DX11 = PhysX killer?

eric_bazile

Weaksauce
Joined
Sep 22, 2009
Messages
73
Much buzz is taking place in regards to the current state of both ATI and Nvidia pertaining to ATI's current 5800 series and Nvidia's newly debuted Fermi architecture. This has brought up greatly heated debates over which side will prevail so on and so forth. I did however ponder over Nvidia's current moves along with held onto some thoughts that struck me curious. Below I shall lay out to my understanding some reasoning to the current ATI/Nvidia situation.

Nvidia trying to place a very significant role on GPGPU applications.

1- They lack the capability to produce a x86 CPU (no license)
2- They have to cover ground to make up for the possibility of lost market share due to ATI's 5800 series release.
3- Attempt a removal of media spotlight on AMD's newly introduced 5800 series
4- Keep the positive attention of investors

Minor downplay of DX11 and Nvidia's role with production of a new GPU in compliance with the architecture.

1- DX11 is specific to hardware only which is in compliance with it, it has no manufacturer BIAS.
2- DX11 tailors towards more then just straight up rendering. Adding GPGPU functionality and focus on multi-threading. This has a great ability to derail progress of PhysX based API's and the use of PhysX in later games.
3- It can help snuff proprietary standards like CUDA. As long as the card supports some of the required features that DX11 can provide you lack the worry of whether or not you have an ATI or Nvidia card to accomplish the GPGPU DX11 task.

Last but not least of all to state once again, Nvidia lacks an x86 license. AMD and Intel are planning to release a CPU/GPU single die package (for Nvidia to disregard that would be idiotic) which will play a significant role in netbooks,notebooks,ultra portables, entry level desktops etc. (That removes a massive market from Nvidia, quote me on that if desired.) Believe me, these CPU's WON'T have an Nvidia GPU attached to them (to state the obvious). I see this race towards GPGPU etc not as an act of promoting new technologies but more as one to stay profitable. They must adapt and or evolve. I see more desperation then innovation taking place.
 
Last edited:
I don't think it is a physx killer but I do think nvidia will loose its battle in physx eventually, because developers would rather use the built in features on dx11 to do physx than do it for specially nvidia's card using their phsyx tech, it just doesn't make sense to me, unless they can make more money from nvidia than selling their game for both nvidia and ati owners.
 
because developers would rather use the built in features on dx11 to do physx than do it for specially nvidia's card using their phsyx tech, it just doesn't make sense to me

I fully agree with you. Unless developers are so deep in bed with Nvidia and their "Way it's meant to be played" program I don't see PhysX being held onto for long after the release of DX11. Place on top of that the backlash towards the removal of Nvidia GPU (and most recently found out Ageia PPU) accelerated PhysX on systems with ATI GPU's and suddenly the water isn't so calm anymore.
 
Once there's middleware available for it, PhysX will most assuredly be a distant memory unless NVIDIA manages to change their approach to hardware physics.
 
It only needs to run x86 code if you want to boot an x86 OS on it.
Other code can be recompiled.
 
It all depends on how much money Nvidia is willing to spend to keep Physx going.
 
I would say that Nvidia being a nazi with their drivers is a "PhysX killer".
 
I think anytime you can move to an industry standard (DX11), the market almost always follows. Nvidia, eh. I have my feeling about them, and they are in trouble to be honest. It's no secret Intel would love to snuff them out, and I somehow doubt AMD/ATI would be all that upset either. That said we need them to stick around and at the least compete. I think they will be in for a rude awakening if they release the new Fermi based stuff at their usual prices in this economy. I think with the economic conditions it's ATI's war to lose. For a long time they been nailing the bang for the buck sales. Physx will go the way of glide, as well it should. Open API's FTW.
 
I think anytime you can move to an industry standard (DX11), the market almost always follows. Nvidia, eh. I have my feeling about them, and they are in trouble to be honest. It's no secret Intel would love to snuff them out, and I somehow doubt AMD/ATI would be all that upset either. That said we need them to stick around and at the least compete. I think they will be in for a rude awakening if they release the new Fermi based stuff at their usual prices in this economy. I think with the economic conditions it's ATI's war to lose. For a long time they been nailing the bang for the buck sales. Physx will go the way of glide, as well it should. Open API's FTW.

I couldn't agree more. IF Nvidia releases the 300 series at its usual really steep price level, it's going to flop. It needs to be competitive for Nvidia to remain a viable company.

And the things that they're doing with the drivers to disable Physx on systems where an ATi card is the main adapter is doing nothing but punishing the consumer. (Although you can get Physx to work with ATi if you use the 185.xx driver ;) ) IMO, that's just really bad politics. That puts a bitter taste in the mouths of consumers.

