When has ATI ever designed their own DSP, and where do you expect an integrated DSP to be of good use inside a general purpose MPU?
ATI has extensive experience with embedded graphic solutions in TVs , cellphones and other handhelds ( check for Imageon )
In other words ATI has the expertise and relations to know what's inside small stuff.Doing a media processor for a cellphone give you some insight after all.
Low power Barcelona has a TDP of 75W, and a max speed of 1.9GHz. I think that precludes any quad core mobile devices for the time being. Meanwhile, Intel has a 45W quad core running at 2.53GHz to be released later this year. They will also have dual core chips capable of 2.53GHz in a 25W power envelope. And these are chips that outperform AMD at similar clock speeds. Unless you have better data, it would seem to me that Intel is doing better these days at hitting low power points.
K8 at low speed has excellent performance/w.I'm thinking more along the lines : what can AMD do to come up fast with a competitor to Atom.
This has been a business choice, rather than as you phrase it, as if it were an engineering limitation. Intel's FSB chips outperform AMD's HTT chips, so the business requirement to go to an integrated memory controller hasn't been compelling enough for Intel to do it earlier. But as memory bandwidth requirements continue to grow and FSB runs out of gas, Intel has developed both Nehalem and Moorestown architectures with IMC to continue to scale. The data shows that the "Build it as you need it" approach has worked very well for Intel. Having sizable revenues from a chipset business has also benefitted them over the years.
More importantly, though - what has AMD integrated besides the memory controller that makes them any more better than Intel in terms of integration? Most of the things that Intel has integrated into the chipset have been innovations that AMD and others have had to follow later. PCI-Express, SATA, USB, integrated ethernet MAC, integrated graphics, and more.
True.AMD showed it was more willing to take risks than Intel when it comes to new technology ( post Netburst anyway ) , they were the innovators more or less.I'm thinking this trend will continue even for low power solutions.
You should expect AMD to have better graphics performance and lower power at 55nm than Intel has at 130nm (965GM chipset from Santa Rosa mobile platform). What is so mysterious or impressive about that, when AMD has 2.5 process generations of advantage? When Intel releases their 65nm chipset (GM45 chipset from Montevina mobile platform), let's see how vast AMD's performance and power lead is, when they only have 0.5 process generations of advantage.
Even if the Intel chipset be made on 32nm and the AMD one on 55nm , the later will be better.
It's not about process , it's about the GPU itself.And it's fairly obvious who has the upper hand , both in uarch and drivers.
Probably many reasons. For one, even at 130nm, the current SCH chipset is a huge improvement over the previous generation, in that it combines both northbridge and southbridge in a single chip that uses a smaller package and is lower power than the previous generation (previous gen is the UMPC McCaslin platform that uses Little River northbridge and ICH-7M southbridge). Second, using 130nm probably uses up remaining 130nm capacity, so that Intel doesn't need to take a charge as these last remaining fabs ramp down. Third might be time to market advantages. Fourth is that the next generation will receive an even larger boost by shrinking the chipset down to a much smaller process (likely 65nm), and putting more features on the CPU.
When you are so thrifty about using your resources ( like older 130nm FABs ) you can end up shooting yourself in the foot.
Whatever the reasons, in the end, the Menlow platform that is now called Centrino Atom is an incredible piece of engineering hitting ultra low power envelopes that many would have thought would be impossible to hit with x86 architecture. And the next generation will be even better.
True.But why weren't they a bit more agressive ? Why did they build a chip that isn't exactly suitable for what they were targeting ( the cellphone market ) ? Instead they went on the create an interim between cellphones and laptops.
So says the rumors, at least. I wonder if ARM can live up to the expectations that people have put on them. Going OOOE is not trivial, and it doesn't come with a trivial amount of power added. To think that it will improve IPC, go to higher clocks, and be half the power of the next generation Atom seems like wishful thinking to me, especially when there's no data to support it.
By the way, as it is, Atom is similar in power to ARM's Cortex-A8, while having much higher performance and faster clocks. Of course, vendors have used ARM with a much higher level of integration, which helps the overall platform to be lower power, but Atom will catch up with that next year as well. It could end up being a close call, and where do you think the market will go, if x86 and ARM have similar power characteristics...?
I think ARM proved itself by now , it's Intel who has to show cellphone makers it has the upper hand.As for x86 vs. Arm at a close call , x86 still needs a true OS for cellphones and mids.
Putting Vista or some stripped down Linux is insane.
Market Data 
Markets 
AI
