Jim Keller leaves AMD - AMD claims Zen "on track"

[Mikjul]

Jim Keller was not at AMD when Bulldozer was designed, so we know for a fact that he was not one of the people who made the crappy design choices from Bulldozer.

So this matter much less then.

[as96]

AMD should just drop/suspend CPU department and focus on GPUs, let's be honest they have already both feet inside a grave if they don't do something the rest of the body will follow.

[RagnarokDel]

Maybe he left so he could join the radeon graphics division.

[pixeldensity]

if zen cpu cannot deliver then I will be switching over to a intel cpu in my next upgrade. I hope zen will deliver and get amd back into the cpu game

[juretrn]

Can I have his job?

Or at least be on his team? Pleeeeease?

[Eroda]

he could have just come on board to restructure their CPU line up and give them a new focus maybe he was always planning on leaving all along we wont know until zen drops whether it will be good enough or not but can it be worse than bulldozer  really even if it was just a clone of the K10 arch with more modern tech in it and double the ammount of cores it still would be better than bulldozer

 

 

AMD should just drop/suspend CPU department and focus on GPUs, let's be honest they have already both feet inside a grave if they don't do something the rest of the body will follow.

 

not something they can do after they spent so much on R&D on zen they will just have to ride out the first generation @ the very least if they were going to cancel it

[MageTank]

Hmmm, though, surely if Zen is going to be good enough to pull AMD back into the game he would have stayed to see it through to completion (granted, he's not solely responsible for its success)? Albeit, I don't know much about his history, so as you say it might just be what he does - being a part of one thing, doing his thing then going off to do it with another. 

 

Just to be cheeky and ask, how come you and airdeano don't fill a full line in a post?

His part of the job is most likely done. He was brought on to redesign it, and tasked with getting a competent team together to handle the development of Zen. According to these articles, he achieved that. Also, if you remember back at K8 days, he left right at the exact same moment. After design phases, right before launch. It's his pattern. Keeping him on at this point would probably be too much of a financial burden, and AMD got what they could out of him.

 

The only thing we have to do now, is wait to see if it was worth it, and if it will work. If their promise of 40% IPC over Excavator was true, and it puts them on par with even Haswell, it will be a great thing. Skylake in and of itself is not much faster than Haswell, so having a consumer chip with 4-8 cores and 8-16 threads to compete with Z97 and X99 at the same would be a great launching point for AMD. The only issue is Zen has been delayed, and will likely be released around the same time as Kaby Lake, which leaves it in a pretty bad position in terms of potential market share. 

 

Either way, consider Zen their "All-In" hand. If it fails, this will mark a terrible fate for AMD.

[SeriousDad69]

The Zen chips are pretty much done at this point because it takes many months after the first prototypes to ramp up production and iron out the last bugs and such.

Lol everyone's always doom and gloom when it comes to AMD. Their GPUs are competitive with Nvidia's in several of the most important price points and with DX12 and Vulkan coming out the need for strong single threaded performance should go down giving all those old 6 and 8 core FX processor new life. I like AMD because at least they're interesting, like a wild card. Intel and Nvidia are so boring. I guess I'm just a sucker for an underdog story.

[Eroda]

His part of the job is most likely done. He was brought on to redesign it, and tasked with getting a competent team together to handle the development of Zen. According to these articles, he achieved that. Also, if you remember back at K8 days, he left right at the exact same moment. After design phases, right before launch. It's his pattern. Keeping him on at this point would probably be too much of a financial burden, and AMD got what they could out of him.

 

The only thing we have to do now, is wait to see if it was worth it, and if it will work. If their promise of 40% IPC over Excavator was true, and it puts them on par with even Haswell, it will be a great thing. Skylake in and of itself is not much faster than Haswell, so having a consumer chip with 4-8 cores and 8-16 threads to compete with Z97 and X99 at the same would be a great launching point for AMD. The only issue is Zen has been delayed, and will likely be released around the same time as Kaby Lake, which leaves it in a pretty bad position in terms of potential market share. 

 

Either way, consider Zen their "All-In" hand. If it fails, this will mark a terrible fate for AMD.

 

id be worried if it was going up against cannonlake  but that is also delayed so maybe it might go ok, i also dont think it was delayed alot i always expected zen q3/4 2016 if it is out q1/2 2017 its not that bad

[ltguy]

 Zen is probably mostly done by now.

 

I personally don't see this as a huge problem, due to the fact that AMD would not let talent like this leave unless zen was already mostly finished. 

