Blogs Intel's Computex 2010 Keynote: Atom Everywhere

http://www.bit-tech.net/blog/2010/06/10/intel-computex-2010-atom-everywhere/

 

They seem to like sandy bridge at least. Poor AMD, if it really is 40-50% faster (or even 20-30%) how on earth will bulldozer compare.

 

You can't bulldoze sand.

Perhaps that's the motivation for the code names.

 

You can bulldoze sand. Very effectively!

 

http://forum.coolaler.com/showthread.php?t=240578

Damn it Intel, you've let me down.

 

WTF? I want the most faster processor ever everytime, and now what I see? A TOY cpu called atom! WTF?

 

Did you read the article?

Atom is only referenced three times in the article, and once was only in reference to a codename.

Sandy Bridge, the latest chip that's aimed at enthusiasts was mentioned twice as much.

 

despite Intel's attempts, i still want a Sandy Bridge i7 CPU without integrated GPU. we'll never use the integrated GPU, so why should we pay for it in the material cost and motherboard backplate space?

 

because some people will buy a sandybridge to get them by, or fall back on should a discrete graphics card fail

 

Rich, it's still only good for certain people. Your typical enthusiast/gamer will disable the IGP the first time he turns on the motherboard, so it's wasted die space, pumping up the price.

 

Unless manufacturers get their act together and allow people with discrete graphics cards to utilise the IGP as a dedicated on-die physics processor. Or run a Folding client on it.

 

I'm not sure that's 'getting their act together'. Dedicated physics calculations (as in PhysX) only run where nVidia wants them to, so it's not Asus/Gigabyte/MSI's problem if nVidia don't want them to. Folding, it wouldn't be a GPGPU, Intel wants people to use the cpu for that, remember?

Neither of those suggestions have any chance of being reality, based on the fact that Intel wants physics calculations to happen on the CPU, as part of their business plan.

Honestly, the gpu would be a waste of space.

 

It would be the exact definition of getting their acts together - an open standard that all manufacturers can follow. **** PhysX. PhysX runs on top of CUDA. CUDA (although not free as in speech) is a free standard that anyone can implement. It should be possible to get PhysX running on an Intel IGP. You could even run it on an ATI GPU if it weren't for the fact they're too busy working on StreamSDK and Brook+.

To be honest, Nvidia, ATI and Intel should just implement OpenCL.

 

OpenCL works fine on all nVidia cards, all ATI cards and all x86 processors.

Intel has no incentive to running GPGPU tasks on it's CPUs. Intel's core business is CPUs, and it's trying to convince people that they shouldn't be using GPUs for GP things. There's no point having a graphics card filling up part of your CPU if it's only good for stuff that could have been done on a dedicated GPU or the part of the CPU that would have taken up that space.