New Mafia II PhysX ON/OFF video.

[SirPauly]

So does Ambient Occlusion, tessellation and all part of it and if someone desires to play with settings maxed with IQ added -- takes resources and a nice system -- this is nothing new.

The key is one can if they have the resources and the more powerful system, well, one is rewarded not just with higher frame-rate, if this is the most important thing, but abilities to take advantage of this raw performance as well to add gaming experience potential, immersion and realism.

Eventually as software and hardware evolves and mature, single GPU's should be able to handle Physics and rendering at the same time at the mid-range and lower-end sku's and gamers may be complaining about something else like raytracing taking too much of a hit. It's a vicious circle that repeats itself but as long as things are moving forward -- fine by me.

 

[Keysplayr]

I don't know if somebody already mentioned, but the video posted by the OP reminded me of the heaven benchmark: DirectX 11 Vs DirectX 10 BS.

I don't know how that is possible seeing as how this thread is about Mafia II PhysX Video comparison and the Heaven bench is all about tesselation.

 

[tviceman]

I thought EyeFinity was a step in the right direction and very welcomed. Also the pro-active nature of trying to get 10.1 and 11 content in gaming titles was as well.

Eyefinity is cool. But dx10.1 and dx11 is just doing what Microsoft says they can do.

 

[GaiaHunter]

So does Ambient Occlusion, tessellation and all part of it and if someone desires to play with settings maxed with IQ added -- takes resources and a nice system -- this is nothing new.

If you turn all those things on and it is too slow, you turn them off and really no biggie.

So as you can see, none of that is something that is selling games or hardware.

Of course at some point, all the cards will be powerful enough to do it and nobody will care much anymore.

Same with physX. Except, no matter how powerful the cards are, you will require a specific brand.

The thing is people are "selling" physX as something that will revolutionize gaming (the same way they did/do with tessellation) and it doesn't, at least for not yet.

It is something you can turn off.

I guess we could say this about most visual features but the fact is there are some visual features that take much less performance hit and don't have such pre-requisites.

 

[Piotrsama]

I don't know how that is possible seeing as how this thread is about Mafia II PhysX Video comparison and the Heaven bench is all about tesselation.

Both comparisons portrait artificial methods to "boost" what they want.
AKA lacking implementation of the lower choice.
AKA Crippling DirectX 10 / crippling No PhysX mode.

 

[zerocool84]

Personal attacks are not acceptable.

If you find reason to believe a poster has made a post that warrants moderation the expectation is that you will report the post, use the comment box to your delight, and refrain from making a public spectacle of the matter.

Moderator Idontcare

 

[the chillmaster]

AMD fans are green with envy.

hahaha I don't think this is up for debate!!

look how many pages of posts there are about the smallest little details of how nvidia is implementing their new physx feature! I've never seen such ardent fanboy-ism on both sides but holy cow I could count 50 different people in this thread who seem to be up all night worried about how each developer chooses to show off new technologies with what % of effort and dev time!!


We allow cussing in P&N and OT, not in the tech forums.

Moderator Idontcare

 

[SirPauly]

Of course at some point, all the cards will be powerful enough to do it and nobody will care much anymore.

Probably for some and a good thing because all the IHV's may offer an ability, be in titles and may be open standards, too, and simply expected and simply part of gaming. But, how did it start?

Virtually every feature started with some kind of resistance -- too much of a hit - -who cares? Only a few titles -- what's the point? Too early -- too much marketing!

 

[ShadowOfMyself]

Personal attacks are not acceptable.

If you find reason to believe a poster has made a post that warrants moderation the expectation is that you will report the post, use the comment box to your delight, and refrain from making a public spectacle of the matter.

Moderator Idontcare

 

[NoQuarter]

Both comparisons portrait artificial methods to "boost" what they want.
AKA lacking implementation of the lower choice.
AKA Crippling DirectX 10 / crippling No PhysX mode.

It's not a bad comparison. The tessellation in Heaven is actually the way tessellation should be implemented if all video cards supported tessellation though. Sometime down the line there will just be very basic models with different degrees of tessellation for the GPU to expand out to.

In an efficiency sense GPU physics is the way all physics should be implemented as well, if all GPU's supported it and people had extra GPU power to spare. However, they don't - so just like tessellated games use high detail poly meshes for DX9/DX10 hardware to look right instead of Heaven-like tessellated models, games need a decent CPU physics implementation to cover for non-PhysX hardware.


In Mafia II, GPU PhysX gives you shells that land on the ground, bouncing and scattering through overly exact algorithms and waiting to be affected by explosive forces later. CPU PhysX should give you shells that land on the ground, possibly bouncing once or twice through a rough guess algorithm and then lie dormant on the ground unaffected by any future force.

Of course that requires twice as much work because the GPU PhysX code is written entirely differently than CPU PhysX code due to the parallel nature of GPU code. You can't just say you'll use half as accurate calculations to gain some speed because the code is incompatible and has to be completely rewritten.

