Jim Keller joins Intel

Koduri... meh, not impressed. But Jim Keller is interesting. He was always around whenever AMD had a good idea. You have to wonder how much of them came from him.
 
  • Like
Reactions: N4CR
like this
I read this, initially thought it was trolling. I mean you know, given his past.
But it makes perfect sense given other indicators regarding Intel's state of affairs and overall, it's definitely a good thing because fuck Tesla, i want this man working on CPUs; any CPUs.

Now on a personal, emotional level? This is bad news, lol. Why Jim, why?
(i was kinda hoping.. maybe.. sometime.. he'd come back and help with Ryzen 2s or 3s)
 
Aenra said:
I read this, initially thought it was trolling. I mean you know, given his past.
But it makes perfect sense given other indicators regarding Intel's state of affairs and overall, it's definitely a good thing because fuck Tesla, i want this man working on CPUs; any CPUs.

Now on a personal, emotional level? This is bad news, lol. Why Jim, why?
(i was kinda hoping.. maybe.. sometime.. he'd come back and help with Ryzen 2s or 3s)
Click to expand...

Well it certainly sucks for AMD. But from a competition perspective, it's probably a good thing. Maybe he'll give Intel a kick in the ass.
 
Oh yeah, definitely; as i mention above, a man with such potential, a proper shame to have him working on anything else, so definitely good news.

Unfortunately, i'm sure it wasn't received thus inside AMD's headquarters; and i can sympathise, this is a man with potential; proper, proven potential. This will most likely hurt them in the long run, a lot. Much as i hate to admit it, lol
 
Aenra said:
Oh yeah, definitely; as i mention above, a man with such potential, a proper shame to have him working on anything else, so definitely good news.

Unfortunately, i'm sure it wasn't received thus inside AMD's headquarters; and i can sympathise, this is a man with potential; proper, proven potential. This will most likely hurt them in the long run, a lot. Much as i hate to admit it, lol
Click to expand...

Who has been leading Zen development since Jim left, anyway?
 
DuronBurgerMan said:
Who has been leading Zen development since Jim left, anyway?
Click to expand...

Mark Papermaster, at least until Zen launch day, got a bit confuddled with TRs (they were dubbed a "passion project" that was supposedly "not officially in the roadmap" [bollocks]) and onwards given the later restructuring, so not sure just how much of a direct involvement Lisa Su has now.
 
DuronBurgerMan said:
Well it certainly sucks for AMD. But from a competition perspective, it's probably a good thing. Maybe he'll give Intel a kick in the ass.
Click to expand...

Will be at least 4 years until you see his work on a new chip, but I am certain he will be a big help to Intel.
 
The Intel Dream Team.
bl4BV2Z.jpg
 
Feels like they're all just building their resume with AMD just so they can join Intel lol
 
Looks like AMD caught intel completely off guard with Zen and they don't feel their roadmap is good enough anymore. They'll have at least 4 years before anything comes out of whatever they are changing. AMD will be on their next gen by then. All is not well at intel.
 
Peppercorn said:
Looks like AMD caught intel completely off guard with Zen and they don't feel their roadmap is good enough anymore. They'll have at least 4 years before anything comes out of whatever they are changing. AMD will be on their next gen by then. All is not well at intel.
Click to expand...

No, it isn't. But give Intel some credit for realizing it was time for a shakeup. I'm glad to see action in this market again, finally. So many years of turds from AMD and little minor tuneups from Intel.
 
realworld said:
Feels like they're all just building their resume with AMD just so they can join Intel lol
Click to expand...

nah Keller is a contract engineer. he gets paid to go into a company design something then move on after a few years to a new company, do the same there and move on to the next.
 
DuronBurgerMan said:
Well it certainly sucks for AMD. But from a competition perspective, it's probably a good thing. Maybe he'll give Intel a kick in the ass.
Click to expand...
not like they need one? the U series mobile chips are now quad core but crazy turbo boost making them actually usable. The single core performance is still wrecking AMD and now they have more cores absolutely everywhere. If they get anymore kick AMD won't exist. AMD is still a good 3-5 years behind on power consumption. I don't see any scenario outside of mining and content creation that I'd tell people to buy AMD. That Vega + NUC is such a weird curveball.
 
mkrohn said:
not like they need one? the U series mobile chips are now quad core but crazy turbo boost making them actually usable. The single core performance is still wrecking AMD and now they have more cores absolutely everywhere. If they get anymore kick AMD won't exist. AMD is still a good 3-5 years behind on power consumption. I don't see any scenario outside of mining and content creation that I'd tell people to buy AMD. That Vega + NUC is such a weird curveball.
Click to expand...

