darry wrote on 2025-03-22, 04:53:

I don't disagree, but I'm still hoping for a wrapper,

The peak of hardware accelerated PhysX usage was in 2010 (after Intel released software based Havoc)
At that time AMD had about 50% of the gamers and many AAA games were released.
If it wasn't created then, it certainly won't be created now.

Also, Arasaka Nvidia corp is not very friendly to open source software.
I think the first attempt to create a repository on GitHub will get a strike.

shevalier wrote on 2025-03-22, 09:38:

Also, Arasaka Nvidia corp is not very friendly to open source software.
I think the first attempt to create a repository on GitHub will get a strike.

They already did that themselves. https://github.com/NVIDIA-Omniverse/PhysX

SDK and documentation are there. With copyright saying you can use and distribute binary if you include the copyright notice.

shevalier wrote on 2025-03-22, 04:44:

...Well, no one is complaining that Microsoft doesn't make more 32-bit systems.

Well, but everyone would complain if legacy 32-bit code completely stopped working on current 64-bit Windows systems... (The PhysX problem is more analogous to this situation.)

shevalier wrote on 2025-03-21, 16:45:

GemCookie wrote on 2025-03-21, 16:04:

The RTX 5090 doesn't leave much room for a second video card, especially on recent motherboards.

So it's time to do as Mondodimotori does 😀

Mondodimotori wrote on 2025-03-21, 15:22:

I'm planning in getting either a 1060 or a 1070 for my Win7 retrobox.

PLS Don't! I'm alredy picturing hoardes of people going after these cards, and prices skyrocketing towards the stratosphere! At least wait for me to get the card.

*hoards GTX 980 Ti, which also works in Windows XP*

On March 25th NVIDIA updated the repo. The GPU simulation kernel source was added, along with the Flow SDK all fully under the BSD-3 license.

With this is it should be possible to port from CUDA over to OpenCL or some other technology that's supported on AMD and Intel GPUs.

https://github.com/NVIDIA-Omniverse/PhysX/discussions/384

Nvidia: "I don't want to play with you anymore"

NVIDIA doesn't give a shit, they're AI company now, right?

Why drop 32-bit PhysX/CUDA specifically, though? They could drop entire 32-bit support, if they wanted, right? No more 3D for 32-bit apps!

UCyborg wrote on 2025-04-07, 11:16:

Why drop 32-bit PhysX/CUDA specifically, though? They could drop entire 32-bit support, if they wanted, right? No more 3D for 32-bit apps!

Probably because those libraries require the biggest workarounds to get 32-bit apps talking to a 64-bit driver, especially with regards to memory allocation/sharing.

jmarsh wrote on 2025-04-07, 11:29:

UCyborg wrote on 2025-04-07, 11:16:

Why drop 32-bit PhysX/CUDA specifically, though? They could drop entire 32-bit support, if they wanted, right? No more 3D for 32-bit apps!

Probably because those libraries require the biggest workarounds to get 32-bit apps talking to a 64-bit driver, especially with regards to memory allocation/sharing.

Yeah, ^This.
CUDA it's a whole developement environment. I'm a 3D and Rendering artist, and CUDA acceleration was pretty swell if you have the right hardware.

Of course, I also haven't used a 32bit CUDA application in more than 15 years. No point using 32bit applications when RAM is of the essence. The second I got a PC with 4GB of RAM in 2009, it was all 64bit for me. Only games remained the exception for years.

Deleted off topic post