XBitlabs: Advanced Micro Devices Set to Unveil New Strategy Next Week

[soccerballtux]

Well I certainly think it would make for an interesting product and a great battle with Nvidia!

1. AMD Engineered Custom ARM CPU

2. Radeon GPU

All contained within the OMAP SOC. I like it! Quite frankly it would probably be my future dream product if it really had that spec!

I just wonder how TI would compete against Nvidia's CUDA on the HPC level If they got rid of x86?

knowing AMD they would leave out key feature XYZ because they were too rushed and management didn't give the design a good once over. It would be something stupid like power gating or the GPU doesn't clock down or the media decoding engine is 95% complete but lacks certain decode instructions which means it can't play 30% of the videos you throw at it.

I give up. I've resigned to just paying more for a good computer, and we've got enough great ARM vendors that we'll be covered if AMD sucks at this, too. AMD is completely off my "excite" radar now.

 

[Idontcare]

Do you have a link or search phrases I could use to read about it?

http://www.reuters.com/article/2010/05/13/us-oracle-idUSTRE64B5YX20100513

Ellison says he has already stopped the carnage at Sun, less than four months after the sale closed in January. "Their management made some very bad decisions that damaged their business and allowed us to buy them for a bargain price," he told Reuters.

Lots more in that link, like:

"The underlying engineering teams are so good, but the direction they got was so astonishingly bad that even they couldn't succeed," said Ellison. "Really great blogs do not take the place of great microprocessors. Great blogs do not replace great software. Lots and lots of blogs does not replace lots and lots of sales."

^ Sound familiar? How did those "Bulldozer goes to 11" blogs turn out?

 

[cbn]

Maybe Project WIN will involve a stronger software GPGPU strategy?

Hiring software engineers to replace some of 1400 employees just released.

 

[cbn]

AMD needs to announce a new chip with 2 bobcat cores, two ARM cores, 80 SP, 2GB of stacked DDR3, and an 8 channel integrated SSD controller with 64 GB of stacked NAND. All for $100. Wake me up when its 2014.

That would be shocking, but @ 0:15 to 1:07 in this You Tube video Rory Read really liked the idea of "convergence of devices". (ie, ARM and x86 in a single device)

 

[Dadofamunky]

http://www.reuters.com/article/2010/05/13/us-oracle-idUSTRE64B5YX20100513

Completely off topic, I think Ellison is very impressive. Based on that article, he has a fount of common sense that explains why he's worth $28B. I also think that AMD is gutting their GPU group - the only one that's actually been successful. I ahve a feeling they're going to regret that.

What I like about Ellison is that he doesn't use mindless marketingspeak. Rory Read's explanations were nothing but. Guys who talk like that are a) not to be trusted; b) usually colossally incompetent.

 

[cbn]

knowing AMD they would leave out key feature XYZ because they were too rushed and management didn't give the design a good once over. It would be something stupid like power gating or the GPU doesn't clock down or the media decoding engine is 95% complete but lacks certain decode instructions which means it can't play 30% of the videos you throw at it.

I am sure the OMAP with AMD would be a great chip.

I am more worried about how this would be addressed--> http://forums.anandtech.com/showpost.php?p=32541980&postcount=121

The irony is that AMD bought ATI to differentiate themselves from Intel.

Unfortunately, With Intel bottlenecking Windows with Poor GPUs for so long there was no reason for MS to aggressively develop their own GPGPU API.

This left Nvidia by itself to innovate with CUDA in order to survive, while AMD was left with OPEN CL as it only hope.

The question is will MS invest more in an easy to use GPGPU API? Or are Intel GPUs still going to be pretty bad at GPGPU for some time to come?

Otherwise we are just talking about Gamer chips here, not something that could take the fight to Nvidia!

 

[Idontcare]

knowing AMD they would leave out key feature XYZ because they were too rushed and management didn't give the design a good once over. It would be something stupid like power gating or the GPU doesn't clock down or the media decoding engine is 95% complete but lacks certain decode instructions which means it can't play 30% of the videos you throw at it.

I am sure the OMAP with AMD would be a great chip.

I am more worried about how this would be addressed--> http://forums.anandtech.com/showpost.php?p=32541980&postcount=121

The irony is that AMD bought ATI to differentiate themselves from Intel.

Unfortunately, With Intel bottlenecking Windows with Poor GPUs for so long there was no reason for MS to aggressively develop their own GPGPU API.

This left Nvidia by itself to innovate with CUDA in order to survive, while AMD was left with OPEN CL as it only hope.

The question is will MS invest more in an easy to use GPGPU API? Or are Intel GPUs still going to be pretty bad at GPGPU for some time to come?

Otherwise we are just talking about Gamer chips here, not something that could take the fight to Nvidia!

Once again notice the stark difference between AMD versus all these other companies.

AMD relies on others to enable their path for success. And when these other companies don't, then AMD falters.

All the other businesses you highlight are in the business of forging their own success.

If you are a 2-bit player then why would anyone invest considerable resources of their own on a gamble that you'll still be around 4-5 yrs from now with an executive team that still makes it a priority to pursue the original market?

