• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Physics X - ... A few questions...

It's getting to the point where individual particles themselves need to be calculated. If we can accurately calculate the airflow over a section of wing then upsize that abit and there we go.

However if a Quad takes a day to calculate one section of wing we need a huge performance speed up :p
 
Concorde Rules said:
It's getting to the point where individual particles themselves need to be calculated. If we can accurately calculate the airflow over a section of wing then upsize that abit and there we go.

However if a Quad takes a day to calculate one section of wing we need a huge performance speed up :p
Click to expand...

It's the rigid body stuff that is so difficult to implement well from a gaming engine perspective, particle based methods such as cloth and ejection are already almost as good as they need to be to suspend disbelief, and they are designed to scale well, it just takes more processing power. However, whether its running on a GPU or a CPU is irrelavent from the devs POV, from my own research PhysX is currently one of the best in terms of it being a library, with Bullet and Havok closely behind.

However, what NV really need to do is stop going down the CUDA route and go straight to OpenCL with SDK 3.0. Having said that though it's looking more and more likely CUDA will become inter-operable with OpenCL. Assuming this is the case, I see no reason a CUDA based HW accelerated PhysX library couldn't also work on OpenCL libraries also...
 
Last edited:
Would I buy a dedicated card for PhysX now? No.
Would I buy a dedicated PhysX card in the future if it developes well? Yes.
Would I keep an old nVidia card for dedicated PhysX processing when upgrading? Yes, it's what I have done.

My GTX285 is my main card and I use my old 8800GT as my dedicated PhysX card. I have not played a game to date that utilises it yet but it's there for when I do. Borderlands might have some PhysX stuff and that's on pre-order :)
 
coupe69 said:
Let's face it Kyle, you'll never invest in Nvidia fullstop :p;)
Click to expand...

That's the point, ;) I was considering buying a cheapo/secondhand 9600 or something to use as a PhysX card as I'm interested in it, but nVidia are doing all they can to make it hard for themselves.

Now they've went and disabled PhysX in the drivers when it's not an nVidia card doing the rendering, no chance of it, and as for buying a "highend" nVidia card, they've not got anything going for them at the moment, and I'm not prepared to pay their rip off prices and support their crazy self destructive ways by giving them my monies.:o
 
Well PhysX can be utilised by anyone on any mobo with 2x pcie slots easily enough without much of a performance hit. Bung in an 8800GT and boom headshot - it's what I'm doing tonight then continuing on with batman!
 
Update on 1920x1200 4xAA Batman results:

Code: 
No PhysX, GTX260:
Min	Max	Avg
50	113	84.363

HIGH PhysX, GTX260:
Min	Max	Avg
19	61	39.486

HIGH PhysX, GTX260 + 8800GT
32	87	61.187

£40 or the 8800GT to get a 22fps boost and nicer visuals = Worth it :)
 
I'll have to test again but that PhysX hit seems way more than what I saw when I tested... tho I was concentrating on CPU v GPU physics at the time rather than no PhysX v GPU PhysX.
 
mrk said:
Update on 1920x1200 4xAA Batman results:

Code: 
No PhysX, GTX260:
Min	Max	Avg
50	113	84.363

HIGH PhysX, GTX260:
Min	Max	Avg
19	61	39.486

HIGH PhysX, GTX260 + 8800GT
32	87	61.187

£40 or the 8800GT to get a 22fps boost and nicer visuals = Worth it :)
Click to expand...

Just tested with my setup - 2048x all settings maxed, 4x AA, 16x AF:

No PhysX, 260GTX SLI:

Min Max Avg
54 117 89

High PhysX, 260GTX SLI:

Min Max Avg
37 94 65

Bigger impact than I thought from physx... also wondering if the results are CPU limited.

EDIT: Overclocking the GPU even more doesn't seem to have any effect on the fps so I'm guessing its CPU limited.
 
Last edited:
PhysX is reliant on processing cores, not their speed as such so overclocking GPU won't help in the math needed to do physics hence why offloading those cores to another card gives the biggest boost!
 
Faster cores will get the job done faster as well as the parallelization.

I was OCing the GPU tho more to see what happened to the non physx one as they seem a bit low.
 
5870CF 1920x1200 4AA16AF MAX settings, Physx off
shippingpcbmgame2009101.jpg


5870CF 1920x1200 4AA16AF MAX settings, Physx max 9600GT @ 600/900
shippingpcbmgame2009101.jpg


5870CF 1920x1200 4AA16AF MAX settings, Physx max 9600GT @ 675/1000
shippingpcbmgame2009101.jpg
 
Woo nice fps... I'm guessing that shoots down my CPU limited theory unless your running a 4.5+gig i7 and the game makes mad use of the HT units.
 
Hmmm SLI seems to be only giving 8% scaling in the benchmark :( hence the poor results...

With a little more digging it seems that PhysX is executed on the primary GPU in SLI mode first, then the rendering which kills SLI performance :( prolly using another GPU for PhysX wouldn't make much odds...
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom