A very dumb question...

[eagerigor]

If say someone had a disposable income, would they be able to use an nvidia tesla as a physx card? do the tesla cards even support physx?

[Electronics Wizardy]

yes they would support it, but drivers might not let you, but no reason to, a geforce card is cheaper and would work well

[RadiatingLight]

It's so impractical I don't think anyone's even asked.

[eagerigor]

ok was just curious as the tesla cards are designed for maths calculations and the like so they would be good for calculating physics interactions or am i just completely missing the mark...?

[The Sloth]

3 minutes ago, eagerigor said:

ok was just curious as the tesla cards are designed for maths calculations and the like so they would be good for calculating physics interactions or am i just completely missing the mark...?

 

 

teslas are specially ordered custom cards mostly, you can buy normal teslas can be slow as a 730 or fast as a titan x sli also can vary in power. the tesla car uses tesla gpu for program learning. 

[eagerigor]

 

1 minute ago, nerdslayer1 said:

teslas are specially ordered custom cards mostly, you can buy normal teslas can be slow as a 730 or fast as a titan x sli also can vary in power. the tesla car uses tesla gpu for program learning. 

ahhh k thx

[Qwweb]

The current Tesla P100 interfaces over SXM2 or PCIe. Nvidia also only sells Pascal Teslas as full DGX1 HPC nodes starting at $129,000 or as blade units through Dell, Cray, or IBM (for supercomputers). It would work (probably) but be much more expensive than using a Titan XP. Also, very few programs/games support PhysX and many do worse with it enabled if you are using a modern GPU.

28 minutes ago, eagerigor said:

ok was just curious as the tesla cards are designed for maths calculations and the like so they would be good for calculating physics interactions or am i just completely missing the mark...?

The calculations that Teslas do have little to do with physics and more to do with plotting massive amounts of points and/or vectors. Teslas were also designed to be used in parallel with NVLink instead of PCIe(4 to 8 cards in one blade unit) and not serial (Think SLI or CFX) this is so they can be put into supercomputer blade units and run in parallel for near 1:1 scaling in specified workloads. Finally, keep in mind that the GP102-400 of the Titan XP, and the GP104-400 of the 1070 and 1080 are just cut-down and non-validated versions of the Tesla's GP102 and GP104.