Jim Keller leaves AMD

[Techhog]

The god himself who was to singlehandedly to save AMD with Zen and K12?

Says a lot about those products.

... Ugh.

Ignoring comments from Intel investors/shills, I wonder what this means for the future of the company and the industry.

 

[Ajay]

His role as chief architect on Zen would have been over about two years ago. What was he doing all this time?

He was seeing the project thru, probably to first silicon. After that it's just bug fixes - respins that the team he built can handle. Allot of the weight falls on GFL now in ramping up HVM (at competitive clocks). I doubt Keller and his team screwed up as was the case with Bulldozer (AMD brought him in because of his ability to deliver).

As to exactly why he left - I imagine we'll find out soon enough.

 

[Ajay]

Lol. If this was planned there would have been a succession plan in a press release.

Are there any statements from him like "I love AMD, they are on a great course and now is the right time to try something new"? Of course there isn't, because when you are terminated without notice there isn't time.

Possible if first silicon came in and it was a bust. The other possibility is he was getting bored with the minutia and gave a relatively short notice. Why somebody from within the team wasn't promoted is really puzzling.

 

[DarkKnightDude]

Maybe the impact on Zen would be minimal? Didn't they tape it out back in July?

Seems like Keller has a habit of doing this. He left after K7 wrapped up in 1999, and then he left Apple after A4/A5 was done.

 

[desprado]

People dont seem to be aware of how big a deal Jim leaving is.
He would have controlled individial teams that would have worked on their specific areas that they were strongest. Zen was his design, and he is probably the only one that has an in depth understanding of how the whole CPU goes together. He would have dismissed ideas that wouldnt have helped with its speed, and encouraged others to come up with what he needed. He wasnt just an outstanding engineer, he was an outstanding manager.
Now hes gone, they will have a leader that doesnt know zen like the back of his hand. Things will either become a mess very quickly, or they will take a long time to implement, and will probably implement good and bad, by which time Intel will leapfrog them again.
I have seen this happen many times over the years, and it is about to happen again!

 

[Idontcare]

Lol. If this was planned there would have been a succession plan in a press release.

Are there any statements from him like "I love AMD, they are on a great course and now is the right time to try something new"? Of course there isn't, because when you are terminated without notice there isn't time.

If Keller's departure was planned and executed by Keller of his own volition then it would have been communicated to management well in advance of today.

In this industry, anyone in an org chart position at Keller's level would have communicated to his chain of command his intentions of leaving the company at least 4-6 weeks in advance. Anything less would be absolutely unprofessional by any measure of a business standard.

The fact that AMD had little option but to fill the vacant with the CTO, versus taking the opportunity to promote one of Keller's direct-reports in advance of Keller's departure says a lot.

Everything here points to this being a top-down decision. The decision makers could not coordinate the promotion of Keller's subordinate in advance of communicating to Keller that he was about to be departing the company, businesses cannot and do not function that way for a good reason.

So I'm with Phynaz on this one. I doubt Keller knew he was leaving the company when he woke up that morning. It was probably just as much news to him as it was to the rest of his organization. But it probably was not news to Mark Papermaster or Lisa Su as they were probably involved in the decision to boot Keller.

The question to ask is why was Keller booted? It could range from something professionally egregious to simple insubordination or an act of defiance that was grounds for immediate termination (which never happens at that level of management, but he would be told he either resigns on the spot or will be terminated thereafter).

It would be odd if the decision was in any way related to Zen. You don't fire chief architects on the basis of timeline slips or failure to meet performance targets with first silicon samples. So I really doubt this change is an indicator of Zen's performance or timeline.

But, had Keller decided to quit, he'd have given AMD his 4 or 6 week notice (it is not a 2 week notice at his level in the org chart) and the press release itself would have noted that Keller was staying on for a period of time to assist in the transition from himself to whoever was being appointed to the position.

If he really did just up and quit the company cold turkey with no advance notice, well that scenario would then say very dark things about AMD (in which case, questions about Zen would then become germane to his departure).

 

[Azix]

I wouldn't make too much of them not putting someone in his place. Even with due notice it may not be smart to hire someone for a job that is no longer essential and at the salary level required. Zen is done and he actually stayed on longer than he had to I think. Compared to k8 where he left 3 years before it came out.

 

[dark zero]

His role as chief architect on Zen would have been over about two years ago. What was he doing all this time?

