DX11 = PhysX killer?

[jeremyshaw]

lol, nah, i know you mean well. everyone is entitled to their own thoughts and opinions. so yep, we'll just have to see what the future holds on this front.



lol, indeed. seeing as how dx11 offers a physics engine, the thread was aptly named.

lol, now if only VIA would release a killer 12 core CPU...

 

[kumquatsrus]

lol, now if only VIA would release a killer 12 core CPU...

heck, make it interesting and have it priced to sell with solid performance and released to market alot earlier than the others, and i would pick one up, lol. go via!

 

[jeremyshaw]

*LOL*

The reason the havok CPU water "link" dosn't work...is because there is NO havok CPU water *ROFL*

I am laughing to hard to type anymore...BBL *ROFL*

Just like Atech to make up stuff that doesn't exist...

 

[kumquatsrus]

Just like Atech to make up stuff that doesn't exist...

haha, someone should get him back in the future for that one.

 

[kadozer]

*LOL*

The reason the havok CPU water "link" dosn't work...is because there is NO havok CPU water *ROFL*

I am laughing to hard to type anymore...BBL *ROFL*

*Scratches head..LOL! bro I haven't laughed this hard since Carrot Top! LOL! Reminds me of this one time this guy pretended to show a link to a new game and it was a rick roll, LOL!.

 

[Atech]

It just goes to show that you should always use the little greys when I post...I don't play "fair"..."nice"...but not "fair"

But I really like this thread..it shows that PhysX has moved from irrelevant, to people wanting more (including alternatives to the API)...and the GPU physics is suddenly not just a buzzword, but a real feature.

Just a shame it should take +3 years...think what would have happend if 3D acceleration first kicked off after +3 years...but alas...things are moving

 

[ToastyBread]

Just a shame it should take +3 years...think what would have happend if 3D acceleration first kicked off after +3 years...but alas...things are moving

Apparently you weren't around much when 3D acceleration first took off... a few major different competing formats, with the supporting companies playing dirtier than they do now in order to try and get support for their standard.

It wasn't until Microsoft finally truly pushed D3D that we finally started seeing some uniformity, and even then, if memory recalls, it wasn't until DX 7 was released.

Nobody is arguing that physics acceleration is a feature that shouldn't be included. What's being argued is the means by which it's implemented.

Some (such as yourself it seems) worship the ground nVidia walks on, and think PhysX should be the standard and all others dropped.

Others want a more open, cross-platform API that isn't controlled/administered primarily by one company. That's why all of the major physics implementations currently available have issues, and why I truly do believe that we won't see any type of uniformity until Microsoft puts its foot down and either supports an existing one, or comes out with their own.

Either way, everyone over-estimates the importance of physics in games. It's a nice feature, sure. But it isn't exactly the same as the jump from 2D to 3D gaming. Maybe someday, but not today.

 

[ToastyBread]

havok fx wasn't announced until 2006, while ageia announced the ppu a year ago in 2005 prior to that. so if anything, havok fx was a response by nvidia/ ati to hardware accelerated physx. so yeah, physx could still be considered the pioneer.

When you say "havok fx was a response by nvidia / agi to hardware accelerated physx", did you really mean that havok fx was havok's response? :-P Cause Havok was working independent of nVidia and ATI, as far as I can remember, on having Havok supported on nvidia/ATI hardware. I do agree though, it was in response to AGEIA's PPU.

I'd consider PhysX the pioneer of hardware-accelerated physics, but not the pioneer of physics implementation by any means.

And in regards to Batman: AA, the original argument was just that once a AAA quality game that supports PhysX was released and sold a million or more copies on the PC, that a case could be made that PhysX may be having an impact (or something like that, I can't recall the specifics of what others were arguing at thi spoint).

