Re: The funny part is Motorola and Samsung use ATI products in their cellphones.
That's not too surprising. Motorola and Samsung want to reach a broad range of consumers, and their own R&D is really best suited for just a subset of the market. No single company has a robust enough product line to cover such a highly segmented market as cell phones. So it's commonplace to use other chip architectures to cover the gaps. My point was that Imageon is not that big of a name in cell phones. TI probably has 10x the market share in cell phones that ATI does.
Re: And the performance delta ?
Let's say K8 gets 3x the performance of Atom at the same clock. I think that's rather generous, and that it's probably closer to 2x, but the point remains the same. Now look at power. Atom is 15x lower power than K8. Simple math would therefore conclude that it is at least 5x more power efficient, when measured in performance per watt.
Re: How much does a 45nm Turion running at 0.8V and 700-800MHz would burn ? Not more than 3-5w.And we're talking about a full desktop x86 core.
We may soon see, but I doubt it is that low. You start running into min Vcc issues at the low end of the power envelope, and most high performance micro-architectures can't scale. I think single core Core 2 Solo chips (which I consider superior to AMD in performance per watt) can currently hit 1.2GHz @ 5.5W. At this power, the chips is now mostly dominated by leakage, so cutting frequency in half doesn't usually cut power in half. At 800MHz ("Low frequency mode"), the chip still has a 4.6W TDP.
Another important aspect is average power, as it applies to battery life. The same Core 2 Solo processor in deeper sleep mode can still draw as much as 4.0A of current (see datasheet above). Meanwhile, Atom processors can draw no more than 0.2A in the same sleep state, while having the capability to go to even lower power C6 mode.
This enormous difference can place huge differences in the power delivery requirements, as well as the specifications of the battery cells. Only Atom will be able to provide decent battery life for ultra small devices, which only have so much physical volume for a battery and power delivery ICs.
Re: Future Imageon products ( to be released in 2009 ) will incorporate an x86 core ( Bobcat IIRC )
That was Anand's speculation, until AMD removed Bobcat from the roadmap after that article was written. If Bobcat does launch in 2009, you wouldn't expect a highly integrated version of it until another year later, at the very least. By 2010, Intel has said they will have their own SoC's, and it's only speculation as to whether AMD can play catch-up in terms of both processor micro-architecture, as well as full chip integration at competitive power levels.
Re: Being an innovator and being successful are not the same thing.Intel too , has , yet to prove it can go beyond its traditional markets.
When you use the term "innovation" for AMD, you use the processor as proof, but when you use the term "innovation" for Intel, you demand proof outside of the processor, and indeed, outside of the traditional PC markets. Why not try holding your commentary to the same standard?
Re: Even with going by Intel's own prediction on G45 vs. older products , I won't [be eating crow].
Indeed you will. G45 is just a single step in the right direction. The roadmap calls for similar large leaps in 2009 as well as 2010, through a combination of process shrinks and a commitment to stronger architectures. AMD scores well with 780G because they are allowing it to cannibalize their value discrete graphics market. AMD's own HD 3450 doesn't score that much higher than IGP graphics, and in fact both devices use the exact same graphics core. Instead, AMD sells on the promise of "Hybrid Crossfire", which combines value discrete graphics and IGP processing to give a boost in performance with low end cards. Unfortunately, the gap between IGP and value graphics has effectively shrunk due to this strategy, and AMD may find that a large portion of their value discrete market will shrink as a result.
And at any rate, Intel can play the same game, and integrate a larger and stronger graphics core into their products. It was largely unimportant to Intel or their customers to provide strong 3D graphics at the expense of silicon costs, but if the market now demands it, Intel can do just as good a job as ATI, when presented with the same power and area constraints, and then beat them with a stronger process. It will happen by 2010.
Re: Well , that's the point.Why wasn't Intel more aggressive in its specifications for ATOM ?
I think I already answered this question, so let's quit with the circular arguments. To recap, we can only speculate, but I think it was due to numerous forces: time to market, technology maturity and risk management, optimization of fab capacity, ability to meet market requirements without it, etc.
Market Data 
Markets 
AI
