*spinoff* PhysX relevance

Is it just me or am I the only person that still yawns over the physics thing?

 

PhysX is currently the "Pimp my ride" of the GPU world. the huge spoiler will make it look cool but it won't do anything but drag down performance.

 

Say's poster who posts the first post on PhysX in a 4770 review thread...

 

I think the same reason people knock on PhysX is because it's the Glide API of this generation.

For the same reason people knocked Glide when OGL and D3D became capable people will be knocking PhysX when OpenCL/Havok/DX11 become capable of GPU accelerated Physics.

Likewise there is no game so far that shows an actual speed increase when enabling PhysX. Developers (prompted by Nvidia maybe?) are making the PhysX enabled features so IN YOUR FACE that rather than enabling more features at the same speed or a faster gaming experience, they are instead overdoing it and slowing down the games.

Take Sacred 2 for example. They could have easily done with less leaves blowing around constantly and had the framerate not take a huge hit (presumably). But rather than enhance the game and keep the same speed. They had to make it so there was no way you could possibly miss the fact that it is there. And in the process the game takes a huge hit in performance.

The conspiracy theorist might say they did this on purpose to make the CPU rendering of it even more horrendous than it needs to be. Or that it's purely driven by marketing, and as such it's more important to make it noticeable than to have an overall positive impact on the game.

But even with all that, just the fact that it is currently the Glide API of this generation is going to make it unpopular with some people. Except while Glide not only made things look better, it also made things run faster (until OGL and D3D caught up). So far PhysX generally makes things look better, but also makes them run slower.

Regards,
SB

 

Glide and the awesome hardware behind it in the mid '90s was what made 3D gaming possible on PC for the most part. Few cared about full OpenGL for realtime rendering for most of the '90s. The drivers for it sucked and it wasn't really suited for the very limited accelerators of the day. Early MS D3D was a big bad joke. Glide had a lot going for it during its heyday.

Physx is, so far, very much a gimmick. What's there to like, exactly? Hyped promises for fancy future stuff? The Physx stuff I've personally tried in UT3 on my 8800GTX was enough to convince me that Physx is not worth making a card judgment over. It was just all so superficial and added little to the game other than "physics eye candy".

I think GPU accelerated stuff like Physx and GPGPU might become useful when all cards support it through DX11 or OpenCL. And when the cards are simply more powerful than they are now.

 

That's true but there was also this...

Giving people a reason to complain even when there isn't any competition. Just like people complained when PhysX was owned and operated by Ageia. It has yet to either...

1. Speed up performance with same level of eye candy
2. Provide greater eye candy with same or similar level of performance as without the extra eye candy

So far, everything is about look what we can do to market PhysX. Look we can do this and CPU can't. Or at least that's the way it feels.

At least with Glide. Not only did you get better eye candy it was also faster. So why would anyone knock it until the competition (OGL and D3D) caught up.

This aversion to PhysX isn't anything new. It was here before Nvidia aquired them, and I'm not sure how they expected to change that perception without improving on the situation from when Ageia owned it.

Regards,
SB

 

Mirrors Edge shows a 20-30% drop in FPS when enabling PhysX on a GTX260. A fact I can see from PCGH's benchmark that a 4870 on E8500 without PhysX is faster than a GTX260 on a QX9770 with PhysX. Also shown is that using an inferior card (8400GS) for PhysX reduces your framerate by about 60%.

 

So your argument went from PhysX not impacting performance at all in Mirror's Edge to it does but depending, and then to a sort of 'but you won't notice the performance impact.' But that last argument is only valid for those with high-end GPUs, and for those with decent GPUs they're of course going to run the game with those 'factors' you mentioned (higher resolutions and detail settings, AA/AF enabled). Also worth noting is that NV did all the heavy lifting on the PhysX code for Mirror's Edge, so not exactly a strong example of developer support suggesting an independent uptake that potential informed buyers should factor into their purchase decisions.

Is Mirror's Edge a better game with the PhysX support? Most would probably say yes. Is the game indicative of impending widespread developer support? Kinda hard to say considering its proprietary nature right now, the performance impact it clearly has when running on even a high-end GPU, and the likely need for a game engine to have physics support factored into its design from the earliest stages of development to properly integrate it into gameplay in any truly significant fashion.

 

This forum is not a valid representation of the consumer market, which is where PhysX will live or die.

And given your continued backpedaling combined with the overall marketshare of the hardware that is capable of sustaining the performance you're having a hard time substantiating, I think the answer is becoming sorely obvious.

 

What I tried to be polite about is that we know this. Nobody gives a damn. Not in the marketplace as a whole, and most definitely not in a thread dedicated to the HD4770, a product that, as you have so helpfully pointed out, has no relation to PhysX.

 

Is this situation more likely to change with the release of DX11 and the tangible performance benefits that will be available to all consumers once NV catches up with hardware support? I think so.

I don't see anyone flaming PhysX. Hell, I own a GTX 285 and my arguments certainly aren't based on envy as you're suggesting. I think people are generally opposed to PhysX because it's not currently an open standard. And ask yourself how many APIs/libraries/standards proprietary to specific hardware have survived over the years in the graphics market.