The problem is, the dev never had any intention of having even basic physics for shells or debris in the game, and nVidia has no interest in writing physics for shells or debris that doesn't require a GeForce GPU, so we end up with all or nothing because without nVidia's support this is what you'd consider a B-quality engine.

This isn't like Crytek or DICE work where physics effects are given serious effort, a game like that isn't one nVidia is going to spruce up because the GPU physics differences would be too minor to market effectively.


So, hooray for nVidia enhancing the game effects, and boo for Mafia 2 having poor effects to begin with.

 

[ModestGamer]

It's not a bad comparison. The tessellation in Heaven is actually the way tessellation should be implemented if all video cards supported tessellation though. Sometime down the line there will just be very basic models with different degrees of tessellation for the GPU to expand out to.

In an efficiency sense GPU physics is the way all physics should be implemented as well, if all GPU's supported it and people had extra GPU power to spare. However, they don't - so just like tessellated games use high detail poly meshes for DX9/DX10 hardware to look right instead of Heaven-like tessellated models, games need a decent CPU physics implementation to cover for non-PhysX hardware.


In Mafia II, GPU PhysX gives you shells that land on the ground, bouncing and scattering through overly exact algorithms and waiting to be affected by explosive forces later. CPU PhysX should give you shells that land on the ground, possibly bouncing once or twice through a rough guess algorithm and then lie dormant on the ground unaffected by any future force.

Of course that requires twice as much work because the GPU PhysX code is written entirely differently than CPU PhysX code due to the parallel nature of GPU code. You can't just say you'll use half as accurate calculations to gain some speed because the code is incompatible and has to be completely rewritten.

The problem is, the dev never had any intention of having even basic physics for shells or debris in the game, and nVidia has no interest in writing physics for shells or debris that doesn't require a GeForce GPU, so we end up with all or nothing because without nVidia's support this is what you'd consider a B-quality engine.

This isn't like Crytek or DICE work where physics effects are given serious effort, a game like that isn't one nVidia is going to spruce up because the GPU physics differences would be too minor to market effectively.


So, hooray for nVidia enhancing the game effects, and boo for Mafia 2 having poor effects to begin with.

I have been looking at writing a pretty big interface project for a alternative OS. All the cards on the market have the math librarys at the bios level to support physx like implementations. the issues becomes hardware power and code efficiency

 

[Dark Shroud]

Not only that I'm pretty sure ATI released tessellation not only before Nvidia but some time ago. I'm not sure if it was true or not but I've even been told tessellation was supposed to be in DX10 but Nvidia asked MS not to put it in there.

What will be really interesting is to see if/how game engines are able to use the GPU cores on Fusion & Sandy Bridge APUs in the future. Having that one core dedicated to just crunching away with Direct Compute or OpenCL paired with a discrete graphics card would be interesting. Combined with the fact that in two years time both Intel & AMD will have multi-threaded Octo-core CPUs on the market is just crazy.

These are great times for tech.

 

[Chumster]

I get the message that we all want to see it advance, but at the same time, I get the message that some want to see it die. And in representing this cause, constantly complain about it. We'll, if they want it to advance, they have a funny way of showing it.

I think you're almost correct - but wrong about what "it" is. I would love to see physics in game become more substantial to game play than what we are currently seeing - Mafia 2 included.

I also feel that as long as one camp is pushing a proprietary engine (PhysX for example), this will never come to pass.

 

[tviceman]

I also feel that as long as one camp is pushing a proprietary engine (PhysX for example), this will never come to pass.

I mostly agree with this statement, which is why AMD pisses me off for never taking the initiative to leverage their hardware in different ways than what is offered in current directx specs.

 

[MrK6]

I mostly agree with this statement, which is why AMD pisses me off for never taking the initiative to leverage their hardware in different ways than what is offered in current directx specs.

Hopefully, their recent successes allow them to hirer the personnel/staff/teams needed to improve their hardware on the software side of things, especially in features and not just performance.

 

[zerocool84]

I mostly agree with this statement, which is why AMD pisses me off for never taking the initiative to leverage their hardware in different ways than what is offered in current directx specs.

They already tried that but Nvidia pretty much forced Tessellation to be held back.

 

[Chiropteran]

Nvidia's stance on PhysX reminds me a lot of 3dfx's "glide". Potentially superior in technical aspects, but destined to fail in the end because no game developer is going to willfully limit their market share. Sad that Nvidia can't learn from the past, even when they bought 3dfx in the end.

 

[finbarqs]

I don't know if this is the thread to put this in, but I have a GTX 480, and a 8800 GTX. I used the 8800GTX for physx, and the GTX 480 for everything else. Mafia 2 actually runs SLOWER with the 8800GTX with PHYSX on high, than just a stand alone GTX 480. Is there something wrong with the PHYSX drivers? I remember something similar, where we would have to install the drivers again.

I did that, but it still runs slower than just one single GTX 480 doing everything... Anyone have any suggestions? Again, I apologize if this is in the wrong section!