Kaby Lake G certainly is a weird curveball. Hades Canyon... because Hell froze over, maybe?

Intel did need a kick in the ass, though. Years of refreshed, slightly faster quads. Since AMD returned to actually giving a shit and making an effort, we've seen so much more movement in the industry from both companies. My point is, perhaps Jim going to Intel means more good shit from Intel - finally.

On power consumption, though, I disagree with you. AMD isn't behind on power consumption per se, so much as it is trying to squeeze every last ounce of performance (from both Vega and Ryzen). If their products operated in the proper power envelope, we'd see a little less performance, but absolutely monstrous levels of power efficiency - especially with Zen (Vega probably less so). But, of course, AMD can't afford to leave performance on the table when it's needed to stay in the high performance race, so we don't see as much of that as we should. The Stilt discussed that after the first Ryzen release last year. AMD pushed past two critical points with power/performance scaling to extract all it could from the design. Zen+ amped that up to the eleven, to the point that overclocking is basically pointless.

With 7nm process next year - supposing this is actually delivered (I have my doubts) we may see a situation where AMD no longer has to push its products to the edge to keep up, and the power consumption and efficiency can shine. If so, yes, Intel may need a kick in the ass. Get that 10nm (probably better than '7nm' process, but whatever) process going instead of stuck in development hell, and maybe finally push past Skylake's IPC.

One can hope, anyway.
 
DuronBurgerMan said:
Kaby Lake G certainly is a weird curveball. Hades Canyon... because Hell froze over, maybe?

Intel did need a kick in the ass, though. Years of refreshed, slightly faster quads. Since AMD returned to actually giving a shit and making an effort, we've seen so much more movement in the industry from both companies. My point is, perhaps Jim going to Intel means more good shit from Intel - finally.

On power consumption, though, I disagree with you. AMD isn't behind on power consumption per se, so much as it is trying to squeeze every last ounce of performance (from both Vega and Ryzen). If their products operated in the proper power envelope, we'd see a little less performance, but absolutely monstrous levels of power efficiency - especially with Zen (Vega probably less so). But, of course, AMD can't afford to leave performance on the table when it's needed to stay in the high performance race, so we don't see as much of that as we should. The Stilt discussed that after the first Ryzen release last year. AMD pushed past two critical points with power/performance scaling to extract all it could from the design. Zen+ amped that up to the eleven, to the point that overclocking is basically pointless.

With 7nm process next year - supposing this is actually delivered (I have my doubts) we may see a situation where AMD no longer has to push its products to the edge to keep up, and the power consumption and efficiency can shine. If so, yes, Intel may need a kick in the ass. Get that 10nm (probably better than '7nm' process, but whatever) process going instead of stuck in development hell, and maybe finally push past Skylake's IPC.

One can hope, anyway.
Click to expand...

Once again. The original tick-tock roadmap was 4-core Skylake (14nm) --> 8-core Icelake (10nm). Then the problems with 10nm did start and Intel had to switch to Tick-tock-optimization-optimization with 14nm+ and now 14nm++...
Don't misinterpret me, Ryzen is a good product, but it is continuously overhyped in media and forums. Processors as the 1800X launched against quad-core Kabylake and 2700X launched against six-core CoffeeLake. We had reviews showing 8-core Zen running circles around 4-core Kabylake on Blender, Handbrake, Cinebench MT, and other multithreaded workloads. Despite the obvious advantage from 'moar cores' (twice more cores at launch) AMD is not kicking Intel really, Ryzen impact on desktop sales was about 3% and only half percentage on datacenter. if Intel had no problem with 10nm and the original roadmap executed flawless, then 1800X had launched against 8-core Icelake and it would lose on everything.

Keller is not a foundry engineer. So he will not solve Intel problems regarding 10nm. Once the yield problems are solved Intel can start releasing everything has maintained in the simulator during last years. Intel cores are more advanced than AMD cores, specially when we add HPC/server workloads that can use AVX 256/512bit. You mention The Stilt; well he has demonstrated how Skylake-X core is 3--4 times faster than Zen core at same clocks in such workloads, because Skylake-X has 512bit SIMD units, whereas Zen has only 128bit.

