Ryzen Architect Jim Keller Joins Intel

[DELETED_195422]

Link

Jim Keller, the VLSI guru who led the team behind AMD's spectacular comeback in the x86 processor market with "Zen," has reportedly quit his job at Tesla to join AMD's bête noire, Intel. Following his work on "Zen," Keller had joined Tesla to work on self-driving car hardware. Keller joins Raja Koduri at Intel, the other big former-AMD name, who led Radeon Technologies Group (RTG).

PC Perspective comments that big names like Keller and Koduri joining Intel could provide clues as to Intel's current state and the direction it's heading in. The company appears to be in a state of shake-up from a decade of complacency and lethargy in its core business. Koduri could be putting together a team of people familiar to him for a new clean-slate project. The last time Intel had a clean slate was ten years ago, with "Nehalem."

 

[voidshatter]

I still think Sandy Bridge was a much more successful milestone than Nehalem.

 

[james_2k]

next you'll be saying @humbug has joined intel too!

madness!

 

[MrRobot]

 

[Rainmaker]

next you'll be saying @humbug has joined intel too!

madness!

Haha! This is interesting news and should yield some nice results in a few years' time. I tell you what though, Messrs Keller & Koduri must be making some moolah... Good on them. Now bring me better CPUs!

 

[Orcvader]

Would take a few years before his contributions crops up though. An Intel Keller designed CPU vs a future AMD design based on Keller's work... Same could be same with the GPU side as well with Koduri... Is Intel becoming AMD inside?

 

[DragonQ]

I still think Sandy Bridge was a much more successful milestone than Nehalem.

Sandy Bridge was their "Zen 2", the first round of improvements to the initial design. Nehalem was a new base (a la Zen), tested in the enthusiast market with Bloomfield (Ryzen 7), then released to the masses as Lynnfield (Ryzen 3 & 5).

Probably bad news for AMD but Intel had to do something: almost zero IPC improvement since 2013, running out of GHz headroom, and long standing issues with 10 nm.

 

[sideways14a]

Tesla must have got up his nose a bit lol.

 

[SiDeards73]

Id be more excited about Keller going to Intel than Raja, Raja is a trainwreck waiting to happen to be honest, and he lies, quite a lot. Too much time spent in the limelight, not enough time spent trying to deliver competitive products, good riddance from AMD tbh.

Keller on the other hand is a man of letting his work do the talking for him.

Interesting times ahead.

 

[4K8KW10]

Zen is not an impressive architecture at all. Yes, it fixed the problems of Bulldozer but nothing special at all. It still lags slightly behind Skylake, Kaby Lake...
So, this news is not worrisome at all.

 

[sideways14a]

lol whut?

 

[MrRobot]

AMD need to ramp up core count fast, 6 core CCX if possible @7nm, beyond 7nm, 8 core CCX.
The small glimmer of hope for AMD, may turn out to be just a dream.

Gonna end up having to pay £600+ for a good Intel CPU, like having to pay £600+ for a good nVidia GPU in 5-10 years time.

 

[DragonQ]

Zen is not an impressive architecture at all. Yes, it fixed the problems of Bulldozer but nothing special at all. It still lags slightly behind Skylake, Kaby Lake...
So, this news is not worrisome at all.

Zen is nothing like Skylake though, it is a modular design that is scalable to 32 cores (currently), is pretty power efficient, and is presumably a lot cheaper to manufacture than Intel's chips. It's not all about raw power. And it doesn't (as far as we know) have an inherently insecure base (Meltdown & Spectre).

 

[MrRobot]

.

 

[ltron]

Wow, I didn't see that coming. This could be bad news for AMD's continued revival as Keller knows Ryzen inside out and will be given many resources to counter it. Intel is a much bigger and richer company after all.

 

[4K8KW10]

Wow, I didn't see that coming. This could be bad news for AMD's continued revival as Keller knows Ryzen inside out and will be given many resources to counter it. Intel is a much bigger and richer company after all.

In December 2016, a news appeared that Intel started designing a new architecture succeeding Core. https://wccftech.com/intel-developing-new-x86-uarch-succeed-core-generation/

I think Intel doesn't need Keller at all. Since they designed the leading x86 without him anyways. And when Keller worked for AMD, he still couldn't deliver something really better.

https://wccftech.com/intel-developing-new-x86-uarch-succeed-core-generation/

The Intel ‘Core’ legacy will come to an end with Tiger Lake in 2019 – To be succeeded by a lean and mean approach to x86
So when they say new architecture, what exactly does that mean. We talk about architectures a lot as far as GPUs go but the fact of the matter is that the architecture of a CPU works in a slightly different way. Where GPUs have historically freely shifted architectures on a very significant scale, Intel really hasn’t shifted from its primary one in ages. As the inventor of the x86 architecture, it has a cross-licensing agreement with AMD to utilize the x86_64 extension of the same architecture (which is basically the 64-bit implementation of x86). It’s a funny state of affairs since Intel cannot offer 64-bit compatibility without licensing x86_64 from AMD and the latter cannot ship 64-bit processors unless it has licensed the base x86 from Intel.

All this time, and through the generations, however, Intel processors have remained 100% backward compatible with all previous iterations. The same basic architecture expanded over and over with new features with time. Every new Intel architecture that we talk about on here (or every tock of Intel’s PAO cadence) is essentially the same underlying x86 architecture expanded with new features with every iteration. For the first time, however, and if this rumor turns out to be true, things might actually change. Intel might introduce an x86 architecture that we can consider to be truly different from what we have seen so far. The reason? It will be a lean mean, x86-on a diet where backward compatibility is no longer assured. Intel’s PAO cadence) is essentially the same underlying x86 architecture expanded with new features with every iteration.

The reason for doing so is simple, by no longer guaranteeing backward compatibility, Intel could save precious die space by removing the hardware for legacy SIMDs and other legacy features. The result would be a much leaner and more efficient x86 architecture that can deliver a bigger bang for lower resources on-die. According to the source, the architecture will be used to replace the current Core series on the Desktop and Enterprise market. It is possible that Intel shifts the 100% backward compatible x86 towards the server side where legacy operations might actually be required on a continuous basis.

 

[JediFragger]

Gonna end up having to pay £600+ for a good Intel CPU

Tbf, it's been proven that a good cpu platform (like 2700k etc) can last a good 7+ years so it's not all bad.

 

[Boomstick777]

This is fine, AMD can just re hire him after his stint with Intel xD

 

[DragonQ]

If they do end up breaking backwards compatibility then they are clearly playing the long game. In the short term, it'll be a disaster and they'll probably have to sell existing chips alongside it. They'll be hoping that in the longer term the (potential) performance gains from starting from scratch will entice people to support the new architecture.

However, the last time they tried it with IA-64, it failed spectacularly.

 

[Boomstick777]

Makes you think, with Keller joining Intel. Intel really are worried about Ryzen. AMD woke Intel up. Now just need to do the same with DGPU, although the evil green empire is trying to stop them at every turn with dirty business tactics xD Might be harder than ever for AMD to compete in GPU space.