AMD might be here 4 yrs from now, but will Rory Read? And will the new AMD have the stamina to stick with the initiatives set out by Rory? Or will they abandon those initiatives, leaving their business partners holding the bag having developed supporting products for something that never materializes from AMD?

There is a reason Microsoft eventually pulled support for Windows on MIPS, Itanium, Power, etc. They aren't in the habit of squandering resources on niche players that have high probability of being less relevant as time goes on.

It is up to AMD to forge their own success as Nvidia did with CUDA and Apple did with the iPhone and the iApps business, if AMD continues to rely on the corporate charity of their business partners then we know where this is headed, we can all connect the dots on this one.

 

[Dribble]

It is up to AMD to forge their own success as Nvidia did with CUDA and Apple did with the iPhone and the iApps business, if AMD continues to rely on the corporate charity of their business partners then we know where this is headed, we can all connect the dots on this one.

That will be tough. Nvidia famously have more software engineers then hardware ones - the whole company seems to see the hardware as just the start, and spends a lot of time trying to *add value* after that. AMD has never had the people or the corporate focus to do this - traditional x86 development needs very little additional software support by AMD as MS/Intel provide it all. Like everything else facing them it seems right now it'll be very hard from a standing start to turn this around.

 

[Rifter]

so its the 9th, where is the news

 

[Rifter]

That will be tough. Nvidia famously have more software engineers then hardware ones - the whole company seems to see the hardware as just the start, and spends a lot of time trying to *add value* after that. AMD has never had the people or the corporate focus to do this - traditional x86 development needs very little additional software support by AMD as MS/Intel provide it all. Like everything else facing them it seems right now it'll be very hard from a standing start to turn this around.

just saw this and i agree 100% this has always been a issue with AMD.

 

[Arachnotronic]

I really, really don't get why everyone thinks AMD should go with ARM cores. The market is flooded with much larger competitors, some of whom can fab their own stuff.

AMD's probably just going to announce that they're going to do exactly what Intel's doing: focusing heavily on the mainstream APUs and then continuing to prune Bulldozer for the server segment.

 

[Rifter]

I really, really don't get why everyone thinks AMD should go with ARM cores. The market is flooded with much larger competitors, some of whom can fab their own stuff.

AMD's probably just going to announce that they're going to do exactly what Intel's doing: focusing heavily on the mainstream APUs and then continuing to prune Bulldozer for the server segment.

I agree, ARM is a bad idea for a company with as little cash on hand as AMD.

 

[peonyu]

I agree, ARM is a bad idea for a company with as little cash on hand as AMD.

Yup. Bad management has been their bane to, they went from the Athlon to the Athlon 64 cpu [both were awesome and great sellers] to the Bulldozer...Good management would have scrapped that cpu at its ugly inception years ago. Its all coming to a head now though.

 

[Genx87]

I really, really don't get why everyone thinks AMD should go with ARM cores. The market is flooded with much larger competitors, some of whom can fab their own stuff.

AMD's probably just going to announce that they're going to do exactly what Intel's doing: focusing heavily on the mainstream APUs and then continuing to prune Bulldozer for the server segment.

Because the markets ARM is currently the defacto standard for are exploding. Where the desktop\laptop markets have stagnated. If AMD wants to see growth it wont be at the expense of the 800 pound gorilla(intel).

 

[Arachnotronic]

Laptops and desktops have not stagnated. They're doing just fine. Also, ARM chips are low margin -- you've gotta be shipping in serious volume for it to be worth it.

 

[Lex Luger]

Although the market for ARM is exploding, I dont think ARM can compete with intel long term.

Atom are poo right now, but intel finally is gonna start accelerating die shinks.

32nm atom 2012

22nm atom 2013

14 nm atom 2014

Intel claims 14nm atom faster than phenom 2 x6, their very words.

Unlike some compainies, intel lives up to their performance promises and their production roadmaps.

Once they hit 14nm, ARM wont be able to compete anymore. It will be like sandy bridge vs bulldozer. You will have no choice but to buy intel.

Intel has the best fabs and will continue to have best fabs for at least the next five years. This is how they will take over.

Their atom chips will be significantly better than everyone elses, just like the full size chips are nowadays.

 

[Genx87]

Laptops and desktops have not stagnated. They're doing just fine. Also, ARM chips are low margin -- you've gotta be shipping in serious volume for it to be worth it.

PC shipments have stagnated.

http://www.zdnet.co.uk/news/desktop...t-shrinks-as-business-spending-dips-40092510/

Phones and Tablets are expected to explode. Tablets are expected to hit 300 million\year up from ~20 million this year. Phones will push into over a billion units\year.

So yeah, you can ship serious volume and the market is expanding rapidly vs the PC market that has stagnated and are forced to fight Intel.

 

[Genx87]

Although the market for ARM is exploding, I dont think ARM can compete with intel long term.

Atom are poo right now, but intel finally is gonna start accelerating die shinks.

32nm atom 2012

22nm atom 2013

14 nm atom 2014

Intel claims 14nm atom faster than phenom 2 x6, their very words.

Unlike some compainies, intel lives up to their performance promises and their production roadmaps.

