AMD at Citi Global Technology Conference (Seeking Alpha)

[AtenRa]

Advanced Micro Devices Management presents at Citi Global Technology Conference (Transcript)

Adeline LeeGreat the second question is how would you characterize your potential productivity gains in your businesses versus one year ago versus a quarter ago, is it better, is it worse, is it the same?


Lisa SuYes so productivity, I think we can look at it as both engineering productivity and manufacturing productivity. On the engineering side, again you think about where we are going as a Company, if you looked at AMD a couple of years ago you could say we were X-86 Company, I mean that’s kind of what we did. If you look at the strategies since Rory Read has joined the Company as CEO which has been about two years, you know I joined about 18 to 20 months ago, Mark Papermaster as our CTO also joined within that timeframe. We really focused the Company on how we transitioned from being a pure X86 Company to be really more of a design Company and what mean is focusing on still our core strength which is in the microprocessor area, our microprocessor design and our graphics capability, our core technology IT strengths where we spend a lot of our engineering and R&D dollars but we are able to reuse those components into a bunch of markets, so those components are reused in the PC markets, they are reused in the game console market, we think there is a growth opportunities in the embedded market than servers. And so from a productivity standpoint the idea of how do you reuse engineering sort of IT such that we get more products out is really what we are focused on. So compared to a year ago, I think we are well on our way in that journey. I would say that it’s a multi-year journey and so we are nowhere near done, and I think you will see some of the key hallmarks for us are productivity means to me execution so when we promise a product we get it out on time, the first half of this year we launched an entirely new client roadmap, we are launching also these new game console chip in the second half of this year and you will see more reuse and more capability as we go forward in the 2014 and 2015 timeframe.

 

[SiliconWars]

I read this earlier, there are a couple of interesting points in it.

http://seekingalpha.com/article/167...lobal-technology-conference-transcript?page=5

I think you’ll also see that our products I said Temash was a four watt product I think you’ll see new products coming out from us at much lower power that will support the 7 to 8 inch screen sizes.

And http://seekingalpha.com/article/167...y-conference-transcript?page=7&p=qanda&l=last

we're not very specific and where the various parts are manufactured only to say that we do use both TSMC and GlobalFoundries for the 28 nanometer node across our production.

This probably wraps up Kaveri for GF.

 

[NostaSeronx]

This probably wraps up Kaveri for GF.

It is for both the Xbox One and Playstation 4, Gaming APUs. Not for Kaveri or Volcanic Islands.

Whole quote:

Yes so in terms of how different are the console chips. They are quite different relative to architecture, they use sort of similar IP, so like our Jaguar core and our Radeon graphics but in terms of the architectures you know sort of how they decided to put it together, they really are custom designs. We run them as two completely separate teams. Given the confidential nature of the business, we have to ensure to ensure two very separate teams and so they are really separate chips from that standpoint twofold design efforts. I am sorry your other questions?

As far as the nodes that…

Yes, so in this particular case both products are starting at the 28 nanometer by technology node that’s actually a good thing because if you think about a 28 nanometers are fairly mature technology node, there has been some discussion about the amount of content that’s on these chips and the die sizes and stuff. The way I think about it is when you are designing a game console, you are really designing for a console that’s going to last as I said per seven year cycle, even if it short to five year cycle, so it’s a long cycle and we make sure in this process that these are future-proof designs that they really have good content, I think relative the technology node because it’s 28 nanometer, it is a stable technology node and so we feel good about the yields in the manufacturing around that and relative to where its manufactured I think we're not very specific and where the various parts are manufactured only to say that we do use both TSMC and GlobalFoundries for the 28 nanometer node across our production.

 

[mrmt]

AMD strategy to counter Silvermont: Pray.

 

[Phynaz]

AMD strategy to counter Silvermont: Pray.

I lol'ed.

 

[piesquared]

AMD strategy to counter Silvermont: Pray.

intel strategy to pump Silvermont: exclusively use misleading, biased benchmarks, limit exposure to only 'friendly' reviewers, and invent new power rating schemes to fool customers.

