisher , did you get to the end of my wall of text alive , that was a long one.

While I agree that people seem to give him a lot of credit and misplace what he's doing (I recall him saying he was at Intel to basically develop the next gen interconnect as that's basically become the key for future computing development), I think its likely a straight up lie that he hasn't used his technical expertise in years, and also is nonsense that he's not a "dreamer" since he has to direct "the real dreamers" to make their dreams coherent. And I'm almost certain that requires a level of technical knowledge up there with most of the "real dreamers", if not broader as he has to be able to turn multiple dreams and direct how they'll work together to set chip designs. So sure, he's not doing the architecture stuff himself, or doing the nitty gritty transistor layouts and material engineering, but there's no doubt he'd need technical knowledge to set a path for those things. Plus, its extra baffling since without someone like that, those dreams wouldn't be reality. There's a reason people know Bill Gates/Microsoft and Steve Jobs/Apple but most people don't care or likely even know about Xerox PARC. Same reason that every company that Microsoft buys ends up with little to show for it. Because there's no real plan for directing any dream into coherency (it seems more "assimilate into Microsoft and then we'll figure out how to integrate your cool stuff in like a decade" which obviously doesn't work out, to the point I don't know why Microsoft keeps doing it, unless they're ending up with tons of talented people we just don't get to see the fruits of their labor very often).

Which, lack of the "dream orchestrator" that's clearly been the biggest problem Intel has been having. Its not a recent on either, its just that they had so much power and engineering resources that they could squash competition. Until they couldn't and then they're left with poor management, and workers used to that poor management not keen on that changing.

Crazy thing is, I'd say IBM and Intel are almost total opposites now from where they were in what the 70s and 80s? Although I'd say that they're also quite similar in that IBM seems to also be lacking a conductor.

 

[RasCas99]

While I agree that people seem to give him a lot of credit and misplace what he's doing (I recall him saying he was at Intel to basically develop the next gen interconnect as that's basically become the key for future computing development), I think its likely a straight up lie that he hasn't used his technical expertise in years, and also is nonsense that he's not a "dreamer" since he has to direct "the real dreamers" to make their dreams coherent. And I'm almost certain that requires a level of technical knowledge up there with most of the "real dreamers", if not broader as he has to be able to turn multiple dreams and direct how they'll work together to set chip designs. So sure, he's not doing the architecture stuff himself, or doing the nitty gritty transistor layouts and material engineering, but there's no doubt he'd need technical knowledge to set a path for those things. Plus, its extra baffling since without someone like that, those dreams wouldn't be reality. There's a reason people know Bill Gates/Microsoft and Steve Jobs/Apple but most people don't care or likely even know about Xerox PARC. Same reason that every company that Microsoft buys ends up with little to show for it. Because there's no real plan for directing any dream into coherency (it seems more "assimilate into Microsoft and then we'll figure out how to integrate your cool stuff in like a decade" which obviously doesn't work out, to the point I don't know why Microsoft keeps doing it, unless they're ending up with tons of talented people we just don't get to see the fruits of their labor very often).

Which, lack of the "dream orchestrator" that's clearly been the biggest problem Intel has been having. Its not a recent on either, its just that they had so much power and engineering resources that they could squash competition. Until they couldn't and then they're left with poor management, and workers used to that poor management not keen on that changing.

Crazy thing is, I'd say IBM and Intel are almost total opposites now from where they were in what the 70s and 80s? Although I'd say that they're also quite similar in that IBM seems to also be lacking a conductor.

I am not sure what is the premise of your reply , but I am not arguing an importance of a leader (on the contrary) , and as you can read in my post , he might be an amazing leader , my point and i firmly stand behind it , he is not involved in the details of the Uarch/designs , he will approve them when presented in the concept stage , he might push back against some of their characteristics (as i stated before , Die size , power , performance) , he doesnt have a drawer full of chip Uarch`s that he dreamed of and just goes around implementing them , as some folks said "give him a blank slate and money and he will build something amazing" , no , its not how it works.

