Physx vs OpenCL

[Thor.]

It looks like Opencl is far superior to Physx and by the looks of it does a have a unified development kit. Now why are developers not utilizing it??

 

http://www.digitimes.com/news/a20130121VL203.html

 

maybe because its to new at the moment or they just Prefer Physx.

 

http://www.tomshardware.com/forum/334290-33-running-physx-natively-card

 

So what do you think of this possible new competition, there is a choice now :thumbsup:

[ZeroKing]

Even if OpenCL/DirectCompute is better, it needs to be more accesible commercially and tools to have, and OpenCL doesn't have a fully developed SDK yet. Plus, OpenCL's implement is portable, which is not necessarily good as you need GPU-discrete implementation for it to be useful. Programmers can't do that, it's inherent in the language and execution pipeline. CUDA doesn't have that problem, a programmer can easily offset calculations in either the GPU, CPU, or both (it means that performance is faster when data is transferred from CPU-GPU).

 

The advantage to OpenCL is that it's open standard, which is a huge plus, and that games perform nicely with OpenCL execution (tested some beta indie games before). The language, despite its rather closed pipeline execution, is rather easier than implementing on CUDA (which is strictly NVIDIA). So for that, I'd vote OpenCL.

Edited July 30, 2013 by ZeroKing

[Aegrus]

That'd be nice; more competition means everyone has to improve and the consumers win. Also, it'd be nice if it was easier to run than PhysX.

 

However, I'm not sure how you get "far superior" out of this. In fact, your links say nothing specific about Open CL's capabilities at all other than that it exists and it's physics.

Edited July 30, 2013 by Aegrus

[Thor.]

They where arguing the fact they could of optimized more for cpu side then having to use discrete gpu hardware. If they wanted to make it a little more efficient they could of used see2 or hyperthreading, or even optimize better fro quads or 8 core cpu's to take the load off of the gpu, they didn't do that do to a business decision. AMD in the other hand will be more friendly to the cpu vs gpu. And optimized for cpu side as well.

Sense its a open platform i could see it being even a more greater threat to Physx, because it can be optimized via third party.

Edited July 31, 2013 by Thor.

[Rennn]

I'll take whichever one offers the more advanced and optimized physics solution. At the moment, that's PhysX, which is why I voted for it.

If OpenCL gets a better development kit and outdoes PhysX in more games, I'd be happy to support that instead.

 

Right now, however, I'm looking at Last Light, Borderlands 2, and Cryostasis as physics powerhouses. All are powered by PhysX.

OpenCL... merely has the Bullet tech demo, and potential for expansion.

 

Red Faction Armageddon actually uses a form of outdated Havok physics, and it's truly impressive as well.

Edited August 1, 2013 by Rennn

[JimboUK]

I'll take whichever one offers the more advanced and optimized physics solution. At the moment, that's PhysX, which is why I voted for it.

If OpenCL gets a better development kit and outdoes PhysX in more games, I'd be happy to support that instead.

 

Right now, however, I'm looking at Last Light, Borderlands 2, and Cryostasis as physics powerhouses. All are powered by PhysX.

OpenCL... merely has the Bullet tech demo, and potential for expansion.

 

Red Faction Armageddon actually uses a form of outdated Havok physics, and it's truly impressive as well.

 

It's mostly being used for intensive applications, Final Cut Pro X, Vegas 11, Powerdirector, VideoStudio, Magix, The Gimp and CS6 all use it.

[Sarissi]

OpenGL is cross platform (OS) and has been in use for years. Its primary use is for VFX applications and only for preview. Actual rendering is done by the CPU, and normally in a render farm.