• 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.

Ati physics news

Man of Honour
Joined
13 Oct 2006
Posts
83,091
Last time I checked bullet it was quite rudimentary compared to havok and physx and had quite a lot of bugs and underlying bad coding... tho for simple stuff it worked very well.

This isn't really a good move - the vast majority of developers with an interest in physics already prefer and have a lot of experience with havok and physx has a reasonable following also and is much more suited to many games implementation wise.


EDIT: From the look of those slides... AMD/ATI really don't _get_ hardware physics - weird - they are trying to implement it as if it was just another graphical effect theres much more to it than that and without a grasp of that concept things aren't going to go well.
 
Last edited:
Soldato
Joined
29 May 2005
Posts
5,622
Location
West London
Last time I checked bullet it was quite rudimentary compared to havok and physx and had quite a lot of bugs and underlying bad coding... tho for simple stuff it worked very well.

This isn't really a good move - the vast majority of developers with an interest in physics already prefer and have a lot of experience with havok and physx has a reasonable following also and is much more suited to many games implementation wise.


EDIT: From the look of those slides... AMD/ATI really don't _get_ hardware physics - weird - they are trying to implement it as if it was just another graphical effect theres much more to it than that and without a grasp of that concept things aren't going to go well.

AMD are supporting OpenCL....they're not directly supporting Bullet etc.

What's your views on the latest revelation on PhysX in Batman? Nv's deliberate crippling of CPU effects?
 
Caporegime
Joined
18 Oct 2002
Posts
33,188
AMD are supporting OpenCL....they're not directly supporting Bullet etc.

What's your views on the latest revelation on PhysX in Batman? Nv's deliberate crippling of CPU effects?

Its all they've been doing from the start frankly, general easy and normal progression completely halted by shifted to a harder to code slower thing so people think they need hardware to use it.

As shown with Batman, their biggest title to date, what they've done can indeed be done on a cpu, easily, with no framerate hit. They are paying people to forcibly kill performance so without messing with the game, which most people don't know how to do, only their hardware will run it fast enough.

It isn't even fantastic effects, just Nvidia paying people to remove utterly normal(at this point) effects and sabotage the competition.

I've yet to see Physx move on in game physics, at all, from years ago. Fullscreen video's of cloth effects, yet in game I've seen nothing that blows away what other games have done. Yes when you focus the entire power of a gpu on one effect on one piece of cloth in a demo you can make it look really pretty. But in game, with other things, and no games being called "laundry day" its done nothing for us at all except hold the industry back.

As for Bullet, its just ONE OF the many things open CL will support, and it is used, as shown, its used quite widely and the point being that if ALL the physics API's except Physx use open CL, open source and free, then the massive majority of the market supports accelerated physics on any cards that use Open CL.
 
Man of Honour
Joined
13 Oct 2006
Posts
83,091
AMD are supporting OpenCL....they're not directly supporting Bullet etc.

What's your views on the latest revelation on PhysX in Batman? Nv's deliberate crippling of CPU effects?

I think someones messed up their configuration... althought I do have to say their CPU implementation seems a bit slower than it should be...

But without _any_ tweaks, physics set to high and GPU acceleration disabled in the control panel I'm seeing good loading across all 4 cores on my Q6600 with increased useage on cores 1 and 3 in areas with heavy physics and a marginal increase on core 2 - core 0 seems to be unused by physics but thats not unusual... not seeing any fps drop below 20 - tho IMO theres not enough physics effects to drop below double that. (Infact similiar physics on my own engine which uses the tokamak source as a base - doesn't drop below 80fps on the CPU).

With GPU physics enabled, all settings maxed at 2048x with 4x AA and 16x AF I don't drop below 60fps except when the engine is loading new data in/out.

EDIT: Not sure if my fps comparisons are fair tho as I used windowed mode so I could check the core usage with external software - which may have disabled SLI and window mode is often lower performance than fullscreen which is where all my own tests were done.
 
Last edited:
Top Bottom