[ChuckNorris89]

For the younger crowd, back in the day before Nvidia bought Ageia, you'd have to buy dedicated PhysX accelerator cards and stick them in your PC in order to use this tech in the few available games that supported it. Wild stuff.

After that, PhysX API was accelerated via CUDA and the dedicated PhysX ASICs were discontinued.

https://www.techpowerup.com/review/bfg-ageia-physx-card/

[rasz]

Dont forget to mention it was pure scam, PhysX PCI "accelerator" actually ran those supported games slower, sometimes Two times slower, than software.

https://www.anandtech.com/show/2001/4

[dagmx]

You miss the important bits:

> As we said before, installing the hardware automatically enables higher quality physics. We can't get a good idea of how much better the PhysX hardware would perform than the CPU, but we can see a couple facts very clearly.

So what you call a scam is down to implementation details. In a true like for like scenario the PPU would usually outperform the CPU implementation as long as your sync boundaries were clean to get stuff back from it, and pipe the transforms back to the GPU.

I wish people like yourself wouldn’t reach for incendiary language by default. It ruins any and all nuance in discussion.

[sillysaurusx]

A more nuanced description is that in most cases, it was slower, and it took immense developer effort to even try to make it faster. And the cases that were accelerated were generally things that consumers didn’t care much about, e.g. fluid dynamics.

The software is the clear winner of that. The team was fantastic, and one of them (John Ratcliff) helped kickstart my career. So it isn’t entirely off base to say that the PPU was… exaggerated.

[rasz]

Its a game, the important bit is playable framerate. There arent any nuances when you are hitting drops to 17 fps.

>down to implementation details

Game is running 2 times slower. Nvidia favorite kind of technology. Tessellation, Gameworks, Hairworks, RTX, all run at lest 2x slower when enabled unless you buy Nvidia flagship hardware.

Did you know Nvidia physx library shipped compiled for FPU? At the time SSE2 was available for 10 years and SSE3 for 5 https://arstechnica.com/gaming/2010/07/did-nvidia-cripple-it... https://www.realworldtech.com/physx87/3/ Nvidia PR person tried claiming Physx actual computation is not a bottleneck, so they didnt bother optimizing :D If its not a bottleneck then why run it on GPU?

[gabereiser]

Even after the acquisition, people still used dedicated PhysX cards before multi-monitor screen stitching made SLI graphics visual based. Prior, you’d have a card in SLI to handle the physics for PhysX and the other for the visual rendering.

This is great news. Bullet and others could benefit from this as well.

[pacaro]

We used the PhysX engine in a simulated environment for robotics development back in that timeframe. I don't think more than one or two people on the team had the accelerator cards, but it worked surprisingly well just on a vanilla hardware

[jansan]

Interesting. What was the target group of these cards? Gamers or engineers and scientists?

[driscoll42]

Mostly gamers, here's a review from Arkham Asylum which showed how offloading PhysX to a GT 220 paired with a GTX 260 would help a fair bit with minimum and average FPS - https://www.tomshardware.com/reviews/batman-arkham-asylum,24...

[the_pwner224]

This is from later (2012) but shows some of the effects that PhysX enables in games. The "apex turbulence" in the second half of the video is pretty cool.

https://www.youtube.com/watch?v=n5qhaEghJ74

[dagmx]

It started out initially for academic usecases till the costs came down and it was then pushed for games.