Maybe the Zen APU...
and I don't be surprised if the responsible of the dissaster is NOT Keller or even Zen.... the real one might be GloFo and TMSC due NOT giving BIG cores like Intel at 14 nm, let alone 20 nm...

Definately AMD without their fabs are paying them off HARD.

 

[Sonikku]

Like rats off a sinking ship.

 

[guskline]

Boy AMD gets a little good PR from Async and DX12 and then BANG Jim Keller gone. The PR machines will be spinning now and I really doubt it will be positive.

 

[nenforcer]

Well he was only at AMD for 1 year from 1998-1999 where he created K8 and X86-64 ISA.

This time he was at AMD for 3 full years from 2012-2015. AMD also had to assure the investors that this in no way affects their roadmap for 2016.

I'm confident Zen will deliver its promised 40% IPC improvements and keep AMD piggybacking off of Intel in the x86 CPU market.

 

[Arachnotronic]

Well he was only at AMD for 1 year from 1998-1999 where he created K8 and X86-64 ISA.

This time he was at AMD for 3 full years from 2012-2015. AMD also had to assure the investors that this in no way affects their roadmap for 2016.

I'm confident Zen will deliver its promised 40% IPC improvements and keep AMD piggybacking off of Intel in the x86 CPU market.

Would 40% IPC improvements at 2.5GHz - 3GHz really make AMD competitive in desktops?

 

[Azix]

Would 40% IPC improvements at 2.5GHz - 3GHz really make AMD competitive in desktops?

40% on 8 or more cores? yes.

 

[jpiniero]

It would be odd if the decision was in any way related to Zen. You don't fire chief architects on the basis of timeline slips or failure to meet performance targets with first silicon samples. So I really doubt this change is an indicator of Zen's performance or timeline.

AMD's kind of at the point where there has to be pro-"Shutdown AMD" forces at work. A bad first silicon sample could be enough to get the plug pulled. Remember that Zen would have to be competitive on perf/watt if it's to be of any consequence.

 

[raghu78]

If Keller's departure was planned and executed by Keller of his own volition then it would have been communicated to management well in advance of today. In this industry, anyone in an org chart position at Keller's level would have communicated to his chain of command his intentions of leaving the company at least 4-6 weeks in advance. Anything less would be absolutely unprofessional by any measure of a business standard.

yeah planned departures are communicated well in advance. I would say 3-6 months. In cases of planned departures the leaving executive also assists in finding his replacement. The general notice period even in lower job profiles at big companies is 3 months.

The fact that AMD had little option but to fill the vacant with the CTO, versus taking the opportunity to promote one of Keller's direct-reports in advance of Keller's departure says a lot.

good point.

Everything here points to this being a top-down decision. The decision makers could not coordinate the promotion of Keller's subordinate in advance of communicating to Keller that he was about to be departing the company, businesses cannot and do not function that way for a good reason.

So I'm with Phynaz on this one. I doubt Keller knew he was leaving the company when he woke up that morning. It was probably just as much news to him as it was to the rest of his organization. But it probably was not news to Mark Papermaster or Lisa Su as they were probably involved in the decision to boot Keller.

yeah. this is a decision from Lisa and does not bode well for AMD. You don't fire the guy who you brought in to design a new CPU architecture and bring you back into competition with Intel.

The question to ask is why was Keller booted? It could range from something professionally egregious to simple insubordination or an act of defiance that was grounds for immediate termination (which never happens at that level of management, but he would be told he either resigns on the spot or will be terminated thereafter).

It would be odd if the decision was in any way related to Zen. You don't fire chief architects on the basis of timeline slips or failure to meet performance targets with first silicon samples. So I really doubt this change is an indicator of Zen's performance or timeline.

But, had Keller decided to quit, he'd have given AMD his 4 or 6 week notice (it is not a 2 week notice at his level in the org chart) and the press release itself would have noted that Keller was staying on for a period of time to assist in the transition from himself to whoever was being appointed to the position.

If he really did just up and quit the company cold turkey with no advance notice, well that scenario would then say very dark things about AMD (in which case, questions about Zen would then become germane to his departure).

Right now for me the PR release reeks of damage control and things are looking so frighteningly bad that I fear AMD might not survive till Zen's release.

"Jim’s departure is not expected to impact our public product or technology roadmaps, and we remain on track for “Zen” sampling in 2016 with first full year of revenue in 2017."