Keller has joined Intel as Senior Vice President. Keller will be heading the engineering group, with an emphasis on SoC development and integration.
 
juanrga said:
Once again. The original tick-tock roadmap was 4-core Skylake (14nm) --> 8-core Icelake (10nm). Then the problems with 10nm did start and Intel had to switch to Tick-tock-optimization-optimization with 14nm+ and now 14nm++...
Don't misinterpret me, Ryzen is a good product, but it is continuously overhyped in media and forums. Processors as the 1800X launched against quad-core Kabylake and 2700X launched against six-core CoffeeLake. We had reviews showing 8-core Zen running circles around 4-core Kabylake on Blender, Handbrake, Cinebench MT, and other multithreaded workloads. Despite the obvious advantage from 'moar cores' (twice more cores at launch) AMD is not kicking Intel really, Ryzen impact on desktop sales was about 3% and only half percentage on datacenter. if Intel had no problem with 10nm and the original roadmap executed flawless, then 1800X had launched against 8-core Icelake and it would lose on everything.

Keller is not a foundry engineer. So he will not solve Intel problems regarding 10nm. Once the yield problems are solved Intel can start releasing everything has maintained in the simulator during last years. Intel cores are more advanced than AMD cores, specially when we add HPC/server workloads that can use AVX 256/512bit. You mention The Stilt; well he has demonstrated how Skylake-X core is 3--4 times faster than Zen core at same clocks in such workloads, because Skylake-X has 512bit SIMD units, whereas Zen has only 128bit.

Keller has joined Intel as Senior Vice President. Keller will be heading the engineering group, with an emphasis on SoC development and integration.
Click to expand...

So you argument it "AMD's victory in [insert specific workloads] doesn't actually mean anything because they have more of [insert hardware structure], but Intel stomps all over them in [insert specific workloads] because they have more of [insert hardware structure]. That means AMD is crap and Intel is the best ever."

Got it.
 
BrotherMichigan said:
So you argument it "AMD's victory in [insert specific workloads] doesn't actually mean anything because they have more of [insert hardware structure], but Intel stomps all over them in [insert specific workloads] because they have more of [insert hardware structure]. That means AMD is crap and Intel is the best ever."

Got it.
Click to expand...

And all that despite I wrote "Don't misinterpret me, Ryzen is a good product". I cannot imagine which would be your reading of my post if I hadn't written that.
 
Last edited:
It's a bit crass but I wonder how much someone like Jim Keller would get paid. I assume he does fixed term project linked contracts, given I know software engineers on half a million a year and vp's engineering who could retire with 2 years wages, it's got to be a good chunk.

In that specialisation, at that level, you could probably sit your peers round a dining room table and there's some very deep pocketed companies looking to gain advantage.
 
juanrga said:
Don't misinterpret me, Ryzen is a good product, but it is continuously overhyped in media and forums.
Click to expand...

Everything is overhyped, man. It's the Internet. Somebody posts a gif of puppies or something, and it gets hyped.

Processors as the 1800X launched against quad-core Kabylake and 2700X launched against six-core CoffeeLake. We had reviews showing 8-core Zen running circles around 4-core Kabylake on Blender, Handbrake, Cinebench MT, and other multithreaded workloads.
Click to expand...

The real innovation of Zen is the CCX concept which allows cheap core scaling. The cores themselves are kind of "meh". They aren't shitty like Bulldozer, but not impressive either. The CCX concept, OTOH, made "moar cores" super cheap. And that IS worth praising.

Despite the obvious advantage from 'moar cores' (twice more cores at launch) AMD is not kicking Intel really, Ryzen impact on desktop sales was about 3% and only half percentage on datacenter. if Intel had no problem with 10nm and the original roadmap executed flawless, then 1800X had launched against 8-core Icelake and it would lose on everything.
Click to expand...

A bit of semantics here. AMD is not kicking Intel's ass. AMD is kicking Intel *IN* the ass. Intel has been lazy on the CPU side of things for a while. Now they are clearly pushing the envelope again. CoffeeLake got pushed out as early as they could get it out the door. More Skylake-X products hit than we were anticipating - and faster. Intel is hiring innovative people to lead new product development. The service AMD provided to us is reminding Intel that competition still exists, and they can't get lazy.

And of course, if the 1800X launched against an 8 core Icelake it would lose on everything. Like I said, Zen's core design is middling. Not impressive on its own. It is the CCX concept that is impressive, because it allows lower cost core count improvement, such that AMD can offer more cores and/or threads at any given price point they care to compete in.