 

[Grooveriding]

I don't know if this is the thread to put this in, but I have a GTX 480, and a 8800 GTX. I used the 8800GTX for physx, and the GTX 480 for everything else. Mafia 2 actually runs SLOWER with the 8800GTX with PHYSX on high, than just a stand alone GTX 480. Is there something wrong with the PHYSX drivers? I remember something similar, where we would have to install the drivers again.

I did that, but it still runs slower than just one single GTX 480 doing everything... Anyone have any suggestions? Again, I apologize if this is in the wrong section!

I have the same experience with two GTX 480s in SLI. Running one card dedicated to physx I get slower performance than with not using a dedicated card.

Physx is unreasonably demanding in this game and seems to tax hardware more than is justified.

 

[Keysplayr]

I don't know if this is the thread to put this in, but I have a GTX 480, and a 8800 GTX. I used the 8800GTX for physx, and the GTX 480 for everything else. Mafia 2 actually runs SLOWER with the 8800GTX with PHYSX on high, than just a stand alone GTX 480. Is there something wrong with the PHYSX drivers? I remember something similar, where we would have to install the drivers again.

I did that, but it still runs slower than just one single GTX 480 doing everything... Anyone have any suggestions? Again, I apologize if this is in the wrong section!

View post # 281 in this thread.

 

[tviceman]

They already tried that but Nvidia pretty much forced Tessellation to be held back.

Wow you cleared that all up. Thanks for explaining this in detail and backing up your grand conspiracy theories with amazing facts, links, and details.

 

[HurleyBird]

Wow you cleared that all up. Thanks for explaining this in detail and backing up your grand conspiracy theories with amazing facts, links, and details.

Well, he's right in a way. An *ironic* way. In the same way that AMD is 'holding back' Nvidia PhsyX by not liscensing it from them, Nvidia held back tesselation by not getting on board with ATI's propriety 'tru-form' tech way back in the day, and then again during the 2900xt-HD4890 era. Not that I blame either company for not jumping aboard on their competitors propriety tech when the options are 1. adopt it and give their competitor a huge advantage, or 2. don't adopt it because it will die without support from both companies.

Tesselation only took off after it became an industry standard. GPU physics will be the same way, and it's about time someone (*cough* microsoft) went through with it already.

 

[zerocool84]

Well, he's right in a way. An *ironic* way. In the same way that AMD is 'holding back' Nvidia PhsyX by not liscensing it from them, Nvidia held back tesselation by not getting on board with ATI's propriety 'tru-form' tech way back in the day, and then again during the 2900xt-HD4890 era. Not that I blame either company for not jumping aboard on their competitors propriety tech when the options are 1. adopt it and give their competitor a huge advantage, or 2. don't adopt it because it will die without support from both companies.

Tesselation only took off after it became an industry standard. GPU physics will be the same way, and it's about time someone (*cough* microsoft) went through with it already.

These guys forget it ALWAYS goes both ways. They blame ATI for not adopting Physx but others can blame Nvidia for not adopting Tessellation way back then.

 

[taserbro]

Well, he's right in a way. An *ironic* way. In the same way that AMD is 'holding back' Nvidia PhsyX by not liscensing it from them, Nvidia held back tesselation by not getting on board with ATI's propriety 'tru-form' tech way back in the day, and then again during the 2900xt-HD4890 era. Not that I blame either company for not jumping aboard on their competitors propriety tech when the options are 1. adopt it and give their competitor a huge advantage, or 2. don't adopt it because it will die without support from both companies.

Tesselation only took off after it became an industry standard. GPU physics will be the same way, and it's about time someone (*cough* microsoft) went through with it already.

Actually, if I remember correctly, truform also died because the developer support for it was poor and the function was a pain in the neck to implement as game designers would need to flag object surfaces on and off individually for it not to give everything the infamous melted plastic in microwave look. In addition, this being a long time before hardware companies would give a hand in writing code, the devs still had little incentive to go through all that hell since the compatible hardware adoption rate was slow and inconsistent.

I'm not gonna try and draw parallels with physics here but making it sound like it was mostly nvidia's fault that truform didn't work out is kind of a stretch. I was one of the first to buy a radeon 8500 and truform in its first iteration was a novelty at best, and worked in only in one titled that I knew of. On the other hand, all gpu functions that could stand on their own and offered immediate and overwhelming advantages like t&l, texture compression, hardware frustum culling or crossbar memory controllers were all pretty much unanimously integrated into all architectures regardless of who came up with it first.

What I do find ironic is that back in those days, people already were arguing strictly along party lines and wagered that tessellation were completely pointless, offered nothing beyond questionable cosmetic value and would probably never take off. Man... what do you say to being shown history repeating itself a generation later unable to stop it?

 

[crucibelle]

I must be blind, because I can't tell much difference between the PhysX on vs off, on the videos. I'm not a game-player, though, so that is probably why.