As a VP in Intel at his level , the amount of time he has to do ANY technical work is just non existent , sorry but thats the reality of someone at his level and job description , again , it does not mean he cant have a big impact on the org he is running , but it wont be because he is doing any technical work , he is not going through the dozens of projects that are under him and provides technical inputs on how to improve them.

"And I'm almost certain that requires a level of technical knowledge up there with most of the "real dreamers", if not broader as he has to be able to turn multiple dreams and direct how they'll work together to set chip design" - FALSE , just pure false , the top talent Uarch folks are usually a minimum Master degree (tons of Phd`s) and are VERY bright and they know their field like no other , the fact he is a VP of the org does NOT make him smarter or have more knowledge then any of them.

"its extra baffling since without someone like that, those dreams wouldn't be reality" - there is a leader in his position in each company that has a design team (or numerous) and a lot of them are doing great things , he is not some magician , he is a high level manager , good/bad we can argue , he is not above the rest of his peers in my opinion , and finally you cannot say Steve jobs and Bill Gates and his name in the same sentence (unless i misunderstood your point ,then forgive me on that) , I got a BSOD just trying to compile that thought.

We can have a discussion on how a chip is "born" if you like and what kind of inputs he might provide from his place in the org , note that are several levels of a "chip is born".

In my opinion Intel biggest problem (aside from the process , which is a huge issue) is that they are not the top dogs in the game anymore in regards to the best companies to work for , 100% not from money POV nor from location POV and not from a future prospect POV.
If I am a good RCG or a Veteran , Intel is one of the last places i would go to work for , that`s just the reality of the situation today , sure they are trying to do better , but they are still not competitive.

 

[Panino Manino]

If Jim is still dreaming or not doesn't matter, we dream with/for him.

 

[DAPUNISHER]

Well Punisher , did you get to the end of my wall of text alive , that was a long one.

You will find many of us are avid readers, when the subject matter is compelling. I hope you will continue to post about your experiences.

 

[Asterox]

Here is quite long interview with Jim Keller.Interview is from February this year, "time stamp Modular design is kind of interesting."

 

[scineram]

Ian is awaiting your questions for a fresh interview by the way.

 

[DrMrLordX]

@borandi

Not sure if you want to collect questions for Jim Keller in this thread, but if so, here goes!

What do you think about the buyout of ARM Ltd. by nVidia, in light of your work on K12 at AMD?

 

[lobz]

A question from AT forums would go like: what are the 3 biggest dark secrets from the power struggle that has taken place within intel in the past 4 years?
We do act like the whole IT world was a soap opera, sometimes.

 

[trexfromouterspace]

A question from AT forums would go like: what are the 3 biggest dark secrets from the power struggle that has taken place within intel in the past 4 years?

Bob Swan downgraded the coffee in the break areas from Starbucks to Seattle's Best. True story.

Also, there's basically no roadmap stability, but the coffee is way more disturbing.

 

[lobz]

The coffee bit is very amusing
But how is the second thing a secret, let alone a big dark secret?

 

[B-Riz]

Bob Swan downgraded the coffee in the break areas from Starbucks to Seattle's Best. True story.

Also, there's basically no roadmap stability, but the coffee is way more disturbing.

I think we can officially say "Throwing excrement against the wall and seeing what sticks" has been Intel's plan since skipping the wide release of 5 series chips and everything that has happened after that; since 2017 everything has looked like a "ermahgerd what can we do?" reaction to what AMD releases...

 

[trexfromouterspace]

But how is the second thing a secret, let alone a big dark secret?

I mean internal product definition roadmaps. Not "10nm has been delayed so Ice Lake won't release" roadmaps, but internal product definition roadmap changes. Kind of like how you'll see videogames where a year from release they try changing the genre and core mechanics.

