Widescreen Gaming Forum

>>Lots of debris = less fps than on vanilla maps... but "slomo" with a card like yours ?? Did you try the latest drivers ?

Yeah, I have the latest 177.whatever drivers direct from the NVIDIA CUDA mini page. Perhaps it's a matter of running the graphics and physics calculations all on one GPU core? I play everything at 1680x1050 resolution, so that might be a little taxing when I start throwing exploding brick walls into the mix. Not that it terribly matters because I don't really like the three maps are designed anyway, haha. The new Facing Worlds map, however, is quite a delight.

>>And that's where people get confused about PhysX : you can't always accelerate it using a GeForce or Ageia card ! It's basically only a part of a game's engine dedicated to physics (gravity and such). Lots of games use it - but very few feature hardware PhysX support or advanced Physx effects... GoW, like all UE3-based games, uses PhysX - but unlike UT3, it doesn't support hardware acceleration of PhysX afaik.

Right. I hadn't thought of that, even though I knew there was software and hardware acceleration. Thanks for the reminder!

I'll try out Switchball too, I'm a big fan of those fun casual games. Off-topic, but ever try out Grid Wars (1 or 2, but I've only tried 2)? It's the freeware ripoff of Geometry Wars...

Don't get your hopes up too high, google force within, you should be able to find your way from there, the Nvidia download manager is leet.

Hm. When did you get the power pack Edward? I keep getting a message stating "The server returned an error. Would you like to retry?" when the download initiates. I'll try later I suppose.

[edit]Nevermind, I saw the link for downloading the files individually.

My opinion is that they've turned their video cards into sales gimmicks. I don't believe for one second that nVidia structured their architecture to fully utilize the current 3d api's while, at the same time, allowed for concurrent processing for the Physx api.

I would place my money on a conversation like this:

nVidia exec: "Hey, Ageia looks like a buyable company, any way we can leverage that?"
Corporate yes-men: "Absolutely, we have processors on our cards, we can use those to run the api. Oooooo! Oooooo! We can even give it an acronym!"
ne: "That's a good idea, I'm glad I thought of it. Call it CUDA, you know, like barracuda, 'cause I like that. Make it an acronym or something. Oh, and buy Ageia."
cy: "You got it, Boss!"

cy: "Here, make the 8x, 9x and new 200 line of cards do this." Hands an engineer a 1 page document explaining more about the acronym than any sort of requirements or even business spec.
nVidia engineer: "Um, this really isn't possible, you see..." He goes on for about 10 minutes explaining all the very good reasons why this is a bad, bad idea. The cy hears "blah, blah, tech-stuff, blah, blah, hope-it's-sunny-for-golf, blah".
cy: "Whatever. Make it happen or your fired."
ne: "Sigh. Ok."

The cards were not made for this. The simple proof is that they almost all have single core processors. Ageia cards have their own processors for a reason, and 1 processor to do the work of 2 will cut performance by about 66% (not half. There's quite a bit of overhead involved in concurrent processing).

Sorry for the extra bile and ire in this post. I have a huge problem with companies that lie to their customers to get them to buy their products, and I see this as exactly that. I also have a huge problem with nVidia now. If you would like to know more about that, I made a blog post here a little while ago that you can read.

I know this is the way most corporations work. I still don't like it.

[quote]Perhaps it's a matter of running the graphics and physics calculations all on one GPU core?.

YA THINK? No seriously, that's exactly what came to mind when I read your comments. With games as resource hungry as they are, many now having lots of destructible objects, I feel you either need a dedicated processor on the video card to run PhysX or run it via the CPU, like physics have been traditionally, esp if it's a multi core CPU.

Back when there were rumors (false of course) of the 8800 having a dedicated physics processor, I assumed from what I'd read about Nvidia's incorporation of virtual physics processing on those GPUs via software laying that their next step would be a dedicated physics processor on board. It appears the route they've gone though will result in your top shelf card suddenly performing like a mid-range one should you want to employ PhysX.

Nice broken quote :P

Sigh. Here we go again.
You can't just have a "PhysX processor" on a graphics card like that - the good old Ageia P100 card has 36 cores... That's why shader programming and CUDA are such a clever way to deal with the problem. That's also why it doesn't make sense to think "a multicore CPU" can handle advanced PhysX effects along with all the rest. It's not only computational power we need, it's friggin' parallelism. We'll see what 18-core Nehalems and whatnot can do for PhysX, the engine itself probably needs optimizing too, but from here on we're entering Speculationland...