Someone then said it had sold 2.5 million copies, so it should count. I simply pointed out that the number of PC copies sold of B:AA was far less than that, since console sales were the predominant source. PhysX wasn't implemented in any way, shape or form for the 360 or PS3 versions, so the PC version, which did have PhysX support added, was the only version that could count in regards to that original argument. It's fairly pointless anyway, since who knows how many PC copies were given away to nVidia card purchasers as part of the voucher program. I'd give the game a try on my PC, but I don't like the idea of it being gimped on my soon-to-be-ATI-based system, so I'll try it on my PS3 or 360.

 

[kumquatsrus]

Apparently you weren't around much when 3D acceleration first took off... a few major different competing formats, with the supporting companies playing dirtier than they do now in order to try and get support for their standard.

It wasn't until Microsoft finally truly pushed D3D that we finally started seeing some uniformity, and even then, if memory recalls, it wasn't until DX 7 was released.

Nobody is arguing that physics acceleration is a feature that shouldn't be included. What's being argued is the means by which it's implemented.

Some (such as yourself it seems) worship the ground nVidia walks on, and think PhysX should be the standard and all others dropped.

Others want a more open, cross-platform API that isn't controlled/administered primarily by one company. That's why all of the major physics implementations currently available have issues, and why I truly do believe that we won't see any type of uniformity until Microsoft puts its foot down and either supports an existing one, or comes out with their own.

Either way, everyone over-estimates the importance of physics in games. It's a nice feature, sure. But it isn't exactly the same as the jump from 2D to 3D gaming. Maybe someday, but not today.

agreed. and the point is the foundation has to be put into place before it can even progress in the future up to where it will be ubiquitous and commonplace in improving gameplay experiences. but yeah, someday. hopefully, someday is sooner rather than later.

 

[kumquatsrus]

When you say "havok fx was a response by nvidia / agi to hardware accelerated physx", did you really mean that havok fx was havok's response? :-P Cause Havok was working independent of nVidia and ATI, as far as I can remember, on having Havok supported on nvidia/ATI hardware. I do agree though, it was in response to AGEIA's PPU.

I'd consider PhysX the pioneer of hardware-accelerated physics, but not the pioneer of physics implementation by any means.

And in regards to Batman: AA, the original argument was just that once a AAA quality game that supports PhysX was released and sold a million or more copies on the PC, that a case could be made that PhysX may be having an impact (or something like that, I can't recall the specifics of what others were arguing at thi spoint).

Someone then said it had sold 2.5 million copies, so it should count. I simply pointed out that the number of PC copies sold of B:AA was far less than that, since console sales were the predominant source. PhysX wasn't implemented in any way, shape or form for the 360 or PS3 versions, so the PC version, which did have PhysX support added, was the only version that could count in regards to that original argument. It's fairly pointless anyway, since who knows how many PC copies were given away to nVidia card purchasers as part of the voucher program. I'd give the game a try on my PC, but I don't like the idea of it being gimped on my soon-to-be-ATI-based system, so I'll try it on my PS3 or 360.

yeah well i think essentially it was nvidia/ ati using havok fx in order to tout gpu physics over ageia's hardware to keep it from becoming relevant, but yes, duly noted.

agreed.

problem is physx is in every version of batman, console and pc alike. the only difference is the consoles get software physx while the pc gets software physx & optional gpu physx. physx is still the underlying physics engine in ue3 which batman is using, whether it be xbox360, ps3, or pc. but yeah, the argument is pretty pointless overall.

 

[Elledan]

Apparently you weren't around much when 3D acceleration first took off... a few major different competing formats, with the supporting companies playing dirtier than they do now in order to try and get support for their standard.

It wasn't until Microsoft finally truly pushed D3D that we finally started seeing some uniformity, and even then, if memory recalls, it wasn't until DX 7 was released.

Ugh... it wasn't until DX9 that DX actually became usable. DX6 was the worst thing a developer could happen. Meanwhile OpenGL was around and very much more usable. The sole reason why DX got popular on Windows instead of OpenGL plus OpenAL and such was that MSFT is just really good at marketing. Which is why even your grandmother uses Office even though she only types the occasional letter for which Wordpad would more than suffice. Consoles never got in the habit of using DX, though, not even the XBox (360).

