Intel will be finally be in the smartphone arena, but 500 million units by 2015 would be 40%-50% of the WW market, depending on which forecast you use. This would be a significant feat for any company, but extremely difficult to achieve when the leading smartphone vendors are developing chips in-house and you are competing against strong industry incumbents.
Intel has huge advantages in terms of process technology, Mfg capacity, software and design teams, They can turn around and verify designs much faster than Fabless companies if they put their business mind on it,. If they just focus on it as a side business they will not be sucessful
Samsung has similar capabilities like Intel, it would be a good competition.
But Intel can beat Fabless if they put their mind to it.
I wish Intel success in whatever endeavor they pursue. They create and sustain so many jobs for the middle class Americans. Most other fabless companies care so much about profit that they outsource jobs to most Asian companies.
Profit-motive is strong for any business let alone fabless or not. Intel has had many tries at the mobile market and they are certainly serious about it and have deep pockets. But that does not guarantee success. It remains to be seen if the ultra-mobile space will accept generic silicon.
With all the resources, fab advantage, deep pockets and the number of years they have been at this to only show their designs with companies like Lava and Safaricom says a lot about the (lack of) traction they are seeing.
Intel SINGLE-core mobile CPU benchmark:
Intel DUAL-cores mobile CPU benchmark:
Compare to what? Even Quad-cores ARM?
Many have dismissed Intel when it wants to make "mobile" processor. Who's laughing now??
And Intel want to make this Atom's line as it's main line (even for micro server clusters). And it will gain out-of-order executions logic. And of course more cores.
Who should be worried? Intel or ARM?
Right now this seems ridiculous. But watch out in several years. Those Xeon-Phi co-processor is a massive simplified x86 with SIMD/MIMD steroids running on a supercomputer ring-bus (like the CRAY's do).
That thing can be scaled for geez.. high performace GPU? How long before you see this incorporated into its line of CPU?
Again, instead of learning 2 machine language (CPU+GPU) you only need to learn 1. And use the same SDK.
Right now, ATI/NVIDIA/IT/ARM seems under-estimate Intel efforts. Let see this in a couple of years.
Blog Doing Math in FPGAs Tom Burke 22 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...