 

[AtenRa]

AMD strategy to counter Silvermont: Pray.

https://i.chzbgr.com/maxW500/3298617856/h3107FEFC/

 

[SiliconWars]

AMD strategy to counter Silvermont: Pray.

Is this the same Silvermont that was supposed to be competing with Jaguar yet is nowhere to be seen while AMD takes large percentages of mobile share from Intel?

I lol'ed.

I fear that says more about your mental age than mrmt's chances of becoming a comedian.


Knock off the personal comments
Markfw900
Anandtech Moderator

 

[Sweepr]

AMD strategy to counter Silvermont: Pray.

When a tablet chip matches your 15W TDP chip in MT floating-point performance and there are server chips with twice the number of cores of that tablet chip (running at similar frequency) with 12W TDP thats pretty much what you can do.

 

[SiliconWars]

When a tablet chip matches your 15W TDP chip in MT floating-point performance and there are server chips with twice the number of cores of that tablet chip (running at similar frequency) with 12W TDP thats pretty much what you can do.

Funnily enough that's not how Intel sees it.

(From the link, if any of you bothered to read it)

http://seekingalpha.com/article/167...lobal-technology-conference-transcript?page=6

Adeline Lee Actually, what’s interesting is that one of Intel’s datacenter briefings they told us that their (...Avoton I guess) product which is something they just unveiled today, that is directly sort of targeted towards AMD's product; I am interested in what’s your response to that.

Lisa Su Should I feel flattered or…

Adeline Lee I think you should.

Seems like Intel is taking it a lot more seriously than you are.

 

[AtenRa]

I dont want to derail the topic but, I wander how many $600+ BayTrail tablets will be sold and how many of this ASUS $250-300 with AMD Temash. :whiste:

 

[mrmt]

When a tablet chip matches your 15W TDP chip in MT floating-point performance and there are server chips with twice the number of cores of that tablet chip (running at similar frequency) with 12W TDP thats pretty much what you can do.

Lisa Su acknowledged that AMD expects a much harder environment after Silvermont is deployed and that was to be expected, but what caught my attention is that they have nothing to show beyond Kabini/Temash and so far we heard nothing of their ARM custom core, while Intel is aggressively pushing Atom roadmap, first with Airmont and then Goldmont one year later and the other ARM companies, Calxeda, Nvidia, Qualcomm, Samsung mainly, are all pushing custom designs for the server markets, exactly the market is aiming with a vanilla ARM core.

As much as 28nm will be low cost there isn't much you can do once you are that outclassed. All in all it's hard to see a sustainable bleeding edge business here, it seems that semi-custom/low margin will be the rule from now on.

 

[Idontcare]

Lisa Su acknowledged that AMD expects a much harder environment after Silvermont is deployed and that was to be expected, but what caught my attention is that they have nothing to show beyond Kabini/Temash and so far we heard nothing of their ARM custom core, while Intel is aggressively pushing Atom roadmap, first with Airmont and then Goldmont one year later and the other ARM companies, Calxeda, Nvidia, Qualcomm, Samsung mainly, are all pushing custom designs for the server markets, exactly the market is aiming with a vanilla ARM core.

As much as 28nm will be low cost there isn't much you can do once you are that outclassed. All in all it's hard to see a sustainable bleeding edge business here, it seems that semi-custom/low margin will be the rule from now on.

Long term roadmaps and the business/product vision that is required to put such roadmaps into action is a luxury that only financially healthy businesses can afford.

Everyone in your list, and rightfully so, has a healthy bottom line and cash flow situation. Not so AMD.

AMD is still very much fighting just to live hand-to-mouth at this stage. That doesn't really give anyone the opportunity to sit in a cubicle dreaming up roadmaps that don't contribute to AMD's bottom-line in the next 90-180 days.

In that environment, when it comes down to the individual who must feel secure in their job and their career path, who wants to find themselves with the job title and job definition that says "I sit around thinking about stuff that won't contribute to AMD's bottom-line for four years, yes I am basically holding my hand up and volunteering myself to make the short-list of people who will be cut in the next round of layoffs".

