Except GPU:s aren't sacrificing their graphical performance for general-purpose computing. Their graphical performance is just wrapped for general-purpose computing.

They most likely based their first compiler on the POWER ISA. They use a combination of old and new elements, and the new elements make up a larger proportion of the whole than Larrabee.

easy maybe

but memory whores

look at visual studio UGH its pain to work even .net with it.

i think intel will have to show first that larrabee can do what they claim it can. x86 chips are good multipurpose chips so good cpus, but if you need just a limited instruction set thats specialised on doing gfx output the current ati/nvidia chips have a lot more power.
So there is still the question how good a highly parralellised x86 based chip (which larrabee should be) can compete with the chips that are specially built for just gfx purposes. you can also see how much power the current gfx chips from nvidia have as an example when you compare physx performance on them with physx performance on a intel cpu or even a physx card (yes nviddia gfx cards destroy even these specialised physics boards in tehse ebnchmarks, and amd gpus would most likely do it too if they would be supported)

add to that the history of intel when it comes to big announcements for gfx chips that neevr lived up to expectations once tehy came out and finally ended in their chipset graphics, and then you see why many in the industry are on a wait and see position when it comes to larrabee. if larrabee can deleiver what intel promises then i'm sure they will sell, but people want to see first if it delivers what it promises.

How will this affect b. compatibility? MSFT, if they do this, will have 3 different GPUs from each of the 3 consoles.

If Intel covers the whole of the hardware side of things - design, chips etc then Microsoft would be free do do whats best and tap into the billions of dollars worth of software engineering talent and interface research and dream up something nice to compete with the Wii2/PS4.

Its quite the partnership which would let Microsoft leverage its x86 strength and it should make it darn easy to develop for.

8 Core Sandybridge x86 (Nehalem successor) vs  Cell 2.... Which one would developers want to use? xD

@Halo gamer it would have full backwards compatibility with Xbox 1 (same architecture) The Xbox 360 would if they spent some time working on the compatibility *I think*

Quaiky "i think intel will have to show first that larrabee can do what they claim it can. x86 chips are good multipurpose chips so good cpus, but if you need just a limited instruction set thats specialised on doing gfx output the current ati/nvidia chips have a lot more power.
So there is still the question how good a highly parralellised x86 based chip (which larrabee should be) can compete with the chips that are specially built for just gfx purposes. you can also see how much power the current gfx chips from nvidia have as an example when you compare physx performance on them with physx performance on a intel cpu or even a physx card (yes nviddia gfx cards destroy even these specialised physics boards in tehse ebnchmarks, and amd gpus would most likely do it too if they would be supported)

add to that the history of intel when it comes to big announcements for gfx chips that neevr lived up to expectations once tehy came out and finally ended in their chipset graphics, and then you see why many in the industry are on a wait and see position when it comes to larrabee. if larrabee can deleiver what intel promises then i'm sure they will sell, but people want to see first if it delivers what it promises."

Intels problem is a chicken/egg relationship. The power is likely going to be there Larrabee is specced to have 1 Tflop of double precision power compared to 120 Gflops from the R700 and 20 from the Cell 1. The problem they have is if they don't get developers working on it, tools won't be developed to exploit its power and if its power isn't exploited then developers won't work on it.

Don't worry about physx, its a highly parellel architecture its likely to do fine on their own brand of physics *Havok* so they don't actually need to run Physx anyway.

a cpu gpu combo would do wonders to reduce the price of a console.. when are they going to have the hard drive built onto the cpu?

Sony will do the Cell cpu part in PS4.  I'd go with 8 (so 64 SPUs) at 6ghz (IBM has already run Cells at this speed).  It'll be really really cheap by then, and devs will be really really good with Cell architecture by then.

Larrabee, who knows.

Except the lack of tools is a one-off mistake that Sony can learn from and overcome.  And it already has, as evidenced by the tools it's developing and giving away, as well as, apparently, one free visit from the Sony CELL experts (I think some are in Santa Monica, some in London) to devs wanting to see how to best optimize their shit.  I certainly wouldn't count on Sony making a similar mistake (lack of early tools/support) next gen.