Nvidia well positioned to take more graphics market share

I suppose nVidia *could* make their Tesla "GPU Computing" engine more a general purpose CPU but it''s not going to run x86 or x86-64 machine code so it''s irrelevant for the PC or general computing market. Not only does Tesla need a compiler to generate code for it, that code and the problem to solve have to be adaptable to a massively parallel architecture. IOW this is very much a niche market and Intel has no reason to lose any sleep.

Sometime ago, I heard an argument that graphic chip can easily incorporate text function and beat Intel's CPU chip. Is this right? nVidia can actually rolls out superchip to include every function Intel CPU chip can offer, right?

When you start talking about heterogeneous multicores, you are talking about SOC. Yea, there are some infrastructure problems to overcme with Hetero-cores. But, the incredible cost pressures....to get as much cost out of the motherboard and silicon as possible...may be the catalyst that causes hetero-cores to come more quickly than we anticipate. Even Intel is taking some SOC approaches these days....as are others. Maybe this whole graphics discussion is really the harbinger of an extensive SOC discussion.

It seems to me that when nVidia were allowed to have a license to the Intel FSB IP, a precious commodity, it was not Intel's intention that they play in the integrated graphics market. IOW Intel wanted the high performance of nVidia's nForce SLI to contribute to their sales of high-end CPUs for gaming, along with discrete GeForce graphics cards. I don't think this move is going to go down well with Intel - it was not their intention to cede the mainstream chipset & motherboard market to the err, upstart. As for the heterogenous multi-core CPU, there's too much hardware & software infrastructure to be put in place for that to be a consideration at this point... too far in the future and unproven as an advantageous "technology" - remember vaporware?? I also suggest that you take note of the long-running persistent incompatibilities and problems with nForce chipsets both for AMD & Intel CPUs. I could make a list.

I think the multicore pehnomenon (with a graphics core) could argue against nV being able to hold or gain market share in graphics for an extended period of time. If Intel or other multicore CPU makers can't buy nVidea, they will try to dissintermediate them (ie, remove them from the value-added chain).

Intel can't buy nVidia. Because it will cost them too much. They don't have that much. And nVidia will never agree, at any price that Intel can give. nVidia is going straight to the top, they will gain more market share and ultimately take the 1-st place in every graphic market.

This question misses the whole point...the present and coming battle of homogeneous vs heterogeneous multicore processors. Intel seems to think they can do some graphics software (and tweek the x86 core), run it on one of their homogeneous cores in their duo or quad and get similar results of an AMD processor with a Barcelona core and an ATI graphics core (ie, a heterogeneous multicore architecture). And that explains why Wall Street analysts are saying Intel must buy Nvidia at whatever price they ask.

Why are you writing this blog when you are confused between chipsets (i.e. mGPUs), for which Nvidia is NOT gaining market share and discrete graphics, for which Larabee is targeting (not processors integrated into CPUs), and you are assuming AMD is not focused on graphics - ATI would have screwed up the R600 whether AMD bought them or not! While Nvidia's discrete graphics market share lead is unquestionable, please dish us some facts to make us feel good about what Nvidia could do going forward.

Nvidia and ATI/AMD rule on all graphics. Integrated graphics are 90% of the market now and much more important given Vista and MacOSX leveraging their 3D capabilities. If you want to see how poorly Intel performs against ATI go to Youtube and search on "AMDUnprocessed".