This make sense. What chip manufacture is working on their next time chip as soon as the finish the current one? (sarcasm)  I sure hope amd doesn't need a chip to follow Zen to stay competitive.

[HalGameGuru]

This make sense. What chip manufacture is working on their next time chip as soon as the finish the current one? (sarcasm)  I sure hope amd doesn't need a chip to follow Zen to stay competitive.

The team Keller left behind should be able to handle iterative follow ups. Keller was there for the trailblazing. They'll be able to implement emergent improvements, and node specific improvements in his absence.

 

 

Edit: 420th post, smoke'em if you got'em.

[Eroda]

The team Keller left behind should be able to handle iterative follow ups. Keller was there for the trailblazing. They'll be able to implement emergent improvements, and node specific improvements in his absence.

 

 

Edit: 420th post, smoke'em if you got'em.

 

Blaze on man...

 

and yeah Zen is their new go to arch that they will be using for the next X years just like intels "Core" arch ie they are reinventing the wheel with each new generation of CPU

[Mahigan]

He's a brilliant engineer. This makes my eyebrow raise.

[Prysin]

He's a brilliant engineer. This makes my eyebrow raise.

and he's done this before.... Although, perhaps not BEFORE the product was launched.

 

perhaps this is a sign of the product is in such a shape that it simply does not need his guidance anymore, leaving him to look for greener pastures....

[crystal6tak]

His part of the job is most likely done. He was brought on to redesign it, and tasked with getting a competent team together to handle the development of Zen. According to these articles, he achieved that. Also, if you remember back at K8 days, he left right at the exact same moment. After design phases, right before launch. It's his pattern. Keeping him on at this point would probably be too much of a financial burden, and AMD got what they could out of him.

Ah I see, that makes much more sense now. And not what I thought which was him just bailing a sinking ship.

[Mahigan]

Ah I see, that makes much more sense now. And not what I thought which was him just bailing a sinking ship.

 

Zen is finalized: http://www.extremetech.com/computing/205086-amd-details-zen-k12-cpu-cores-confirms-next-gen-memory-support

My eyebrow is raised because I'm quite certain someone is about to purchase AMD... I'm thinking Keller didn't want to work with these folks.

[ltguy]

Intel is working on their next two generations of chips before they even finish the current one. The fanboyism is real. Expect another 5 years between processor refreshes again. AMD has to bring in help to design the chip, they have someone else fab them, what exactly do they do on their own now?

[escape093]

Keller saw Zen through like 80% of its development. They hardly need him there to babysit the rest of the guys at AMD.

[Mahigan]

 

Intel is working on their next two generations of chips before they even finish the current one. The fanboyism is real. Expect another 5 years between processor refreshes again. AMD has to bring in help to design the chip, they have someone else fab them, what exactly do they do on their own now?

 

 

Jim Keller was brought in, from DEC, to work on the K7 and K8. He then left to join a few ventures (most notably Apple). He's built router SOCs as well as ARM SOCs. He tends to leave once the work is done.

[MageTank]

and he's done this before.... Although, perhaps not BEFORE the product was launched.

 

He left AMD back in 1999, right before K8 came out, and he was the lead designer on that. Pretty sure he is mostly used as a paid consultant on these things. He comes in, enlightens people on how things should be done, ensures a team can get the job done, then gets his check and moves on. While he is not the god people are making him out to be, he is the man to get this ball into motion. If his design puts AMD on par with Haswell, then it was a success. AMD will then have a solid platform to improve upon, something they have been lacking since the Phenom days. 

 

They can then adopt Intel's tick-tock style, releasing just enough of an improvement to keep people wanting more, while trying to stay competitive. Again, this all hinges upon whether Zen is capable of competing with Haswell/Skylake. It also depends on if Kaby Lake will be enough of a successor to be an enticing offer as well. AMD has a very narrow path to walk in, and its getting smaller as time passes. This is the all or nothing moment for them, and it's almost time to deliver.

[alpenwasser]

This make sense. What chip manufacture is working on their next time chip as soon as the finish the current one? (sarcasm)  I sure hope amd doesn't need a chip to follow Zen to stay competitive.

 

 

The team Keller left behind should be able to handle iterative follow ups. Keller was there for the trailblazing. They'll be able to implement emergent improvements, and node specific improvements in his absence.

 

 

Edit: 420th post, smoke'em if you got'em.

 

Yeah, based on his previous behavior, I kinda get the impression that he's not too interested

in iterative designs, he's more the guy who does the first design when a new architecture gets

developed, then leaves the smaller iterative steps to other folks.

 

My eyebrow is raised because I'm quite certain someone is about to purchase AMD... I'm thinking Keller didn't want to work with these folks.

