AMD Q2 2013 Market Share up 2.2% Q to Q (Mercury Research)

[Phynaz]

Intel cannot maintain their margins while competing in mobile. It's just not possible.

Margin is a function of volume.

How else do you think Qualcomm does it?

 

[SiliconWars]

http://forums.anandtech.com/showpost.php?p=35375910&postcount=32

Nobody expected Intel to get zero design wins, and Samsung tend to use a bunch of different components so that's no surprise at all. Judging by the bad reception their Clover Trail tablet got, this one might be as horrible though.

Intel needs to be in high profile, high volume designs but they've missed them all as far as I can tell. Next year they'll have the A12 and A53 to compete against and it will probably be the same story.

 

[Phynaz]

They're not the only ones in Facebook.

Irrelevant.

 

[Blandge]

My argument would be simply, Intel has never faced the kind of competition they are trying to usurp. Whatever money Intel can throw around, Qualcomm can double it.

Qualcomm is firmly entrenched. I don't believe the onus is on me to prove Intel will fail - I do believe they are failing however and their lack of traction in mobile is all the proof I personally need. The onus is on Intel to prove they are capable of breaking the mobile market, something they are currently as far away from as they've ever been.

Even if the ARM crowd can take just 20% of Intel's server market, the damage will be huge. The ARM guys can afford to lose some tablet and even some amount of phone share to Intel, but Intel can't afford to lose server share to the ARM crowd.

Uhh... Yes they have. Try Intel muscling its way into the server market against big iron at turn of the millennium. This was a very similar situation where Intel was going up against the likes of IBM (Who was an $80b dollar company in 1997) and SPARC. In 1997 Intel had almost zero revenue from servers, and now by 2013 they are pulling in billions of dollars every quarter. All the while IBM has been trying to crush them like the puny bug they are.

Intel has repeatedly proven they can push into new markets throughout its history. Hell Intel was a DRAM company for many years before inventing the microprocessor with the 4004.

Sure Intel may not be able to break into the mobile market the same way it did with servers (there's never a guarantee), but lately it seems like people's memory only goes as far back as the Athlon 64/Pentium 4 era.

 

[blastingcap]

Competition is good for consumers.

 

[AtenRa]

AMD wasnt the first with an integrated memory controller. They wasnt even close by decades.

AMD wasnt the first with SOI either, it was IBM.

AMD wasnt the first with a native quadcore either. They was atleast a couple of years late there.

You sure like hypes.

We are talking x86

 

[Phynaz]

We are talking x86

Which would still make you wrong about the memory controller.

And as far as native quad core (whatever that is), AMD has stated that going with a monolithic quad core when they did was a mistake.

SOI...They were "first"(and only) to a manufacturing technology that everyone - including themselves - is abandoning. It actually makes Intel look smart for not going SOI.

Technology for technologies' sake isn't a win.

 

[BallaTheFeared]

Barcelona was a mistake, the beginning of the end so to speak.

 

[Charles Kozierok]

Which would still make you wrong about the memory controller.

How so, at least in the x86 world?

 

[Phynaz]

Intel's 386SL was the first x86 CPU with an integrated memory controller.

 

[Charles Kozierok]

Intel's 386SL was the first x86 CPU with an integrated memory controller.

Thanks, didn't realize that. But really, it seems to have been more of an oddball than anything and Intel never followed through with it; AMD really popularized the IMC in the x86 line.

 

[krumme]

What could have helped BT - and to a smaller degree Temash - to get off the shelves was success for win8. But its a failure. Both on desktop and tablets. Its a bad os. The user dont want it even for free.

Then bt have to compete using that os against ss android and a ios. Its not going to work. That leaves LG and Sony and android competing with s800 that is firmly established here already and have great succes.

Android on s600 and up simply fly on the user experience. Not in any single way is it slow. There is plenty of power and most users wouldnt even notice if it was cut in half.

 

[mrmt]

What could have helped BT - and to a smaller degree Temash - to get off the shelves was success for win8. But its a failure. Both on desktop and tablets. Its a bad os. The user dont want it even for free.

Not a problem. You can experience Bay Trail in Android flavors and win8/android hybrids.

 

[krumme]

Not a problem. You can experience Bay Trail in Android flavors and win8/android hybrids.

Yes
"That leaves LG and Sony and android competing with s800 that is firmly established here already and have great succes.

Android on s600 and up simply fly on the user experience. Not in any single way is it slow. There is plenty of power and most users wouldnt even notice if it was cut in half."

 

[mrmt]

SOI...They were "first"(and only) to a manufacturing technology that everyone - including themselves - is abandoning. It actually makes Intel look smart for not going SOI.

I wonder why people credit AMD so much for SOI. It's not their technological choice, it wasn't developed by them, it wasn't even deployed first by then. In fact, AMD never developed an entire process node by themselves. AMD first used Motorola's node technology (Hello, Ruiz!) and then IBM's. Praising AMD for SOI is the same thing as praising the car pilot for the race track design.

 

[AtenRa]

SOI...They were "first"(and only) to a manufacturing technology that everyone - including themselves - is abandoning. It actually makes Intel look smart for not going SOI.

Technology for technologies' sake isn't a win.