FYI using your main graphics card OR a PPU to do the PhysX will give similar framerates - it even seems PhysX on GeForce definitely beats the PPU with newer drivers !
http://www.firingsquad.com/hardware/physx_performance_update/

The problem is advanced PhysX always taxes framerates to begin with - but don't forget the UT3 maps like all "advanced PhysX" features are still tech demos, and as such they're pretty unrefined and resource-hungry. PhysX can probably get better than that... the problem is, when will it actually start to happen ? :?

PhysX can probably get better than that... the problem is, when will it actually start to happen ?

When there is something unified (probably in DirectX) that every system can call to, whether it has an Intel CPU, and AMD one, ATi VPU or nVidia GPU. ;)

stuff

I stand corrected in my knowledge of the technology.

My original point is still valid, however. The cards were not made for this and I still think it's a gimmick.


Or maybe I'm just bitter. :?

OR my english vocabulary is somewhat limited :oops:

Some nvidia official stated in an interview that PhysX on GPU is a bit like AA : "using it may have a performance cost, but it's still a nice feature, right ?" Given the few titles we can test at the moment, all in all, I tend to agree. Plus I have a killer gaming rig and have very playable framerates no matter what on all UT3 PhysX maps. (So yes, be bitter!) :lol:

The problem with PhysX is you can only hate it as long as you didn't actually see it in action - I pity the ones who never saw an Ageia card at work managing the insane amount of debris in the old, broken, unoptimized and then again kickass and promising Cellfactor games !

I know I sound like a fanboy, and again, blame my english. I just wish to explain what I know about this ever-underrated, now-controversial technology. As for picking a graphics card because it supports PhysX or DX10.1 or not, it's simply foolish at the moment.

YA THINK? No seriously, that's exactly what came to mind when I read your comments. With games as resource hungry as they are, many now having lots of destructible objects, I feel you either need a dedicated processor on the video card to run PhysX or run it via the CPU, like physics have been traditionally, esp if it's a multi core CPU.

Back when there were rumors (false of course) of the 8800 having a dedicated physics processor, I assumed from what I'd read about Nvidia's incorporation of virtual physics processing on those GPUs via software laying that their next step would be a dedicated physics processor on board. It appears the route they've gone though will result in your top shelf card suddenly performing like a mid-range one should you want to employ PhysX.

That's what I initially thought Frag Maniac, but then I remembered that that's really only true of last-generation technology (althought I wouldn't mind having a second GTX in my case). I think it's important to remember that the 8 + 9 + GT200 cards all have significant multithread processing capabilities, and with 240 stream processors it's inexcusable that I can't run Heatray_PhysX beyond 30 FPS on the most scalable game engine next to Source. It's all very interesting really, if not confusing. I've seen all the Tom's Hardware and [H]ardOCP and whatever-other-site-out-there benchmarks showing GeForce-based PhysX beating out the original PCI-based PPU, so it seems a little crazy that Epic (and other PhysX partners) and nVIDIA wouldn't work a *little* closer on integration into the game, as opposed to blaming the technology, especially when you see gameplay videos for, say, Cellfactor. Understandably, CUDA's real usefulness lies in computing - like with the simulations for protein structure synthesis at Stanford or whichever university.

That's just my theory: that it's a combination of a need to properly implement PhysX and immaturity of the technology, rather than a simple matter of not having enough video cards stuffed into your rig.

OR my english vocabulary is somewhat limited :oops:

Some nvidia official stated in an interview that PhysX on GPU is a bit like AA : "using it may have a performance cost, but it's still a nice feature, right ?" Given the few titles we can test at the moment, all in all, I tend to agree. Plus I have a killer gaming rig and have very playable framerates no matter what on all UT3 PhysX maps. (So yes, be bitter!) :lol:

More titles will come. :) Age of Conan is getting PhysX as well:
http://www.bit-tech.net/news/2008/08/28/age-of-conan-directx-10-content-looks-stunning/1

All new features that can gives us more joy when gaming is welcome. :D

It will be interesting to see what happens with the physics in the coming year.

Intel have its Havoc, which support most games at the moment. Their new processor series might gives us great physics without sacrificing FPS, AA and AF.

DirectX 11 brings physics to every user.

Nvidia's PhysX can be used on more then Nvidia cards:
http://www.tgdaily.com/content/view/38137/135/

It all depends what directions they are chosing to take.