Keller is not a foundry engineer. So he will not solve Intel problems regarding 10nm. Once the yield problems are solved Intel can start releasing everything has maintained in the simulator during last years. Intel cores are more advanced than AMD cores, specially when we add HPC/server workloads that can use AVX 256/512bit. You mention The Stilt; well he has demonstrated how Skylake-X core is 3--4 times faster than Zen core at same clocks in such workloads, because Skylake-X has 512bit SIMD units, whereas Zen has only 128bit.
Click to expand...

Never said he would. He's an architecture guy. But that means Intel is working on architectural improvements (perhaps Ocean Cove?). That's fucking exciting, dude. I know we won't see the fruits for some years yet, but come on. That's cool. As for AVX, yes we all know Zen's implementation is inferior. I think that was a mistake on AMD's part. But it's not one that (so far) has caused much difficulty for them. AMD kept saying that kind of work should really be offloaded on to a GPU. I *sort of* disagree. Still, the uses still aren't very common. That being said, it's something I have my eye on and hope AMD has addressed with Zen2, or intends to address down the road.
 
I read dream team and thought LeBron James joined the Golden State Warriors or Cristiano Ronaldo joined Barcelona.
 
DuronBurgerMan said:
Everything is overhyped, man. It's the Internet. Somebody posts a gif of puppies or something, and it gets hyped.
Click to expand...

There are levels of hype: from barely above optimistics to crazy nonsense.

DuronBurgerMan said:
The real innovation of Zen is the CCX concept which allows cheap core scaling. The cores themselves are kind of "meh". They aren't shitty like Bulldozer, but not impressive either. The CCX concept, OTOH, made "moar cores" super cheap. And that IS worth praising.
Click to expand...

Quad-core clusters connected with a high-latency interconnect aren't something new. Even the AMD-made consoles have a CPU is built from two quad-core clusters

DuronBurgerMan said:
Intel is hiring innovative people to lead new product development. The service AMD provided to us is reminding Intel that competition still exists, and they can't get lazy.
Click to expand...

AMD isn't the center of the Universe. Koduri was hired to fight Nvidia, for instance.
 
juanrga said:
There are levels of hype: from barely above optimistics to crazy nonsense.



Quad-core clusters connected with a high-latency interconnect aren't something new. Even the AMD-made consoles have a CPU is built from two quad-core clusters



AMD isn't the center of the Universe. Koduri was hired to fight Nvidia, for instance.
Click to expand...

Actually, considering we need competition, AMD is the center of that Universe. Otherwise, we would still have Quad core as mainstream only and single digit improvements for $100's of dollars.

Also no, the consoles do not have 2 quad core clusters on their cpu's, that is completely wrong.
 
not that long ago there were dismissals as to how much of a role Keller actually had with Ryzen, even went to praising others more, but apparently it is a dream team now.
 
ManofGod said:
Actually, considering we need competition, AMD is the center of that Universe. Otherwise, we would still have Quad core as mainstream only and single digit improvements for $100's of dollars.

Also no, the consoles do not have 2 quad core clusters on their cpu's, that is completely wrong.
Click to expand...

he is technically right about the consoles being 2 jaguar CU's which are 4 cores each.
 
juanrga said:
There are levels of hype: from barely above optimistics to crazy nonsense.
Click to expand...

Yeah. So?

Quad-core clusters connected with a high-latency interconnect aren't something new. Even the AMD-made consoles have a CPU is built from two quad-core clusters
Click to expand...

Mmhmmm. None that have been brought to the mainstream CPU market. At least, none that didn't suck elephant farts (Bulldozer was kind of modular like that - sort of - but who cares, it was garbage).



AMD isn't the center of the Universe. Koduri was hired to fight Nvidia, for instance.
Click to expand...

You tell me where I said they were. I don't get you, man. I'm excited to see what Intel can do with some new talent - maybe have a jump like Nehalem and Sandy Bridge again - and you're sitting here talking about 'AMD isn't the center of the Universe.' Lolwut?
 
OrangeKhrush said:
not that long ago there were dismissals as to how much of a role Keller actually had with Ryzen, even went to praising others more, but apparently it is a dream team now.
Click to expand...

Anonymous: "Who had the biggest role in the creation of Ryzen? Was it you? Jim Keller? Someone else?"

Lisa Su: "In terms of the creation of Ryzen, I am really really really PROUD of our team. To build something like Ryzen takes really smart people coming together around a big, audacious goal and the Ryen team did it. The lead architect on Ryzen was a guy named Mike Clark and together with the entire global team, made Ryzen a reality."