Others want a more open, cross-platform API that isn't controlled/administered primarily by one company. That's why all of the major physics implementations currently available have issues, and why I truly do believe that we won't see any type of uniformity until Microsoft puts its foot down and either supports an existing one, or comes out with their own.

I'm a game developer and I would not want to use an API developed by MSFT, nor would I care about what they 'support'. I want something which is completely independent, just like OpenGL (maintained, not created by Khronos), something even game developers can contribute to. Bullet is the closest thing we have to this right now, but even it is far from this goal, still.

Either way, everyone over-estimates the importance of physics in games. It's a nice feature, sure. But it isn't exactly the same as the jump from 2D to 3D gaming. Maybe someday, but not today.

Oh, how wrong you are. Do you have any idea of the jumps in gameplay possibilities with even reasonably accurate not to mention ubiquitous physics for any and all objects in a scene? The possibility of doing animation in real-time instead of using mo-cap/key-frame animation, thus saving millions in development costs and countless hours of work.

The possibilities truly are astounding.

 

[ToastyBread]

Ugh... it wasn't until DX9 that DX actually became usable. DX6 was the worst thing a developer could happen. Meanwhile OpenGL was around and very much more usable. The sole reason why DX got popular on Windows instead of OpenGL plus OpenAL and such was that MSFT is just really good at marketing. Which is why even your grandmother uses Office even though she only types the occasional letter for which Wordpad would more than suffice. Consoles never got in the habit of using DX, though, not even the XBox (360).

DX6 was horrible. DX7, not so much. DX7 started showing a good number of improvements, which ultimately led into DX8, which was *definitely* the point at which MS had secured widespread development support behind DirectX. Sure, maybe it was just marketing, but you can't argue with the results (it seems like the only ones really still gung-ho about OpenGL at that point were a few independent developers and the few Apple gaming fans that existed, lol).

I can remember the huge marketing push that occurred with GPU vendors advertising DX8 compatibility, and the huge impact that had on the market. Call it whatever you want, but MS did succeed in getting uniformity present.

And don't underestimate what Office brought. Is WordPad more than enough for what many casual users do? Sure. But look at the early-to-mid 90s, with all of the major competing word processing and office productivity suites, and the lack of compatibility between them. People scoff at Office now, but most of those people never had to use the likes of WordPerfect and Lotus Notes. I did, and those were much, much worse by comparison.

I'm a game developer and I would not want to use an API developed by MSFT, nor would I care about what they 'support'. I want something which is completely independent, just like OpenGL (maintained, not created by Khronos), something even game developers can contribute to. Bullet is the closest thing we have to this right now, but even it is far from this goal, still.

So then you should not want PhysX to be the dominant standard, as it's just as much of a "non-independent" API as DirectX.

And whether or not you dislike anything provided by MS, the fact remains that many do like DirectX, for all of its misgivings, especially since MS has been in the console space and using DX on their consoles.

Oh, how wrong you are. Do you have any idea of the jumps in gameplay possibilities with even reasonably accurate not to mention ubiquitous physics for any and all objects in a scene? The possibility of doing animation in real-time instead of using mo-cap/key-frame animation, thus saving millions in development costs and countless hours of work.

The possibilities truly are astounding.

Did you even read the full line, including the part where I said "maybe someday, but not today?". As of today, as in, right now, and for the last few years, physics implementations in games have been essentially nothing more than a gimmick.

I never said physics would always be that that way. In fact, I hope it does become the next big thing for gaming, because expanding the game play experience is generally always a positive thing.

However, you cannot sit there ,and tell me that as it is right now, that physics implementation has had anywhere near the impact on gaming as the change from 2D to 3D did. I remember when the likes of Super Mario 3D, Tomb Raider, hell, even the original Star Fox, were coming out, and the *huge* hype those were producing in both the gaming industry and the mainstream media. It was billed as this huge advancement in gaming. Has any type of physics implementation had anywhere near that impact? No. That's what I meant, and I still stand by it.

