NVIDIA Increases Foundry Outsourcing to TSMC and UMC

[DaedalusHelios]

i think ibm has their own x86 licence tho

not yet...... and thats a fact

Which is why i believe Intel you do it's best to try and block an ATI (or nVidia) buyout by IBM because that would mean they could try and make a CPU out of a GPU. Or am i thinking wrong?

Its totally different. Tesla is great for floating point calculations with oil discovery but thats not really a CPU architecture and it works in totally different ways.(thats the closest thing to what you were trying to say I think )

 

[Wile E]

i think ibm has their own x86 licence tho

No, they don't. That's supposedly one of the things keeping a deal from happening between them and AMD.

 

[HTC]

Its totally different. Tesla is great for floating point calculations with oil discovery but thats not really a CPU architecture and it works in totally different ways.(thats the closest thing to what you were trying to say I think )

If IBM does not have a x86 license, then what i said is totally untrue, and i'm not referring to to your point though, currently, it's 100% correct.

What i meant was IF IBM had a x86 license and managed to buy either ATI or nVidia, they would be in a position to create a new CPU based on a GPU that could use the GPU's full power but with CPU instructions.

Do i explain this properly?

 

[DaedalusHelios]

If IBM does not have a x86 license, then what i said is totally untrue, and i'm not referring to to your point though, currently, it's 100% correct.

What i meant was IF IBM had a x86 license and managed to buy either ATI or nVidia, they would be in a position to create a new CPU based on a GPU that could use the GPU's full power but with CPU instructions.

Do i explain this properly?

Intel already is finalizing Nehalem.... which is what you are referring too.

 

[HTC]

Intel already is finalizing Nehalem.... which is what you are referring too.

I was referring to a company other then Intel or AMD: in this case, IBM.

 

[DaedalusHelios]

The CPU and GPU have different instruction sets because they are completely different. Although Nehalem is a processor with a memory controller and a GPU built into it. That is why I mentioned Nehalem. To be gaming grade graphics it would consume a ton of power and create enough heat to need TEC cooling. So I believe Nehalem would have to feature basic graphics acceleration that falls short of performance found in the discrete graphics market.

IBM would just use existing fabs at first. Eventually after all the reorganization would be done we would see some interesting breakthroughs in technology.

 

[HTC]

The CPU and GPU have different instruction sets because they are completely different. Although Nehalem is a processor with a memory controller and a GPU built into it. That is why I mentioned Nehalem. To be gaming grade graphics it would consume a ton of power and create enough heat to need TEC cooling. So I believe Nehalem would have to feature basic graphics acceleration that falls short of performance found in the discrete graphics market.

IBM would just use existing fabs at first. Eventually after all the reorganization would be done we would see some interesting breakthroughs in technology.

Possibly: dunno, really

Imagine: to be able use a 3870x2 or a gt200 full power as a CPU ... i'm drooling here ...

 

[spud107]

i was reading something about that,
using stream processors/unified shaders to compute stuff,
nvidia has cuda for this, not sure bout ati,
an 8800 has around the same computing power as some supercomp in france