Recognizing Keller merits is one thing, inventing he was the lead architect of Zen, K8, K7, Cyclone,... is another thing. Lisa Su confirms the lead architect for Zen was Mike Clark, in another thread I gave the names of the lead architects of K7 and K8.

Keller joined Intel as a senior vice president to lead silicon engineering, with a focus on "system-on-chip (SoC) development and integration". I think that Keller will be a bridge between the CPU teams and the GPU/accelerator teams, because Intel’s chief engineering officer and group president of the Technology, Systems Architecture & Client Group wrote:

"We have embarked on exciting initiatives to fundamentally change the way we build the silicon as we enter the world of heterogeneous process and architectures. Jim joining us will help accelerate this transformation."

DuronBurgerMan said:
I'm excited to see what Intel can do with some new talent - maybe have a jump like Nehalem and Sandy Bridge again - and you're sitting here talking about 'AMD isn't the center of the Universe.' Lolwut?
Click to expand...

Excitation isn't the word I would to use, but I am interested in what Intel will produce.

CPU core microarchitectures did hit both IPC and frequency walls. With enough resources and time, engineers can optimize here and there and obtain a 5% IPC gain or 300MHz extra. That is all. There is no revolutionary architectures in the pipeline, there is no quantum jumps in performance per core coming in next few years.

The only way to substantially accelerate workloads in the mobile/desktop space is by abandoning the traditional x86 paradigm and using accelerators/GPUs for certain applications or regions of code. I guess that must be the reason why Intel's chief engineering officer mentions "heterogeneous process and architectures". Keller's words also seem to confirm this:

The world will be a very different place in the next decade as a result of where computing is headed. I am excited to join the Intel team to build the future of CPUs, GPUs, accelerators and other products for the data-centric computing era.
Click to expand...
 
Last edited:
juanrga said:
Anonymous: "Who had the biggest role in the creation of Ryzen? Was it you? Jim Keller? Someone else?"

Lisa Su: "In terms of the creation of Ryzen, I am really really really PROUD of our team. To build something like Ryzen takes really smart people coming together around a big, audacious goal and the Ryen team did it. The lead architect on Ryzen was a guy named Mike Clark and together with the entire global team, made Ryzen a reality."


Recognizing Keller merits is one thing, inventing he was the lead architect of Zen, K8, K7, Cyclone,... is another thing. Lisa Su confirms the lead architect for Zen was Mike Clark, in another thread I gave the names of the lead architects of K7 and K8.

Keller joined Intel as a senior vice president to lead silicon engineering, with a focus on "system-on-chip (SoC) development and integration". I think that Keller will be a bridge between the CPU teams and the GPU/accelerator teams, because Intel’s chief engineering officer and group president of the Technology, Systems Architecture & Client Group wrote:

"We have embarked on exciting initiatives to fundamentally change the way we build the silicon as we enter the world of heterogeneous process and architectures. Jim joining us will help accelerate this transformation."



Excitation isn't the word I would to use, but I am interested in what Intel will produce.

CPU core microarchitectures did hit both IPC and frequency walls. With enough resources and time, engineers can optimize here and there and obtain a 5% IPC gain or 300MHz extra. That is all. There is no revolutionary architectures in the pipeline, there is no quantum jumps in performance per core coming in next few years.

The only way to substantially accelerate workloads in the mobile/desktop space is by abandoning the traditional x86 paradigm and using accelerators/GPUs for certain applications or regions of code. I guess that must be the reason why Intel's chief engineering officer mentions "heterogeneous process and architectures". Keller's words also seem to confirm this:
Click to expand...


That is the point, it isn't really a dream team as the perceived idea that AMD = Keller is not true, Keller is going to Intel for a purpose and when he has fulfilled that purpose he will be out of there, so I don't expect this to be long term, nor do I see it as being radical. SOC makes sense, Keller was big on the ARM/AMD SOC development so it makes sense for Intel to cash in on that expertise. As for dream team in quantum leap IPC architecture, probably not much will change, Intel will limp the core I architecture to its end.
 
Let see, Intel hired Jim to prevent AMD in rehiring him back once they got the $'s, Intel still wanted to be relevant when AMD next CPU design came out --> :D

Jim talents and drive maybe more akin to Intel's Quantum computing work which would be server based. Just a thought.
 
You must log in or register to reply here.
Back
Top