New AMD CEO readies strategy shift

[MentalIlness]

http://techreport.com/discussions.x/21647

Report: New AMD CEO readies strategy shift
by Cyril Kowaliski — 9:42 AM on September 15, 2011


In case you haven't heard, there's a new man at the helm of AMD: freshly minted CEO Rory Read, who joined the chipmaker late last month from Lenovo. According to Fudzilla, which quotes an "insider hint," Read is working on a "twist in AMD's long-term strategy." The site says that twist may very well involve mobile devices:

We heard that Rory Read was famous for turning things around and that he helped Lenovo with some key transitions. Since AMD believes that smartphones and tablets are getting more important, and that they need even better mobile chips, its logical to assume that this new strategy can get AMD closer to these categories.​

The lack of a solid game plan to tackle the ultra-mobile and handheld markets is said to have contributed to the dismissal of AMD's previous CEO, Dirk Meyer. Unsurprisingly, during Read's introductory conference call on August 25, many asked the new CEO about AMD's plans in those areas. Read and other AMD execs shied away from sweeping statements, but they did express a desire to go after ARM in "lower power bands." Perhaps we'll hear more about that soon, then.

 

[frozentundra123456]

Maybe their new strategy is to get their new chips out on time!!

 

[Idontcare]

With Win8 going for ARM, its guaranteed that AMD will secure an ARM license and get going after the same market that Tegra targets.

No way they can ignore that market segment when they are the 3rd largest graphics provider (behind Intel and Nvidia) and both Intel and Nvidia are going after it.

edit: and props to the OP, Thundercats FTW :thumbsup:

 

[AtenRa]

They can go the x86 way in that market as Intel

 

[Ancalagon44]

Win8 and Android both run on x86 and ARM anyway. Why go after an ARM license when they can just use low power x86 cores? The only reason they might want an ARM license is if the ISA is better suited to low power environments, which, being a RISC ISA, it might just be.

 

[Idontcare]

They can go the x86 way in that market as Intel

Intel has leading edge process tech to differentiate their products in the mobile space from ARM and lagging foundry process tech...AMD does not have this option.

If you are going to compete while using less stellar process tech then you need a more stellar processor design.

How is Tegra doing compared to Atom in the mobile sector?

 

[sm625]

I dont see what they have to lose by adding two ARM cores to their two bobcat cores and 80 shaders. If win8 is smart enough to not use x86 until I open an Excel spreadsheet, then more power to em. Mobile devices already work like this. If I run my ipoop touch in airplane mode I can read ibooks and use ReadItLater for about 6 hours. But if I fire up infinity blade it drains my tiny battery in 45 minutes. I imagine x86 cores will work the same way... ok maybe not that bad.

 

[dma0991]

I do think that the current strategy with BD in the server space and Bobcat in the mobile space is effective and a solid plan. The problem is that the plan is so poorly executed with delays and nullifying its effectiveness.

AMD should have an ARM license ages ago to supplement their mobile strategy side by side with Bobcat. I doubt an x86 arch can have a long standby time like an ARM arch would. Even Intel knows that x86 can't standby much therefore much improvment have been made to have faster boot times.

 

[Tuna-Fish]

How is Tegra doing compared to Atom in the mobile sector?

Atom is barely a blip, Tegra is a major player now. Atom needs at least one more shrink, or a major redesign, to be competitive.

 

[podspi]

I dont see what they have to lose by adding two ARM cores to their two bobcat cores and 80 shaders. If win8 is smart enough to not use x86 until I open an Excel spreadsheet, then more power to em. Mobile devices already work like this. If I run my ipoop touch in airplane mode I can read ibooks and use ReadItLater for about 6 hours. But if I fire up infinity blade it drains my tiny battery in 45 minutes. I imagine x86 cores will work the same way... ok maybe not that bad.

That would be awesome, but I'm not convinced that it is actually possible...