I think we can officially say "Throwing excrement against the wall and seeing what sticks" has been Intel's plan since skipping the wide release of 5 series chips and everything that has happened after that; since 2017 everything has looked like a "ermahgerd what can we do?" reaction to what AMD releases...

Actually, no. If they were throwing stuff at the wall to see what sticks, they'd actually have released some interesting products by now. Intel's problem is that a while back they decided to cut down pre-silicon validation efforts in order to be more agile, which ended up having the opposite effect because of -Redacted- course it did.

Now they don't have enough people to validate the stuff they want to throw at the wall, then Swan decided to institute a hiring freeze to stop them from getting more people to validate, meaning that they can't throw stuff at the wall because they can't get it to work.

The level of mismanagement is mind-boggling.

Profanity, even adding **'s to it is not allowed.

Daveybrat
AT Moderator

 

[TransientStudent]

Now they don't have enough people to validate the stuff they want to throw at the wall, then Swan decided to institute a hiring freeze to stop them from getting more people to validate, meaning that they can't throw stuff at the wall because they can't get it to work.

The level of mismanagement is mind-boggling.

Gelsinger has talked about hiring more engineers. Do you think he is fixing the problems, or is it so bad that it will take a long time to right the ship?

 

[moinmoin]

Gelsinger has talked about hiring more engineers. Do you think he is fixing the problems, or is it so bad that it will take a long time to right the ship?

By all accounts Intel doesn't lack good engineers but a corporate structure that fosters productive use of their output instead pitting teams against each other. Chip development from design to launch takes at least 3 years (like Lakefield and Alder Lake are Intel's first direct reaction to the MCM products AMD launched since 2017) and sustainably changing the corporate culture may take quite some time as well.

 

[trexfromouterspace]

Gelsinger has talked about hiring more engineers. Do you think he is fixing the problems, or is it so bad that it will take a long time to right the ship?

"A long time" is somewhat relative. Realistically it'll probably take about 5 years to get everything patched up and see the results.

They've already been making management changes, but it'll take time for them to modernize their design flow.

There's also the issue of rampant poaching and brain drain, since Intel's also been underpaying their guys for a while now. When Microsoft made their SoC team they basically poached half of the people working on Meteor Lake.

Some turnover is normal and healthy for the industry, and I'll always advocate for the engineers to get paid what they're worth. But from Intel's point of view, when a huge chunk of your teams are getting poached by a single competitor, you're doing something wrong.

 

[CluelessOne]

A question from AT forums would go like: what are the 3 biggest dark secrets from the power struggle that has taken place within intel in the past 4 years?
We do act like the whole IT world was a soap opera, sometimes.

So it's Days of Our Lives and the Young and the Restless? 😂😂😂

 

[TransientStudent]

"A long time" is somewhat relative. Realistically it'll probably take about 5 years to get everything patched up and see the results.

They've already been making management changes, but it'll take time for them to modernize their design flow.

There's also the issue of rampant poaching and brain drain, since Intel's also been underpaying their guys for a while now. When Microsoft made their SoC team they basically poached half of the people working on Meteor Lake.

Some turnover is normal and healthy for the industry, and I'll always advocate for the engineers to get paid what they're worth. But from Intel's point of view, when a huge chunk of your teams are getting poached by a single competitor, you're doing something wrong.

This sounds like Intel may have some bumpy years ahead still. Maybe Alder Lake will be great, but their best chance for a reset almost seems to be at the point with 3 nm tiles from TSMC.

 

[TransientStudent]

By all accounts Intel doesn't lack good engineers but a corporate structure that fosters productive use of their output instead pitting teams against each other. Chip development from design to launch takes at least 3 years (like Lakefield and Alder Lake are Intel's first direct reaction to the MCM products AMD launched since 2017) and sustainably changing the corporate culture may take quite some time as well.

Yes, engineering resources alone without a clear plan is not enough.