I could see AMD splitting up into a CPU and GPU section again (though considering how tightly

they're integrating their stuff on their APUs, I would be kinda surprised even about that),

but I'm kinda skeptical on another company buying AMD, mainly due to the cross-licencing

agreement between them and Intel:

 

[...] Intel hereby grants to AMD a non-exclusive, non-transferable [redacted] worldwide license, without the right to sublicense, under Intel's [...]

Source is the redacted version of their cross-licensing agreement, found here: http://corporate.findlaw.com/contracts/operations/patent-cross-license-agreement-advanced-micro-devices-inc-and.html

The tricky thing I see in there is the "non-transferable" part, which could end up meaning

that a potential buyer of AMD would need to renegotiate such an agreement with Intel (depending

on whether a buyout of AMD counts as a transfer or not, opinions on the web seem to be divided

on that one, so I'll just raise the question without providing an answer), potentially causing

significant headaches for anyone interested in buying out the CPU part of AMD.

And this doesn't even take into account whether such patents could be licensed to certain

foreign companies, such as Samsung, or whether the US government would prohibit that (let

alone AMD's own patents then being owned by a foreign entity). If so, then only a US company

would be allowed to buy AMD, reducing chances of a takeover even further.

I'm not saying it's impossible, but going by the info I can find, call me extremely skeptical

until an actual official announcement is made.

[Mahigan]

 

 

 

Yeah, based on his previous behavior, I kinda get the impression that he's not too interested

in iterative designs, he's more the guy who does the first design when a new architecture gets

developed, then leaves the smaller iterative steps to other folks.

 

I could see AMD splitting up into a CPU and GPU section again (though considering how tightly

they're integrating their stuff on their APUs, I would be kinda surprised even about that),

but I'm kinda skeptical on another company buying AMD, mainly due to the cross-licencing

agreement between them and Intel:

 

Source is the redacted version of their cross-licensing agreement, found here: http://corporate.findlaw.com/contracts/operations/patent-cross-license-agreement-advanced-micro-devices-inc-and.html

The tricky thing I see in there is the "non-transferable" part, which could end up meaning

that a potential buyer of AMD would need to renegotiate such an agreement with Intel (depending

on whether a buyout of AMD counts as a transfer or not, opinions on the web seem to be divided

on that one, so I'll just raise the question without providing an answer), potentially causing

significant headaches for anyone interested in buying out the CPU part of AMD.

And this doesn't even take into account whether such patents could be licensed to certain

foreign companies, such as Samsung, or whether the US government would prohibit that (let

alone AMD's own patents then being owned by a foreign entity). If so, then only a US company

would be allowed to buy AMD, reducing chances of a takeover even further.

I'm not saying it's impossible, but going by the info I can find, call me extremely skeptical

until an actual official announcement is made.

 

Hypothetically speaking. If Microsoft were to purchase AMD... they already have most of the x86 OS marketshare, through the Windows OS, and would also add x86-64 tech license to their portfolio.

This would give them quite a bit of leverage over Intel. As there are no competitive x86 processor making companies, Intel might be obligated to provide an x86 license to Microsoft for anti-trust reasons (oh the irony).

This also comes to mind: http://www.forbes.com/sites/patrickmoorhead/2013/06/26/the-real-reasons-microsoft-and-sony-chose-amd-for-consoles/

[alpenwasser]

Hypothetically speaking. If Microsoft were to purchase AMD... they already have most of the x86 OS marketshare, through the Windows OS, and would also add x86-64 tech license to their portfolio.

This would give them quite a bit of leverage over Intel. As there are no competitive x86 processor making companies, Intel might be obligated to provide an x86 license to Microsoft for anti-trust reasons (oh the irony).

Yeah, MS is probably one of the few companies who might be on the list of potential buyers.

But I honestly don't know enough about their business plans and general strategy to know how

much sense this would make for them, or if the potential headaches would be worth it. They

might also run into regulatory issues due to their dominating position in the desktop OS market

which could make buying AMD less attractive.

And yeah, I think Intel would probably be required to license their stuff to whoever buys

AMD (I actually think they were required to license to AMD as well, but I haven't looked this

up, so don't quote me on that), but that doesn't automatically mean that such an agreement

would be easy to reach.

As said, bottom line for me at the moment: It's possible, but I'm skeptical and will continue

to be until I see anything official.

[Agosto]

RIP in pepperoni AMD

[soaringchicken]

I keep having the horrible feeling that this is going to be the last hoorah for AMD CPU's, I hope not.....