Thanks [H]

Status
Not open for further replies.

Mister E

2[H]4U
Joined
Sep 14, 2004
Messages
2,661
For linking the physics section to Video Cards again. I hope both companies can still do something incredible with this tech.
 
Doesn't this really belong in nVidia section? or does ati do physics

i think it should remain separate, even though nvidia is now responsible for supporting the ppu hardware, because it is still a piece of discrete hardware separate from gpu physics, and the ppu card itself was designed by ageia, not nvidia. as far as ati, they do not currently do gpu physics atm. in the near future, who knows?
 
I liked it when it was in it's own section... seeing how physics could be processed via CPU, PPU, or GPU. But since the shift was made from the dedicated PPU to a GPU solution I can see why the change was made to put it in with the Video Card Section (and that the forum was not all that active). However, if it was put back in it's own section for any reason it would be welcome by me (I miss being able to see new posts from the main forum page).
 
Now we only need an Intel section in Videocards, for when Larrabee hits ;)
 
I liked it when it was in it's own section... seeing how physics could be processed via CPU, PPU, or GPU. But since the shift was made from the dedicated PPU to a GPU solution I can see why the change was made to put it in with the Video Card Section (and that the forum was not all that active). However, if it was put back in it's own section for any reason it would be welcome by me (I miss being able to see new posts from the main forum page).

Strictly taken there is little difference between a GPU and PPU. Both are massively parallel vector processors. It is far more cost-effective to use one design for both tasks, so I doubt we'll ever see dedicated PPU cards again (aside from tweaked GPU designs, like nVidia's Tesla cards).

So I'm all okay with moving this forum to the videocards section :)
 
Strictly taken there is little difference between a GPU and PPU. Both are massively parallel vector processors. It is far more cost-effective to use one design for both tasks, so I doubt we'll ever see dedicated PPU cards again (aside from tweaked GPU designs, like nVidia's Tesla cards).

So I'm all okay with moving this forum to the videocards section :)

Um... not sure why you quoted me. You basically restated points I made without actually agreeing or disagreeing.
 
I rather see it in een seperate section.
physics is physics and first done by CPU. some time ago also a PPU now GPU.
So it's a CPU PPU GPU thing.

instead PPU it could be re named Physics Computing CPU PPU GPU.
But because physX is hot these day's and linkt with nV and its GPU. it's now in the GPU section.
Altho it's not my preference I don't mind. There are con's and pro for both.

As for nV. It's a realy bad thing that physics middle ware is owned by iNtel en nV. this hold back innnovation and market adoption.
if Havok and physx where independand we gamer get more of it sooner then later.

Dev hav enow to doeel witch Middle ware to choose. Havok no GPGPU support on perpous. iNtel killed HavokFX. PhysX. You must have a CPU only path and a GPGPU.
With havok FX. Dev could go for low GPGU lod to high. would stimulate gameplay PhysX adotion by dev's.

As for nV nextgen GPU hope that the put a higher unified shader ratio to rops and TMU because of PhysX.

Also the bad thing of GpGPU for Physics. The shaders are halve the diesize. But the TMU rops etc are always idle power drawers.

I wonder if on tesla these parts are fully disabled.

The TMU and rops are bundled with in a shader block. So not easy to disable if they share ondie power traces. And arent seperated bij a modular system.
 
It is staying here for now. Things are always subject to change in reaction to the market and technology.
 
Status
Not open for further replies.
Back
Top