I honestly hope that all of the possibilities you mentioned, such as real-time animation, etc., it's all potentially very exciting and game-play changing in the future. But that's just it, that's the future, and not today. Thanks ;D

 

[Elledan]

So then you should not want PhysX to be the dominant standard, as it's just as much of a "non-independent" API as DirectX.

I don't care much for PhysX, true. I'm actually preparing an effort to rectify this and hope to find more developers with the same interest.

And whether or not you dislike anything provided by MS, the fact remains that many do like DirectX, for all of its misgivings, especially since MS has been in the console space and using DX on their consoles.

That's the myth, but in reality the XBox consoles do not use DX. They use a lower-level API which is relatively easy to port to and from DX, but it isn't DX.

Did you even read the full line, including the part where I said "maybe someday, but not today?". As of today, as in, right now, and for the last few years, physics implementations in games have been essentially nothing more than a gimmick.

[..]

I honestly hope that all of the possibilities you mentioned, such as real-time animation, etc., it's all potentially very exciting and game-play changing in the future. But that's just it, that's the future, and not today. Thanks ;D

It just seemed to me that you were downplaying the use of physics, something which is a bit of a pet peeve of mine Sorry if that wasn't what you intended.

Most PC gamers actually have GF8 or better nVidia cards, so one developer just has to ignore the 30% or so of the gaming market which do not have GPU physics capabilities. It could be done today.

 

[sethk]

Most big budget games today are not PC-specific, so PhysX needs to be added to the PC port - lots of examples of this happening in the games that support PhysX.

As part of TWIMTBP (worst acronym ever, BTW,) developers receive help with integrating PhysX into their PC ports, and every bit of help counts in keeping costs and development time down. This is part of the reason for PhysX success so far under Nvidia - it's a selling point and it doesn't cost that much to add with Nvidia's help.

However, I think they're being short sighted in keeping it so closed. They should just open it up and let ATI make their own implementation if they want, chances are they wouldn't right away, and if they did it would take them time to catch up in features and performance. It would take the 'closed' platform stigma away from PhysX. Their recent move to block PhysX on mixed GPU environments (Windows 7) is baffling, I'm hoping it's just a short term thing while they figure some stuff out, because it's just stupid and will turn Nvidia owners against them.

OpenCL / DirectCompute is capable of PhysX-like behavior if the programmers have time to create a physics engine, but it's the difference between saying "C++ is capable of raytracing' versus actually using a free raytracing library. You would be starting from scratch. A free high end Physics library with no strings attached would be a no-brainer, people would use it more if it was supported across the board. Even CPU PhysX right now is limited to a single thread. With 12 thread / 6 core CPUs on the near horizon, this is monumentally stupid, but Nvidia is trying to avoid articles where a 6 core Nehalem is shown to soundly thrash a low end GPU at PhysX.

It doesn't help that Nvidia is currently behind ATI in the GPU game. Maybe they'll free some of this stuff up once they have returned to a more competitive environment.

 

[kumquatsrus]

Most big budget games today are not PC-specific, so PhysX needs to be added to the PC port - lots of examples of this happening in the games that support PhysX.

As part of TWIMTBP (worst acronym ever, BTW,) developers receive help with integrating PhysX into their PC ports, and every bit of help counts in keeping costs and development time down. This is part of the reason for PhysX success so far under Nvidia - it's a selling point and it doesn't cost that much to add with Nvidia's help.

However, I think they're being short sighted in keeping it so closed. They should just open it up and let ATI make their own implementation if they want, chances are they wouldn't right away, and if they did it would take them time to catch up in features and performance. It would take the 'closed' platform stigma away from PhysX. Their recent move to block PhysX on mixed GPU environments (Windows 7) is baffling, I'm hoping it's just a short term thing while they figure some stuff out, because it's just stupid and will turn Nvidia owners against them.