So why is AMD positioned so differently than the others in your list - quality of personnel who occupy the specific positions that have those responsibilities, and the fact anyone who can't tie their daily work hours directly to some existing revenue-generation product line is basically begging to be laid off in the next 6 months.

 

[SiliconWars]

what caught my attention is that they have nothing to show beyond Kabini/Temash and so far we heard nothing of their ARM custom core

Untrue. On my first post in this thread I copied her comments which clearly points to AMD developing a lower power chip. We've known about Beema and Mullins for months as well and will surely hear more about that during the developer summit in November.

As much as 28nm will be low cost there isn't much you can do once you are that outclassed. All in all it's hard to see a sustainable bleeding edge business here, it seems that semi-custom/low margin will be the rule from now on.

We've seen the anticipated clock speeds of Airmont - at up to 2.7 GHz that's 12.5% higher performance than Silvermont. AMD can beat that on 28nm with a turbo. Graphics are another matter and it's possible that Intel might finally be able to match AMD with a two node lead and dual-channel memory.

The gains are lessening each node, "good enough" stays the same, Intel spends more area on graphics in order to catch up. Nothing will change on 14nm.

 

[mrmt]

Long term roadmaps and the business/product vision that is required to put such roadmaps into action is a luxury that only financially healthy businesses can afford.

I expected her to give a silver lining. It's easy to fathom that AMD will not be able to invest to be bleeding edge on every business they are into now, but I did expect her to be clearer about the roadmaps even if in one or two segments. As you said, the ship is still on fire.

The other issue is the foundry. Kumar was pretty straightforward last november when he answered that all CPU production would move to Globalfoundries and some of the GPUs would, but every time the subject is raised in the Q&A and conferences AMD managements avoid the issue completely.

If this the transition was a done deal and on track, why not disclose a roadmap or even say "this will be manufactured at TSMC, that at Globalfoundries"? The way they behave denotes that things are very fluid in the WSA negotiation, whether it is because of the quality of GLF nodes or because of financial issues it remains to be seen. We are approaching the end of 2013 and they have things agreed until Q114, so we should hear some news until the end of the year.

 

[krumme]

I expected her to give a silver lining. It's easy to fathom that AMD will not be able to invest to be bleeding edge on every business they are into now, but I did expect her to be clearer about the roadmaps even if in one or two segments. As you said, the ship is still on fire.

The other issue is the foundry. Kumar was pretty straightforward last november when he answered that all CPU production would move to Globalfoundries and some of the GPUs would, but every time the subject is raised in the Q&A and conferences AMD managements avoid the issue completely.

If this the transition was a done deal and on track, why not disclose a roadmap or even say "this will be manufactured at TSMC, that at Globalfoundries"? The way they behave denotes that things are very fluid in the WSA negotiation, whether it is because of the quality of GLF nodes or because of financial issues it remains to be seen. We are approaching the end of 2013 and they have things agreed until Q114, so we should hear some news until the end of the year.

There is a roadmap running, and Amd have been more or like in this situation for the most of 30 years. Its absolutely more a permanent state than not.
There is probably multiple reasons the roadmap is not disclosed. Wsa, pressure on gf and mubadala and the derived uncertain consequences for actual execution.
But i think the most obvious reason is the most straight forward. They want to keep it secret for competitive reasons. As they dont want to go directly against Intel anymore, taking that decision looks pretty obvious.

 

[mrmt]

But i think the most obvious reason is the most straight forward. They want to keep it secret for competitive reasons. As they dont want to go directly against Intel anymore, taking that decision looks pretty obvious.

I think the most obvious reason is that they don't have much to show.

 

[Sweepr]

We've seen the anticipated clock speeds of Airmont - at up to 2.7 GHz that's 12.5% higher performance than Silvermont. AMD can beat that on 28nm with a turbo. Graphics are another matter and it's possible that Intel might finally be able to match AMD with a two node lead and dual-channel memory.

