• 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.

PhysX87: Software Deficiency

What I do find quite odd is that that article specifically states that PhysX doesn't make use of SSE intrinsics, yet, I have the source code (bar the integration core), and when I compile my own versions of the library I know for a fact that there are system specific SSE compilation pathways in the headers.

Odd.

Having said that, I think rather than it being a conspiracy, it's more a case that the code is indeed now a mish mash of old stuff from over a decade ago, mixed in with countless bug fixes from the past 10 or so years of the libraries development.

Supposedly SDK 3.0 was going to bring a totally new, cleaned up and CUDA compatible code base. That was sometime around a year ago that was first due...

Who knows what's going in inside NVIDIA tbh.

EDIT: I should also say that, if PhysX's CPU code (i.e. most of it...) is so poor, then why, as a library, does it perform as well as all the other physics engines and in some cases better. I have absolutely no doubt that the code could be better written now, but as for all this NVIDIA are deliberately being evil tosh, it's just not true, they just aren't perfect in what they are doing, far from it in fact.

If anyone wants to have a look into the workings of these engines, I suggest pulling apart Bullet and perhaps ODE. The way that Bullet is put together is quite similar to PhysX, albeit its a bit more elegant in places and on the bleeding edge more often. Code wise however, they are both of a similar standard.
 
Last edited:
Hmmmmmmmm, I just noticed something about those Mafia 2 specs. It says a 480 and 285 for max, but it says that you can do that on Win XP? I wonder what Directx version the game uses.
 
That’s a good marker to show where each companies (Nvidia & ATI) priorities’ are at. Nvidia is determined to push and market it’s own proprietary IP whereas ATI want to push developers down the DX11 and open CL route.
 
Freddie1980 said:
That’s a good marker to show where each companies (Nvidia & ATI) priorities’ are at. Nvidia is determined to push and market it’s own proprietary IP whereas ATI want to push developers down the DX11 and open CL route.
Click to expand...

Cool that DX11 is non proprietary now , awesome!
 
Rroff said:
Depending on the game environment - I've not found bullet to work very well in typical game useage (slows down pace for the player or becomes clunky/awkward when they have to interact with the physics). I'm using a modified version of ODE myself but its a little prone to "exploding" if I'm not really careful - but its a bit easier to implement into the game flow.
Click to expand...

I can't say I've ever used ODE myself. Once I found Bullet I kinda stopped looking for other physics libraries.
 
Freddie1980 said:
That’s a good marker to show where each companies (Nvidia & ATI) priorities’ are at. Nvidia is determined to push and market it’s own proprietary IP whereas ATI want to push developers down the DX11 and open CL route.
Click to expand...

I don't know of any OpenCL GPU accelerated games, I own several PhysX games. (Metro2033, BatmanAA, Cryostasis, Mirrors Edge, Darkest of Days, Dark Void, GRAW, Warmonger, UT3)

Intel seems to have buried Havok GPU accelerated physics with Larrabee.

Makes ATi's decision not to buy Havok and Ageia when they were considering it seem like a bad move in retrospect, if they wanted to have hardware accelerated physics for their customers. If they don't care, they've certainly published a lot about it since 1900XT days.
 
Isn't the point that the games you mention have nothing in them that couldn't be done on a decent CPH

ow come with hardware PhysX, you still need a great CPU? Looking at mafia 2 specs for example, you need an i7 for full physics even though the 285 should be doing the work.
 
Last edited:
metalmackey said “Isn't the point that the games you mention have nothing in them that couldn't be done on a decent CPU. “
That’s not really true is it? I have yet to see any games with full dynamic cloth that tears and liquids that run decently on the CPU. But I have to admit I have not been keeping up on CPU physics in recent games. It was certainly true a few years back then CPU's could not do the work at playable speeds

The only problem I see with PhysX is that it’s locked to NVidia. What we need is a hardware physics API that works on all GPU’s. What they should do is add PhysX to DirectX and let everyone fully use it. Or come up with a new hardware physics API that everyone can use.
 
Jethro Bodine said:
I don't know of any OpenCL GPU accelerated games, I own several PhysX games. (Metro2033, BatmanAA, Cryostasis, Mirrors Edge, Darkest of Days, Dark Void, GRAW, Warmonger, UT3)
Click to expand...

To be fair the OpenCL standard was only ratified in the middle of last year. With game development cycles being at least 18 months I wouldn't expect any until the end of this year.

Most games take longer than 18 months to develop so it will be a slow trickle at first.
 
Pottsey said:
metalmackey said “Isn't the point that the games you mention have nothing in them that couldn't be done on a decent CPU. “
That’s not really true is it? I have yet to see any games with full dynamic cloth that tears and liquids that run decently on the CPU. But I have to admit I have not been keeping up on CPU physics in recent games. It was certainly true a few years back then CPU's could not do the work at playable speeds

The only problem I see with PhysX is that it’s locked to NVidia. What we need is a hardware physics API that works on all GPU’s. What they should do is add PhysX to DirectX and let everyone fully use it. Or come up with a new hardware physics API that everyone can use.
Click to expand...

You are right, I cant think of any with tearable cloth and dynamic liquid either. But of course, theres sooo many games with PsysX that have it, oh wait... 3 at the most LOL
 
It doesn’t matter how few they are. The point is the CPU’s cannot handle the physics and until we get widespread hardware physics we are stuck with outdate poor physics. There are so few games with full full dynamic cloth that tears and liquids because we are stuck with CPU’s. There are also more than 3 games that use those effects.

If everyone had GPU hardware PhysX or some other hardware physics then game physics would be much better then today and we would have far more interesting games. The CPU is holding us big time as far as physics goes.
 
You must log in or register to reply here.
Back
Top Bottom