Rockstar CPU architect Jim Keller Becomes CTO at Tenstorrent: "The Most Promising Architecture Out There"

KarateBob

Gawd
Joined
Aug 4, 2004
Messages
803

juanrga

2[H]4U
Joined
Feb 22, 2017
Messages
2,801
"Rockstar" is the correct title here, because his contributions to computer science/technology are tiny. It is all smoke and propaganda compared to the true people that has really advanced the field.

Again Anandtech gives an incorrect curriculum for Keller, stealing the contributions made by other engineers to the field, for instance pretending that Keller was the "lead architect" of K7 and K8. No he wasn't.

The K7 was made by a combined pair of teams leaded by Dirk Meyer and Fred Weber, with Meyer being the lead engineer. And the famous K8 was designed by a team leaded by Fred Weber. I have posted the links and the official presentations elsewhere in this forum, and I even discussed this kind of stuff with Ian once in twitter, but again the historical facts are being replaced by hype.

I am not going to repeat everything once more, because it is boring to have to correct the same mistakes for years. I am only re-posting a copy of the MPR cover page mentioning how Dirk Meyer, the lead architect of the K7, described the chip in the official presentation made by AMD at Microprocessor forum:

1b4dda60012dda22ddd84f74ca479b96b714752d770de8d0db.png
 
Last edited:

juanrga

2[H]4U
Joined
Feb 22, 2017
Messages
2,801
Here comes juanrga right on que to correct everyone.
Yeah Jim Keller is just a nobody who didn't do anything useful to every company he touched (especially AMD).
VERY coincidental that when Jim Keller joined AMD the first time, K7 and K8 was made (Arches take years to make).
Keller left and things petered out with phenom and then bulldozer.
Then Keller joined AMD again and now we have Ryzen...

I think its safe to say Jim Keller = success for AMD

AMD did not mention Keller's name during the official presentation of the K7 in the microprocessor forum. The official presentation credits both Dirk Meyer and Fred Weber, with Meyer introduced to the public as the chief architect of the K7, but it seems like you know more than AMD, so there is no point arguing with you.

Keller left AMD before the design of the K8 began, which makes your point completely moot.

Keller returned to AMD to work on the K12 and Skybridge projects. He was hired by Rory Read because AMD engineers were not familiar with the ARM ISA, unlike Keller (recall he worked at Apple). Keller gave talks on K12 and the benefits of the ARM ISA over x86. He also shared anecdotes about how the Skybridge team (made up of x86 engineers) was a bit intimidated by the ARM ISA. Lisa Su was once asked who played the main role during the development of Zen. She was explicitly asked about Keller and she gave no credit to him. She confirmed that all the credit goes to the Zen team, and especially to the Zen lead architect: Mike Clark. But you seem to know more than she does about AMD internals, so there is no point in arguing with you anymore.
 
Last edited:

deruberhanyok

[H]ard|Gawd
Joined
Aug 22, 2004
Messages
1,998
Kept reading the article trying to find out what this exciting new architecture was.

ah, it’s another machine learning thing.

AT sure is posting a lot of press releases and passing them off as news lately. This one has the added bonus of including Keller’s past 40 years of employment history in chart form. Uh, Okay?
 
Joined
Apr 11, 2017
Messages
876
AMD did not mention Keller's name during the official presentation of the K7 in the microprocessor forum. The official presentation credits both Dirk Meyer and Fred Weber, with Meyer introduced to the public as the chief architect of the K7, but it seems like you know more than AMD, so there is no point arguing with you.

Keller left AMD before the design of the K8 began, which makes your point completely moot.

Keller returned to AMD to work on the K12 and Skybridge projects. He was hired by Rory Read because AMD engineers were not familiar with the ARM ISA, unlike Keller (recall he worked at Apple). Keller gave talks on K12 and the benefits of the ARM ISA over x86. He also shared anecdotes about how the Skybridge team (made up of x86 engineers) was a bit intimidated by the ARM ISA. Lisa Su was once asked who played the main role during the development of Zen. She was explicitly asked about Keller and she gave no credit to him. He confirmed that all the credit goes to the Zen team, and especially to the Zen lead architect: Mike Clark. But you seem to know more than she does about AMD internals, so there is no point in arguing with you anymore.
I wonder what Keller worked on then first time at AMD. Also, a possible good reason Keller wasn't mentioned by Lisa Su is because he was no longer at AMD at the time of the interview.