There's much more to it than a single number. We dont know if this 2.7GHz refers to: base clock or turbo clock? tablet line, notebooks or desktops? 2.7GHz Turbo and something like 2.0-2.2GHz base clock would be a significant improvement vs Z3770's 1.46-2.4GHz that already matches a 1.5GHz quad-core AMD Jaguar @ Cinebench. The graphics side will get a very significant boost and I hope AMD has something better than 128 GCN cores @ 200-300MHz (<10W TDP) or they will get outclassed real fast.

Not to mention the same DigiTimes rumour you picked that 2.7GHz number from also mentions that Intel's next generation Goldmont microarchitecture + Gen 9 graphics could be launching in late 2014. Unlikely in my opinion, but amazing if true.

 

[krumme]

I think the most obvious reason is that they don't have much to show.

It doesnt matter if there is "much" to show or not whatever "much" is in your interpretation. They have obviously chosen not to disclose it.
There can be many reasons but that reason is just the most unlikely.

 

[Abwx]

Some think that after Kabini was released AMD enginers
were given inderterminate vacations.?..

 

[mrmt]

It doesnt matter if there is "much" to show or not whatever "much" is in your interpretation. They have obviously chosen not to disclose it.
There can be many reasons but that reason is just the most unlikely.

Yeah, that's pretty much my interpretation. By much I mean something relevant to the consumer market, something that can push boundaries in one way or another, and their silence doesn't seem to mean they are working on something like that.

They might be working on some new "semi-custom" design that will pay the bills for another couple of years but that will be a yawn from the consumer POV. Given that their forecast is for embedded/semicustom to have a bigger share in their revenues, I wouldn't be surprised if that's the case.

 

[pablo87]

AMD in simple terms:

ASSETS
- GPU technology and IP
- Bobcat and successors, IP integration, integration with TSMC
- contracts with Microsoft and Sony

LIABILITIES
- WSA
- Big Core
- PC industry shrinking
- unstable work environment
- trying to keep up with Intel across the board

Its no coincidence that the Console designs were won and got done using AMD's assets - GPU, Bobcat and TSMC.

The WSA killed the business. It forced AMD to continue focusing on big core to pay for the WSA when they should have put their efforts on Bobcat and its successors. And it was and is so financially punitive that to survive AMD resorted to continuous layoffs, causing the unstable work environment which in turns causes the cloudy roadmap...

Mubadala ATIC should have protected their AMD investment and let Gloflo sink or swim.

 

[SiliconWars]

AMD would still be making big cores regardless of the WSA. They have the wages of thousands of big core guys to pay and big cores are still the majority of the companies revenues.

Throwing money and people at problems is no guarantee of success either. Intel got none with Larrabee, Nvidia is getting none with Tegra.

AMD are to blame for their own problems. Bulldozer should never have existed and that is why the big cores are dragging the company down - it's not the WSA or anything else. Intel didn't stop making big cores after P4, AMD shouldn't stop after BD. They just need to stop making the really stupid mistakes they made in the past, and the current management team appears to be light-years ahead in that respect.

 

[krumme]

I hope you have read Anands take on it after talking to rr a few months ago:

http://www.anandtech.com/show/7281/understanding-amd-semi-custom-strategy

All the focus on margin is as i have stated before just plain stupid. Who else operates business with 40% margins. I never understanded it. Like we only live in an Apple world and not where the state bonds are 1%.

The plain problem with the new semi custom strategy is then what do they do?

What comes after the obvious consoles?

It needs to be both complex but not overly so. I cant see the market materialize. It might be there in theory in spades but thats not enough.

 

[krumme]

Unlike Intel Amd have no history or competence in the more business side of things. It means they have to build it. They seem to do that aquiring ip for eg security outside. Fine. But they have to do it without using any money of importance and they have to do it quick unlike the nature of such developing.

Looks hopeless to me if it wasnt under such circumstances innovation often trives - lol - but its some risky journey. But at least they are moving away from copying Intel like they have done since day one.