Yeap, Intel spend Billions to make FinFets in order to have the same advantages/characteristics SOI (FD) brings.

AMD was first to use SOI in x86, it gave them a complete technological advantage over Intel. They give it up because after 20nm they will use Fin-Fets(no use of SOI needed), plus the higher volume that comes with small cores makes FD-SOI not viable.

 

[Charles Kozierok]

AMD was first to use SOI in x86, it gave them a complete technological advantage over Intel.

What "complete technological advantage" would that be, please?

And if they have that advantage, why have they not been able to develop any chips to exploit it?

They give it up because after 20nm they will use Fin-Fets(no use of SOI needed)...

So you rolled your eyes at Intel developing Fin-FET technology instead of SOI, but AMD abandoning SOI for Fin-FET is somehow a good thing?

At any rate, the technology does not matter by itself, only what the company does with it.

 

[Idontcare]

SOI more or less single-handedly won AMD the performance advantage for the entire 90nm node, a bit of the 130nm node, and a bit of the 65nm node. (about 4 yrs all summed up)

The advantage was really low leakage, comparatively, to the bulk-Si nodes Intel was pursuing.

The problem was it ran out of gas and stopped delivering outsized gains at around 45nm, and even more so at 32nm.

This was actually Intel's expectations as well based on long-range pathfinding research they did back at the 180nm node when AMD was also considering going with SOI for 130nm production. Intel's conclusion was that it wouldn't be a long-term solution so it wasn't worth developing, opting instead to pursue HKMG (albeit on a delayed entry timeline of 45nm).

There's no denying AMD bundled together a lot of great ideas, even if they weren't there own, into making the most of the 130-90-65nm progression. (not so much the 65nm, but at least the start anyways).

 

[mrmt]

There's no denying AMD bundled together a lot of great ideas, even if they weren't there own, into making the most of the 130-90-65nm progression. (not so much the 65nm, but at least the start anyways).

What did they bundle here? They had no process node research, it was acquired from IBM and Motorola. More important, who would have sold them a non-SOI node if they wanted? Was SOI AMD's choice or was the only thing available on the market for purchasing? Maybe TI?

 

[AtenRa]

What "complete technological advantage" would that be, please?

And if they have that advantage, why have they not been able to develop any chips to exploit it?

See IDC post bellow

So you rolled your eyes at Intel developing Fin-FET technology instead of SOI, but AMD abandoning SOI for Fin-FET is somehow a good thing?

There is not good or bad, there is viable and not viable. AMD Could use FD-SOI even at 32nm and have way better characteristics, but it would bring cost up, spend more time/money in R&D and bring Time to Market its products longer times.

Bulldozer could be way better but it would cost more and it would be more than a year late. AMD made a decision not to invest in process after the 90nm most probable because they didnt have the money. So after that they lost any advantages they had with SOI over Intel.

Not to mention that PD SOI lost a lot of its steam after 65nm. It would take AMD to use FD-SOI at 45nm in order to have a technological advantage over Intel but they couldn't do it at the time.

 

[AtenRa]

What did they bundle here? They had no process node research, it was acquired from IBM and Motorola. More important, who would have sold them a non-SOI node if they wanted? Was SOI AMD's choice or was the only thing available on the market for purchasing? Maybe TI?

You can have SOI or non SOI for every process. You can even have SOI with Fin-Fets. AMD Chose to go with SOI at the time because it gave them an advantage. But they didnt continue to pursue that process advantage with FD-SOI later on, so PD-SOI at 45nm and beyond was not that impressive, giving them less and less on each node.

 

[Blandge]

You can have SOI or non SOI for every process. You can even have SOI with Fin-Fets. AMD Chose to go with SOI at the time because it gave them an advantage. But they didnt continue to pursue that process advantage with FD-SOI later on, so PD-SOI at 45nm and beyond was not that impressive, giving them less and less on each node.

Funfact: Intel's finFETs are FD. Not sure if you knew that.

 

[AtenRa]

Funfact: Intel's finFETs are FD. Not sure if you knew that.

yes i know that, it is one of the reasons why AMD will not use SOI after 20nm.

 

[Blandge]

yes i know that, it is one of the reasons why AMD will not use SOI after 20nm.

Wasn't 32nm the last node with SOI?

 

[Idontcare]

What did they bundle here? They had no process node research, it was acquired from IBM and Motorola. More important, who would have sold them a non-SOI node if they wanted? Was SOI AMD's choice or was the only thing available on the market for purchasing? Maybe TI?

By bundle I mean they pulled together a number of features from a varying number of sources. Like an OEM.

Look at the Athlon 64 X2, a combination of incorporating an IMC, an Athlon core, and SOI, plus all the in-house development on expanding the ISA and microarchitecture to 64 bits.

Its no different with cars. Toyota may not have invented the automobile, or even the electric car, or seatbelts, airbags, tires, internal combustion engine, etc...but there is no denying the Prius is a rather special product that the world is better off for Toyota having developed by bundling tons of pre-developed ideas.

And at the same time no one else has quite figured out correctly how to bundle all those together and make a product that people like. Chevy Volt, Honda Insight, Nissan Leaf, etc...all falling short.

So I'm just saying that for a number of years AMD really pulled it together with their product development, and their profitability showed that too for the time.