If it turns out to be, perhaps we'll start seeing ARM CPUs in most laptops/desktops, so that most Windows apps will work regardless of what arch they were developed for.

Or perhaps Metro apps use some sort of JIT? I haven't heard of anything like this...

 

[Tuna-Fish]

Win8 and Android both run on x86 and ARM anyway. Why go after an ARM license when they can just use low power x86 cores? The only reason they might want an ARM license is if the ISA is better suited to low power environments, which, being a RISC ISA, it might just be.

ARM is better suited to low power, mainly because it's instruction decoding is very simple, which basically saves millions of transistors and lots of power on the front end.

 

[trollolo]

Maybe their new strategy is to get their new chips out on time!!

sssshhh, keep those ground-breaking ideas to yourself, otherwise their CEO might steal them :sneaky:

 

[sonoran]

I doubt an x86 arch can have a long standby time like an ARM arch would.

History is littered with doubters who have been proven wrong, time and time again.

 

[Despoiler]

Read was hired specifically to get into the mobile (smartphone) space. Dirk was fired primarily because he did not and would not consider getting into smartphones. The only twist here is the board twisting arms to get the strategy they want to see.

 

[lol123]

I doubt an x86 arch can have a long standby time like an ARM arch would.

http://www.anandtech.com/show/4820/rattner-shows-off-near-threshold-voltage-intel-architecture-cpu

I don't think it's really a question of whether Intel can get x86 CPUs down to ARM level power consumption in the next few years, it's a question of whether AMD can, with their much lower R&D budget and inferior process technology. However the alternative of licensing ARM technology will make them a competitor to Samsung and Qualcomm rather than Intel, who at least give AMD the lower 10-15% of the market practically for themselves. The ARM market on the other hand has cut-throat price competition at all segments. There is no way that AMD would ever make any decent money there.

I think that the x86 license is still AMD's biggest asset.

 

[Arkadrel]

I doubt an x86 arch can have a long standby time like an ARM arch would.

Apperntly Haswell has a 20x better standby power usage (than the one before it).
Thats bound to but it very close to ARM cpus standby.

and before anyone goes Ohhh but it still uses twice as much!,....

yeah.. but twice as much of nothing is still nothing.


Intel might reach the point where, power use isnt a issue anymore (thanks to be better process nodes, or better battery technologies). How many hours of gameing/working does a laptop/tablet really need? I think once you reach the 12hours+ of avg use, with gameing included ect, the need for faster cpu/gpu > longer battery life.

*IF* that happends, then it becomes just a matter of who makes the faster chips. 86x has a clear edge in the performance department.

 

[AtenRa]

Intel has leading edge process tech to differentiate their products in the mobile space from ARM and lagging foundry process tech...AMD does not have this option.

If you are going to compete while using less stellar process tech then you need a more stellar processor design.

How is Tegra doing compared to Atom in the mobile sector?

AMD can do the same thing they did with Bobcat and Llano, differentiate in the iGPU, where Intel hurts the most, Fusion is the future they say.

Same with Tegra, difference vs Qualcom and Samsung ARM SOCs is in the iGPU mostly and not the ARM Core.


On the manufacturing Intel lead,

GloFos Fab 8 in NewYork will start the production of 22/20nm process in 2013, that's close to when the new 22nm Atom release, i believe AMD could only be two Quarters at the most behind Intel if they will use the 20nm process for Bobcat's replacement, they will also start shipping Bobcat at 28nm in a few Quarters from now.

32nm Atom will suck big time not only because of its own design, but because AMD will not lag behind in manufacturing process vs Intel with Bobcat 2 at 28nm.

What exactly do you mean how Tegra is doing vs Atom ?? in revenue ?? margins ?? market share ?? performance ?? power usage ?

 

[Genx87]

Win8 and Android both run on x86 and ARM anyway. Why go after an ARM license when they can just use low power x86 cores? The only reason they might want an ARM license is if the ISA is better suited to low power environments, which, being a RISC ISA, it might just be.