The fact that this happened before AMD's quiet period is also disturbing. Normally you want to question management on these kind of events. So its expected that this kind of news be released on a earnings call or just before it. Somehow I have very bad vibes about the exit of Jim Keller.

 

[raghu78]

Would 40% IPC improvements at 2.5GHz - 3GHz really make AMD competitive in desktops?

The lesson we all learnt from Bulldozer is never take anything what AMD says seriously. Doubt them and believe only the final product when reviewed by the tech press. Here the situation is so bad that we have to doubt if AMD will even be around as a company to ship Zen.

 

[Phynaz]

"Jim’s departure is not expected to impact our public product or technology roadmaps, and we remain on track for “Zen” sampling in 2016 with first full year of revenue in 2017."

Looks to me Zen's schedule may have slipped. I haven't paid close attention, when AMD has said 2016 for Zen, did they say only samples?

 

[Arachnotronic]

40% on 8 or more cores? yes.

Summit Ridge goes up only to 8 cores. I think it will be handily outclassed by Skylake-E, which I think will clock better, deliver more perf/clock, include more cores and have more robust integrated I/O (Summit Ridge has PCIe Gen 3 x16 only). If Zen is anywhere near half decent, Intel should be able to price Skylake-E in such a way to win the perf crown.

 

[Arachnotronic]

The lesson we all learnt from Bulldozer is never take anything what AMD says seriously. Doubt them and believe only the final product when reviewed by the tech press. Here the situation is so bad that we have to doubt if AMD will even be around as a company to ship Zen.

Completely agree :thumbsup:

 

[Despoiler]

The same people that are always wrong are still wrong.

http://www.eetimes.com/document.asp?doc_id=1327734

 

[Azix]

Summit Ridge goes up only to 8 cores. I think it will be handily outclassed by Skylake-E, which I think will clock better, deliver more perf/clock, include more cores and have more robust integrated I/O (Summit Ridge has PCIe Gen 3 x16 only). If Zen is anywhere near half decent, Intel should be able to price Skylake-E in such a way to win the perf crown.

Perf crown does not matter much for cpus. It's price/performance. Because cpus are so central, a majority of users are getting systems with them already inside. The companies building these systems aren't wooed by crowns.

I don't get where all this doubt comes from about Zen. It seems only common sense that it will be competitive. They could have minutely modified bulldozer on a smaller process and been competitive. The process change alone would be significant for performance

Intel hasn't exactly been making huge strides. So their mainstream is well within reach. The question is their 6 core up offerings. If things work out we'll see those replace the mainstream in price

This news has no real bearing in Amd at all.what would be strange is if he stayed much longer

 

[jpiniero]

Perf crown does not matter much for cpus. It's price/performance.

In what market? Without an IGP OEM's will ignore it. Without competitive perf/watt they will ignore it in servers. So you are looking at maybe gamers? And that's the best case scenario where GloFo don't screw up at some point. Plus now they are talking about Zen CPU being delayed until 2017. If that's true, when is the APU coming? The end of 2017?

 

[Idontcare]

This news has no real bearing in Amd at all.

What an odd thing to write.

Obviously it has a bearing on AMD. AMD felt that the action of hiring Keller was a big enough impact to merit making sure everyone knew they had done so. And they obviously (given that they did a press release) felt that his departure was also significant enough to merit letting everyone know he had left.

To say it has no real bearing in AMD is the same as saying AMD was overplaying the significance in announcing the hiring of Keller, as well as them over-estimating the relevance of his departure. Those are two assumptions I personally wouldn't make, but they seem implicit in your statement quoted above.

what would be strange is if he stayed much longer

Yeah, cuz the last thing AMD would have wanted is to retain their top-notch CPU architect to create an even more superior CPU to that of Zen for post-2017 sales. :|

That kind of thinking on AMD's (or Keller's) part would have just been strange...

 

[VirtualLarry]

I hate to say it, but "sinking ship" comes to mind. I have high hopes for Zen, and I hope that AMD does get some financial help from the private equity partners rumor.

 

[eton975]

Yeah, cuz the last thing AMD would have wanted is to retain their top-notch CPU architect to create an even more superior CPU to that of Zen for post-2017 sales. :|

That kind of thinking on AMD's (or Keller's) part would have just been strange...

Ehh... I don't really think so... Keller tends to come and go for major, breakthrough designs, and then leave one the chip's been taped out. I mean he left after the (breakthrough) K7/Athlon, but AMD rode on with the evolutionary K8.