OpenCL / DirectCompute is capable of PhysX-like behavior if the programmers have time to create a physics engine, but it's the difference between saying "C++ is capable of raytracing' versus actually using a free raytracing library. You would be starting from scratch. A free high end Physics library with no strings attached would be a no-brainer, people would use it more if it was supported across the board. Even CPU PhysX right now is limited to a single thread. With 12 thread / 6 core CPUs on the near horizon, this is monumentally stupid, but Nvidia is trying to avoid articles where a 6 core Nehalem is shown to soundly thrash a low end GPU at PhysX.

It doesn't help that Nvidia is currently behind ATI in the GPU game. Maybe they'll free some of this stuff up once they have returned to a more competitive environment.

physx doesn't need to be ported if physx is what is being used in the first place for a multiplatform title since physx is already supported on the consoles. gpu physx is just a bonus that nvidia is leveraging for gpus that are compatible with it.

as far as gpu physx being closed, amd doesn't want to support it since it belongs to their competitor, which is perfectly reasonable.

in regards to cpu physx being single threaded, i would say that might depend on the game, b/c even an old game like city of villains shows a large performance improvement from single to dual:

http://www.anandtech.com/video/showdoc.aspx?i=2828&p=3

havent seen an article where a 6 core nehalem is shown to thrash a low end gpu at physx, but if that was the case, it would have to be a pretty low end gpu i imagine (which is handling the rendering as well), and physx would have to be capable of scaling beyond a single core, which i guess it does according to the link. but yeah, i would think a 6 core cpu should handle it better than a low end gpu already being bogged down with graphics + physx. in that case, it would be better to run physx on the cpu, which the option already exists for both cpu & gpu implementation of physx.

again, it's just a matter of waiting for the competition to arrive for hardware accelerated physics before support can become more widespread.

 

[sethk]

physx doesn't need to be ported if physx is what is being used in the first place for a multiplatform title since physx is already supported on the consoles. gpu physx is just a bonus that nvidia is leveraging for gpus that are compatible with it.

as far as gpu physx being closed, amd doesn't want to support it since it belongs to their competitor, which is perfectly reasonable.

in regards to cpu physx being single threaded, i would say that might depend on the game, b/c even an old game like city of villains shows a large performance improvement from single to dual:

http://www.anandtech.com/video/showdoc.aspx?i=2828&p=3

havent seen an article where a 6 core nehalem is shown to thrash a low end gpu at physx, but if that was the case, it would have to be a pretty low end gpu i imagine (which is handling the rendering as well), and physx would have to be capable of scaling beyond a single core, which i guess it does according to the link. but yeah, i would think a 6 core cpu should handle it better than a low end gpu already being bogged down with graphics + physx. in that case, it would be better to run physx on the cpu, which the option already exists for both cpu & gpu implementation of physx.

again, it's just a matter of waiting for the competition to arrive for hardware accelerated physics before support can become more widespread.

While I read earlier this year than Nvidia is now licensing PhysX to PS3 developers, my understanding is that it is not available on Xbox360. I don't know of any shipping console game with PhysX, if you do, I would love a link. As of now, most PC ports get PhysX added to the PC port.

It makes sense that games would benefit going from one core to two - some of the basic system threads driver processing versus game specific code can be separated without issue - that's not proving that PhysX scales. There are articles showing that CPU PhysX do not scale currently, because the PhysX CPU driver only does PhysX calculations on a single core.

 

[Atech]

While I read earlier this year than Nvidia is now licensing PhysX to PS3 developers, my understanding is that it is not available on Xbox360. I don't know of any shipping console game with PhysX, if you do, I would love a link. As of now, most PC ports get PhysX added to the PC port.

It makes sense that games would benefit going from one core to two - some of the basic system threads driver processing versus game specific code can be separated without issue - that's not proving that PhysX scales. There are articles showing that CPU PhysX do not scale currently, because the PhysX CPU driver only does PhysX calculations on a single core.

Complete list of games, including Xbox 360 games:
http://physxinfo.com/