GeForce 8 is going to get PhysX acceleration

From the ex PPU supporter perspective, isn’t it ironic how to this day there is not a single GPU physics title released but everyone cannot seem to help themselves from criticising the PPUs limited support?
 
The Ageia PPU is dead! Long live GPU physics!

It's pretty odd how things have turned out. Ageia's PPU is dead not because it was unsuccessful in the market (which it was), but because nvidia killed it like it did to ULi's products after ULi's acquisition.

Havok FX was stillborn, if it was ever considered to be born at all. ATI killed it's own GPU physics plans for the meantime, probably because Havok FX served as a warning of the lack of interest from developers.

I don't really see how nvidia will reignite interest in accelerated PhysX development unless it does the work for the developers itself (TWIMTBP). In the meantime, if the CUDA port of PhysX works well, the mostly unremarkable library of PhysX accelerated games will be opened to a larger audience. :p It would be too bad if this CUDA project turned into nvidia's GLide (which I guess it also owns). Maybe ATI is right and generalized GPU physics isn't coming until DX11. It's still the infancy of accelerated physics anyways.
 
AnnoyedDragon said:
From the ex PPU supporter perspective, isn’t it ironic how to this day there is not a single GPU physics title released but everyone cannot seem to help themselves from criticising the PPUs limited support?
Click to expand...

I fail to understand your bitterness. The PPU isn't dead, it's been reborn with more processing power than it could possibly imagine. PhysX has now got 128 stream processors at its disposal, how is this not good? The PhysX API will be implemented in GeForce 8 cards, and all current and future PhysX games will be hardware accelerated by the fastest video card out there. How is this not anything but good? Sure, the dozen or so early PPU adopters got burned, but they knew they would be going into it, and this is what the real mainstream gamers have been waiting for!
 
InorganicMatter said:
Sure, the dozen or so early PPU adopters got burned, but they knew they would be going into it, and this is what the real mainstream gamers have been waiting for!
Click to expand...
:p Dozen.

I wonder if nvidia will just bury the remaining stock of cards or liquidate them. I'm going to guess bury because they did that with 3dfx stock (nvidia had 3dfx cards pulled from shelves) and i'm sure nvidia doesn't want to support the cards any more than they have to.
 
InorganicMatter said:
I fail to understand your bitterness. The PPU isn't dead, it's been reborn with more processing power than it could possibly imagine. PhysX has now got 128 stream processors at its disposal, how is this not good? The PhysX API will be implemented in GeForce 8 cards, and all current and future PhysX games will be hardware accelerated by the fastest video card out there. How is this not anything but good? Sure, the dozen or so early PPU adopters got burned, but they knew they would be going into it, and this is what the real mainstream gamers have been waiting for!
Click to expand...
Oh I could add a little criticism to that minefield of questionable benefits you just made, but I was only really commenting on people making fun of Ageia’s game support while GPU physics still has none.
 
“Sure, the dozen or so early PPU adopters got burned, but they knew they would be going into it, and this is what the real mainstream gamers have been waiting for!“
Do we get burned? With a bit of luck its all the same API so we can use our PCI card without taking away power from the GPU.
 
Pottsey said:
“Sure, the dozen or so early PPU adopters got burned, but they knew they would be going into it, and this is what the real mainstream gamers have been waiting for!“
Do we get burned? With a bit of luck its all the same API so we can use our PCI card without taking away power from the GPU.
Click to expand...

Exactly my question- does this really leave our seperate PPU dead now? or will the same API still be shared.
 
Diseaseboy said:
Exactly my question- does this really leave our seperate PPU dead now? or will the same API still be shared.
Click to expand...

I imagine it will continue working for this version of the PhysX API, and will be obsoleted when a new PhysX API gets released. Kind of like how GeForce 7 cards don't support DX10.
 
Diseaseboy said:
Exactly my question- does this really leave our seperate PPU dead now? or will the same API still be shared.
Click to expand...
I think a clue to that will be what happens to the current PCI PPU. If nvidia immediately cancels it, I would imagine that driver updates for the card will probably stop soon after.

That's not an immediate problem, but as the PhysX library is updated, newer games may not work on the PCI PPU without updated driver support.

Or, nvidia may actually support the card. I wouldn't hold my breath after seeing what happened to 3dfx and ULI though.
 
