• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Will nvidia ever enable dual-GPU (or more) Physx support?

W

Wag

Diamond Member
You may think this is useless now. But I imagine many of their next gen of cards will be dual-gpu, and as they increase in number you would think many people would want to use the whole power of their card for Physx.

I for one would think that at least those of the current generation who own a GTX 295 and decide to wait out their next bleeding edge card (GT300- DX11, dual-GPU, etc) might want to use their current card as their Physx card. If possible, why waste a single GPU, why not use both?
 
"Will nvidia ever enable dual-GPU (or more) Physx support?"

Almost certainly.* The real question is when.

*Assuming PhysX survives that long
 
We can all hope that PhysX goes the way of the dodo, and we don't have to worry about this anymore. With 6-8 cores around the corner, and most games run fine with 2, let's just use the extra CPU cores for this and not waste GPU piower (or an extra add-in card) for this.
 
Originally posted by: ExarKun333
We can all hope that PhysX goes the way of the dodo, and we don't have to worry about this anymore. With 6-8 cores around the corner, and most games run fine with 2, let's just use the extra CPU cores for this and not waste GPU piower (or an extra add-in card) for this.
Click to expand...

We can all hope HW accelerated physics starts to use a vendor independent standard, so it's available to all. We can also hope that games are able to make the best balanced use of our gpu's and cpu's to give us the best experience. This will probably mean the gpu does most of the physics work because it's an order of magnitude better at it then the cpu (even with more cores).
 
Considering it's been confirmed that the bigger upcoming games will support some sort of CUDA Physx support (and possibly PPU) on the PC:

Batman: Arkham Asylum
Borderlands
Dragon Age: Origins
Darkest of Days
Dark Void
Mafia II
etc

It doesn't look like PhysX is going away all that soon. That's why I ask. It seems more and more developers are jumping on the PhysX bandwagon (for whatever reason). I plan on purchasing many of these games and I definitely plan on purchasing nvidia's highest-end card of the next gen. I really hope they implement multi-GPU PhysX support on a single card (at least) by then.
 
Originally posted by: Dribble
Originally posted by: ExarKun333
We can all hope that PhysX goes the way of the dodo, and we don't have to worry about this anymore. With 6-8 cores around the corner, and most games run fine with 2, let's just use the extra CPU cores for this and not waste GPU piower (or an extra add-in card) for this.
Click to expand...

We can all hope HW accelerated physics starts to use a vendor independent standard, so it's available to all. We can also hope that games are able to make the best balanced use of our gpu's and cpu's to give us the best experience. This will probably mean the gpu does most of the physics work because it's an order of magnitude better at it then the cpu (even with more cores).
Click to expand...

@Exar 6-8 cores? For parallel tasks like physics, GPUs basicly have 200+ cores and as the physics added to games gets less trivial one can assume it's going to matter.

Whether it's microsoft, intel, or nvidia (or some different named evolution of such) that gets adopted as the standard it's all the same. Physx atm probably has the best chance of becoming the standard since there's developers jumping onboard, and you can bet your ass if it gets a strong enough foothold AMD will change their we don't want to deal with nvidia and instead suck up to intel stance, which is probably more of a delay tactic than anything else. Physx is only vendor specific because no other manufacturers are adopting it. At least some software developers are though so there's a chance AMD/Intel will end up being forced to adopt it in the future.
 
Originally posted by: ExarKun333
We can all hope that PhysX goes the way of the dodo, and we don't have to worry about this anymore. With 6-8 cores around the corner, and most games run fine with 2, let's just use the extra CPU cores for this and not waste GPU piower (or an extra add-in card) for this.
Click to expand...

CPU cores cant do what a GPU can do. I wouldnt be surprised if an 8400 destroys an i7 920 in any kind of real workload centered around physics. 8400 costs under 100 bucks. i7 920 requires an entire platform that will be north of 400.

I believe a lot of the physics workload will be done by functional units that wont be doing any rendering? It is possible Nvidia will release a monster of a chip that wont be much faster at gaming than the smaller ATI part. But start feeding it calculations and watch it explode.
 
Originally posted by: aka1nas
If you have SLI, you can already have PhysX load-balanced along with rendering on the SLI set.
Click to expand...
I think you misunderstand.

1. New nvidia card comes out that's really fast, I decide to upgrade and keep my 295 as PhysX card.
2. Now if I understand correctly, the way nvidia currently implements PhysX, if I use my 295 I can only use a single GPU of the two on the card and the other will go to waste. Or am I incorrect?

 
Originally posted by: Wag
Originally posted by: aka1nas
If you have SLI, you can already have PhysX load-balanced along with rendering on the SLI set.
Click to expand...
I think you misunderstand.

1. New nvidia card comes out that's really fast, I decide to upgrade and keep my 295 as PhysX card.
2. Now if I understand correctly, the way nvidia currently implements PhysX, if I use my 295 I can only use a single GPU of the two on the card and the other will go to waste. Or am I incorrect?
Click to expand...

Dual-GPU cards might be a weird corner case, but I think you just assign a specific display adapter to PhysX from the NVCP. Does the 295 usually show up as two GPUs in the NVCP?
 
Originally posted by: aka1nas

Dual-GPU cards might be a weird corner case, but I think you just assign a specific display adapter to PhysX from the NVCP. Does the 295 usually show up as two GPUs in the NVCP?
Click to expand...
Just one. And as I understand it it only uses one of the two GPUs even if you use a 295 as a PhysX card.

 
There is already scaling of PhysX up to a GTX 280.

http://www.hardforum.com/showthread.php?t=1420948

The problem would be writing the drivers to use more than 1 GPU at a time. Quad-SLI still doesn't give a performance boost to 3D Vision yet. I wonder what kind of CPU overhead is involved with SLI graphics and SLI physx? Or even quad-SLI graphics and SLI physx. lol An i7 might be required for all of that overhead especially it's QPI.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top