You could be 100% correct that's for sure...
I think a better question is why do all the media outlets get it wrong all the time when the answer is so "obvious"...
 

sram

[H]ard|Gawd
Joined
Jul 30, 2007
Messages
1,404
"Rockstar" is the correct title here, because his contributions to computer science/technology are tiny. It is all smoke and propaganda compared to the true people that has really advanced the field.

Again Anandtech gives an incorrect curriculum for Keller, stealing the contributions made by other engineers to the field, for instance pretending that Keller was the "lead architect" of K7 and K8. No he wasn't.

The K7 was made by a combined pair of teams leaded by Dirk Meyer and Fred Weber, with Meyer being the lead engineer. And the famous K8 was designed by a team leaded by Fred Weber. I have posted the links and the official presentations elsewhere in this forum, and I even discussed this kind of stuff with Ian once in twitter, but again the historical facts are being replaced by hype.

I am not going to repeat everything once more, because it is boring to have to correct the same mistakes for years. I am only re-posting a copy of the MPR cover page mentioning how Dirk Meyer, the lead architect of the K7, described the chip in the official presentation made by AMD at Microprocessor forum:

View attachment 316356
Hmmm, I used to admire the guy! He revolutionize wherever he lands when it comes to CPU architecture. He should be very smart considering he only has a BS degree.

I'm now rethinking all this though ! Are you sure?
 

OutOfPhase

2[H]4U
Joined
May 11, 2005
Messages
3,898
I will just suggest that contributions can be many things. A modern CPU is a wildly complicated beast. Many subcomponents, conflicting ideas, and yet a need for coherency in the final product, else you have icky hacks which make a worse product.
I've worked on many teams like that over the years. Every team made an "ideal" component as far as they were concerned, but they lacked the vision to see there was no way to make the optimal end product from all these theoretically optimal components. Sometimes there needs to be give and take in a given area for the overall success. The Needs of The Many, and all that.

I think this speaks to his value. He's seemingly good at helping teams to gel, organize, mesh and have a coherent vision. It doesn't need to be he that designed everything, or even any sub-component himself. Helping all the groups be an effective team is a Real Thing. The teams he works on seem to have that, so I would speculate that is a big part of his value-add.
 

[Spectre]

[H] Admin
Staff member
Joined
Aug 29, 2004
Messages
17,583
I will just suggest that contributions can be many things. A modern CPU is a wildly complicated beast. Many subcomponents, conflicting ideas, and yet a need for coherency in the final product, else you have icky hacks which make a worse product.
I've worked on many teams like that over the years. Every team made an "ideal" component as far as they were concerned, but they lacked the vision to see there was no way to make the optimal end product from all these theoretically optimal components. Sometimes there needs to be give and take in a given area for the overall success. The Needs of The Many, and all that.

I think this speaks to his value. He's seemingly good at helping teams to gel, organize, mesh and have a coherent vision. It doesn't need to be he that designed everything, or even any sub-component himself. Helping all the groups be an effective team is a Real Thing. The teams he works on seem to have that, so I would speculate that is a big part of his value-add.

In other wrods a good leader and manager. Two things that are hard to find.
 

juanrga

2[H]4U
Joined
Feb 22, 2017
Messages
2,801
Hmmm, I used to admire the guy! He revolutionize wherever he lands when it comes to CPU architecture. He should be very smart considering he only has a BS degree.

I'm now rethinking all this though ! Are you sure?

No one here is debating if he is smart or no, simply mentioning that the "Work experience" table in the AT article is misleading and that the mentioned designs were leaded by someone else.

What really impresses me is Keller's ability to work with different architectures and microarchitectures. He adapts very well and accepts different challenges. Keller has worked with Alpha, x86, ARM, and MIPS ISAs. He has worked from high-performance server-class chips to efficient mobile-class chips.
 
Top