Diseaseboy said:
Exactly my question- does this really leave our seperate PPU dead now? or will the same API still be shared.
Click to expand...

PhysX thus far has been about broad multi-platform compatibility, x86, PS2, XBOX360, and the PPU, it doesn't seem tied to any particular architecture.
 
GotNoRice said:
PhysX thus far has been about broad multi-platform compatibility, x86, PS2, XBOX360, and the PPU, it doesn't seem tied to any particular architecture.
Click to expand...

but no other hardware has run it so good as the PPU.
My quad screams under loads that the PPU just cruises through.
And I am talking 100% CPU on a quadcore, with 1-4FPS, against 15-20% CPU and 30-60 FPS with the PPU.
AFAIK the PS2 cannot run the PhysX API, Sony made a deal with AGEIA...but that was for the PS3...
The Xbox360...well look at gears of wars...the game enigie had to be cut down from the PC version to run
Consoles are overhyped in preformance.
 
This is because CPU's are general platforms, with lots of legacy components (it is just now getting to the point that modern CPU's can render games like Quake 2). Video cards and PPU's are dedicated components, which are very good at certain things. If the PhysX engine can be ported to CUDA, I don't see any reason the video card couldn't easily handle it.
 
Game developers should be more motivated than any other time then now to implement GPU based physics since it's now open to a large array of gamers.
 
BioSergeant said:
Game developers should be more motivated than any other time then now to implement GPU based physics since it's now open to a large array of gamers.
Click to expand...

Define large?
Oh, you mean anyone owning at 8x00 GPU or better?

That is ofcourse first AFTER the PhysX eventualy gets proted to CUDa...no ETA on that yet.
 
My question is this...

How will this effect FPS, if at all?
Secondarily, if it's going to effect the FPS, is this just another marketing idea to try and get more people to go SLi? (Since I'm sure that would be nVidia's "answer" or "cure")
 
First, the title of this thread is somewhat misleading.

The G92 series of GPU's are not being modified in any way, hardware wise, to provide PPU support. Instead, the actual PhysX software API is now the property of nVIDIA. This means that the previous work for GPU accelerated physics can continue. nVIDIA really wanted the Software API's (originally Novadex that Aegia then purchased). The only portion dealing with the hardware aspect I'd expect to see is the migration of the PPU R&D team to the GPU teams. Primarily a way to ensure that GPU phsyics continues to make some headway.

Havok had been purchased by Intel, they had quashed the HavokFX portion that was the software API designed to accelerate certain features with the GPU. Intel is more focused on the multi-core CPU physics acceleration in order to encourage future CPU sales.

So initially expect the similar results, games that use the PhysX API to add.... more debris and fancy explosions once ported to CUDA. As for true immersive physics interactions, expect the level to progress much further in the short term. Game developers are going to need another 6 months to a year with that development kit with the CUDA support is released for their game titles based on that tech. Also expect ATI to license PhysX tech from nVIDIA due to their also being locked out of Intel's Havok Physics API's. Long term, this is a good move for all gamers, short term you'll notice little differences than the current gaming market and the "effects" physics versus the phsyical interaction physics.

Now the physics argument will change from PPU versus GPU to GPU versus mult-core CPU as game developers decide on using PhysX or Havok for their software APIs.
 
RLich8 said:
My question is this...

How will this effect FPS, if at all?
Secondarily, if it's going to effect the FPS, is this just another marketing idea to try and get more people to go SLi? (Since I'm sure that would be nVidia's "answer" or "cure")
Click to expand...

When using the CUDA functions, once completed, the use of in game physics will have some reduction in FPS. This is due to physics calculations being different in function than the graphics functions. It's one thing to have the stream processors all working on different functions for the final display frame (or multiple frames), but PPU functions would have to be performed in between.

Likely there will be an adjustable range in the game to select the level of physics processing. For those that go SLI, or CrossFire (as I believe it will be licensed to AMD as well) the reason to add an additional GPU beyond the second would be for physics. As you've seen recently in the CrossFire and SLI demonstrations of more than two GPU's, there is significantly reduced returns on additional GPU's. However, this could change if the third or fourth GPU is being utilized exclusively for physics.

So, yeah, it's marketing to encourage additional card purchases, especially in light of the significant reduction in FPS returns beyond two GPU's.
 
You must log in or register to reply here.
Back
Top