Low poer x86 cores are still a lot more power hungry than ARM designs. And in the mobile and ultra mobile market. Battery longevity does matter.

 

[Genx87]

Apperntly Haswell has a 20x better standby power usage (than the one before it).
Thats bound to but it very close to ARM cpus standby.

and before anyone goes Ohhh but it still uses twice as much!,....

yeah.. but twice as much of nothing is still nothing.


Intel might reach the point where, power use isnt a issue anymore (thanks to be better process nodes, or better battery technologies). How many hours of gameing/working does a laptop/tablet really need? I think once you reach the 12hours+ of avg use, with gameing included ect, the need for faster cpu/gpu > longer battery life.

*IF* that happends, then it becomes just a matter of who makes the faster chips. 86x has a clear edge in the performance department.

Twice as much means half as long.

 

[lol123]

Twice as much means half as long.

Are you serious? Do you think that a processor that idles at 50mW instead of 25mW in a device that uses maybe 5W in total will cut battery life in half?

 

[Genx87]

Are you serious? Do you think that a processor that idles at 50mW instead of 25mW in a device that uses maybe 5W in total will cut battery life in half?

No, my response was to illustrate having a processor consume twice as much is a processor that consumes twice as much. In a space where literally every minute counts. Having a design that consumes twice as much doing nothing and a lot more during actual useage will lead the design to be looked over. Much like Intel's x86 offerings in the tablet and phone market.

 

[AtenRa]

Are you serious? Do you think that a processor that idles at 50mW instead of 25mW in a device that uses maybe 5W in total will cut battery life in half?

Idle power means nothing, operational power its what it counts, how many hours can you serf the net, watch TV, watch DVDs or even BlueRay movies, play games etc etc.

Intel X86 at 22nm could have close or the same Idle power, but when you start to do things it will consume more than ARM SOCs

We'll have to wait and see, but Intel will have a big competition in the Mobile sector, not what she is used to with AMD in the x86 market.

 

[lol123]

No, my response was to illustrate having a processor consume twice as much is a processor that consumes twice as much. In a space where literally every minute counts. Having a design that consumes twice as much doing nothing and a lot more during actual useage will lead the design to be looked over. Much like Intel's x86 offerings in the tablet and phone market.

The current x86 offerings are looked over because they use orders of magnitude more power than ARM, both at load and in standby, not because they're just marginally worse, which they obviously are not. If you really think that the same rules will apply in few product generations when x86 has cut its power consumption possibly by a factor in the tens then you are seriously deluded. The only thing going for ARM at that point (although that very well might prove decisive) will be its already established position as the dominant ISA in the ultramobile market.

 

[AtenRa]

The current x86 offerings are looked over because they use orders of magnitude more power than ARM, both at load and in standby, not because they're just marginally worse, which they obviously are not. If you really think that the same rules will apply in few product generations when x86 has cut its power consumption possibly by a factor in the tens then you are seriously deluded.

And ARM SOCs will stay in the same power usage/performance ??

Dont you thing that ARM Cortex A15 at 28nm will be much better in power usage/performance than current A9 at 40nm ??

As x86 will evolve the same will happen with ARM SOCs

 

[lol123]

And ARM SOCs will stay in the same power usage/performance ??

Dont you thing that ARM Cortex A15 at 28nm will be much better in power usage/performance than current A9 at 40nm ??

As x86 will evolve the same will happen with ARM SOCs

Probably, but the improvements are not going to be nearly as great as for x86 designs. Both due to diminishing returns in the architecture (ARM has been optimized for low power for decades whereas x86 manufacturers have just recently started paying attention to power consumption) and due to Intel's large process technology and R&D advantage. I hope you saw the link about an Intel x86 chip that can run at 10mV at idle and 400-500mV at load. Can you show me an ARM design that can do the same thing?