Hello I am new here but have a few questions and ideas regarding Cuda.
The first thing is regarding Nvidia’s hardware Physx.I know that it is hardware accellerated on G8 GPU’s and above.I have a 9800 GX2 currently and have tested it out with Unreal 3 both with and without a BFG/Agea pci Physics card.And of course without hardware Physx enabled.Anyways,it does make a hefty difference with it on and just the video card itself but with an actual pci physx card there is a major improvement overall.
So what I was wondering is,Is Nvidia’s Physx being ran on the gpu as an extra load,and if so,is it compatible with there Tesla PCI-e cards?
The reason I am asking is I have had a vision,so to speak,involving some certain high end hardware,and was thinking if a few things were set in place that the potential of the combined hardware could very well excede modern supercomputers in a desktop system.A bit on the pricy side yes,but not out of reach for consumers wanting the very best technology has to offer in one system.
For example,lets say Nvidia’s drivers include the support for cuda based Physx acceleration,as well as AI acceleration and any other type of acceleration posible if Microsoft,with there upcoming Direct X 11.0 api,includes the cuda base as a hardware accelerated api as well.
So at this point what you would have is the ability for any of the 3 pci-e motherboards that are out,just pop in a Tesla card and get a boost in graphic performance like none we have seen to date.And at the same time,have the processing power of the Tesla card as well for other apps that support it.
Now imagine Intell’s Skulltrail motherboard with 3 single slot BFG GTX 280 watercooled video cards in 3way SLI.And in that 4th slot,A Nvidia Tesla C1060 card.You still have two pci slots to put in an Asus Xoner sound card and a bigfoot Killer M1 gaming lan card.Not to mention dual Xenon 3.2ghz extreme CPUs, the 9775’s I believe.And of course you have to top that off with two OCZ 250gig SSD’s in raid 0 for your OS drive.Vista 64bit of course.And I almost forgot to mention the ability to have up to 16gigs of ram,fully buffered ecc no less.
For strictly gaming yes this is major overkill but,if you need an all in one system with massive power for things like Adobe’s CS 4 that is GPU accellerated and one of those h264 cuda accellerated video transcoders not to mention other cuda apps as well.Including digital 3d rendering,and cinimephotogphy,then this is pretty much top of the line.
At app. 12k in price,and all housed in one full tower case,you have to tell the wife,“Honey,I shrunk the car”!
These guys can build it.We just need driver support.