Once they hit 14nm, ARM wont be able to compete anymore. It will be like sandy bridge vs bulldozer. You will have no choice but to buy intel.

Intel has the best fabs and will continue to have best fabs for at least the next five years. This is how they will take over.

Their atom chips will be significantly better than everyone elses, just like the full size chips are nowadays.

People have been saying this for years. Yet where can I buy an intel phone? Or what % of the tablet does Intel own? Hint, it is under 5%.

And you act as if ARM designs wont also be movign to new process nodes and have performance increases right along with Atom.

 

[smartpatrol]

Although the market for ARM is exploding, I dont think ARM can compete with intel long term.

Atom are poo right now, but intel finally is gonna start accelerating die shinks.

32nm atom 2012

22nm atom 2013

14 nm atom 2014

Intel claims 14nm atom faster than phenom 2 x6, their very words.

Unlike some compainies, intel lives up to their performance promises and their production roadmaps.

Once they hit 14nm, ARM wont be able to compete anymore. It will be like sandy bridge vs bulldozer. You will have no choice but to buy intel.

Intel has the best fabs and will continue to have best fabs for at least the next five years. This is how they will take over.

Their atom chips will be significantly better than everyone elses, just like the full size chips are nowadays.

ARM doesn't have to beat Intel on performance in order to displace them. If ARM can offer good enough performance at a significantly cheaper price than Intel, then ARM can do to x86 exactly what x86 did to Alpha/Sparc/IA64/etc.

Not only that, but ARM seems extremely attractive from an OEM standpoint. Would you rather buy chips from Intel, or have your choice of several ARM vendors who are all competing for your business?

I honestly don't think we're too far away from the day when x86 is just for high-performance workstations and servers. In the very near future, we'll have Tegra 3, Qualcomm S4, and various Cortex A15 SOCs. I think at that point, ARM will be more than powerful enough for most ordinary users' everyday computing needs.

 

[cbn]

That will be tough. Nvidia famously have more software engineers then hardware ones - the whole company seems to see the hardware as just the start, and spends a lot of time trying to *add value* after that. AMD has never had the people or the corporate focus to do this - traditional x86 development needs very little additional software support by AMD as MS/Intel provide it all. Like everything else facing them it seems right now it'll be very hard from a standing start to turn this around.

I've been wondering how quickly it would take AMD to catch up to Nvidia in GPGPU programming?

Is CUDA something that is likely to slow down in development, while AMD is on the fast learning curve?

 

[lol123]

ARM doesn't have to beat Intel on performance in order to displace them. If ARM can offer good enough performance at a significantly cheaper price than Intel, then ARM can do to x86 exactly what x86 did to Alpha/Sparc/IA64/etc.

x86 designers didn't beat RISC with "good enough" performance and lower prices, they beat them with significantly higher performance, similar RAS features and much lower prices. That's the kind of advantage that you need to displace any architecture that carries with it an enormous legacy and software ecosystem. And this didn't really start to happen until the early 2000's (and the transition is still not completed).

ARM is the dominant and a highly entrenched architecture in the mobile (and parts of the embedded) space and I don't expect it to go away, especially in the spaces where Intel does not want to compete on price alone. What x86 can do, and what I expect it to do given enough effort from Intel, is to push ARM back into washing machines and cheap Nokia phones for the third world where it belongs. Intel has, at this point, the best designers, excellent management and vastly superior process technology to the rest of the industry (and the gap is rapidly growing at the transition to sub-32nm processes) and the inherent advantages of a "pure" and "clean" RISC architecture (a criterion that the ARM ISA actually doesn't even remotely live up to) are insignificant and overstated at best.

Also, the idea that Intel somehow can't make low-power chips and high-performance graphics is ridiculous. It's just that until very recently they didn't care.

 

[Olikan]

Although the market for ARM is exploding, I dont think ARM can compete with intel long term.

Atom are poo right now, but intel finally is gonna start accelerating die shinks.

32nm atom 2012

22nm atom 2013

14 nm atom 2014

Intel claims 14nm atom faster than phenom 2 x6, their very words.

funny, than ARM moves to 28nm in 2012, and say "u mad intel?"

 

[AtenRa]

We will see Intel's x86 (probable Atom 32nm) at Tablets with android-x86 in H1 2012 and Win 8 in H2 2012.

That's coming from Intel themselves.

http://www.android-x86.org/

 

[Genx87]

Intel already has tablets out there. Their battery time at idle is about half what ARM tablets are at load.

 

[lol123]

Intel already has tablets out there. Their battery time at idle is about half what ARM tablets are at load.

Yeah, we know, since that's what you keep repeating on and on and as loudly as you can. What you apparently want everyone to forget is that while the present Atoms aren't at all optimized for ultramobiles (they suck, to put it bluntly, and everyone agrees on that), Medfield, really as the first Atom SoC, will be:

http://www.anandtech.com/show/4345/...-architecture-group-14nm-airmont-atom-in-2014

It's amazing how some people (specifically the crazed Intel haters) expects Intel to have low-power (smartphone grade) processors on the market since before they were even committed to that market. And since they don't, Intel simply must lack the capability to design such a chip. Brillant conclusion.