However, Nvidia needs to stick around so that we, the consumers have good competition and fair prices. IF either one goes under, GPU's will get very expensive very quickly.
 
Not again....the [H] is slipping.

There is NO physics API in DX11.

Repeat after me:

There is NO physics API in DX11.

DX11 means nothing to PhysX.
You need another physcis middleware to come along a beat PhysX.
Like Havok...or Bullet...but PhysX already got them beat:
http://www.hardforum.com/showthread.php?t=1451856

/thread.
 
I agree, [H] is slipping, but I think it's due to too many fanboys. nVidia ones in particular.
 
Hey, PhysX, what's up?

Remember Glide?

PhysX? PhysX?! What's wrong, PhysX!!!! You don't look so good!
 
Not again....the [H] is slipping.

There is NO physics API in DX11.

Repeat after me:

There is NO physics API in DX11.

DX11 means nothing to PhysX.
You need another physcis middleware to come along a beat PhysX.
Like Havok...or Bullet...but PhysX already got them beat:
http://www.hardforum.com/showthread.php?t=1451856

/thread.

While not a direct feature, ComputeShaders could THEORETICALLY be implemented for use in GPU-based physics.
 
I don't think it is a physx killer but I do think nvidia will loose its battle in physx eventually, because developers would rather use the built in features on dx11 to do physx than do it for specially nvidia's card using their phsyx tech, it just doesn't make sense to me, unless they can make more money from nvidia than selling their game for both nvidia and ati owners.

its exactly a physX killer, but its by no means the death of CUDA.

If your a developer starting a project right now (aim for a 2012 fall release) looking to implement a gameplay mechanic that revolves around "physics", which API are you going to use: one which has (and will continue to have) a maximum of 40% PC market penetration, or one with 70% market penetration? Granted, currently PhysX is substantially more mature than any OpenCL or DirectCompute physics API, thus if you're a studio looking to start a project now and want a serious physics implimentation, you're going to have to build/contract your own.

Nvidia seems to me to be targeting CUDA at very localized implimentations --one outside of your typcal [H]consumer market. PhysX seems to me to be more of a demonstration as to just how cool CUDA really is. Nvidia's biggest problem is that, yeah, they dont have an X86 licence, and they'll never get one, and programming for 512 cuda cores in C++ ties itself to an entirely different set of programming skills, than programming for your standard G++ compiler.
 
You forgot to mention OpenCL. What is going to make Physx obsolete is Bullet OpenCL.
 
You got it all wrong. OpenGL 3.2 is the PhysX killer. It runs everywhere, even outside Windows and on consoles. It's so going to crush PhysX.

What do you mean, it has nothing to do with physics?

DX11 has DirectCompute, which still has nothing to do with physics. At most PhysX/Havok/Bullet/whatever could be implemented on top of it, but I wouldn't know why anyone would bother because DC only works on Windows. You want OpenCL or even CUDA. DC sucks. Nobody would want to use it for anything serious. Not unless they ever hope to port it to a console, or a Mac, or Linux.
 
Not again....the [H] is slipping.

There is NO physics API in DX11.

Repeat after me:

There is NO physics API in DX11.

DX11 means nothing to PhysX.
You need another physcis middleware to come along a beat PhysX.
Like Havok...or Bullet...but PhysX already got them beat:
http://www.hardforum.com/showthread.php?t=1451856

/thread.

PhysX is a proproetary standard that is not open to the Industry.

OPENCL is a standard open to the industry.

If you honestly think for a single second a middleware wont be built upon that foundation your smoking something heavy.

When we say "OpenCL will kill PhysX" we more than understand a middleware needs to exist. We also fully understand it will.
 
PhysX is a proproetary standard that is not open to the Industry.

OPENCL is a standard open to the industry.

If you honestly think for a single second a middleware wont be built upon that foundation your smoking something heavy.

When we say "OpenCL will kill PhysX" we more than understand a middleware needs to exist. We also fully understand it will.

If PhysX is built on top of OpenCL how will it kill PhysX?
 
You got it all wrong. OpenGL 3.2 is the PhysX killer. It runs everywhere, even outside Windows and on consoles. It's so going to crush PhysX.

What do you mean, it has nothing to do with physics?

DX11 has DirectCompute, which still has nothing to do with physics. At most PhysX/Havok/Bullet/whatever could be implemented on top of it, but I wouldn't know why anyone would bother because DC only works on Windows. You want OpenCL or even CUDA. DC sucks. Nobody would want to use it for anything serious. Not unless they ever hope to port it to a console, or a Mac, or Linux.

I am amazed at the sheer number of people not getting it...it's like saying OSX will kill EAX...most people really havn't got a clue about this and they mix (and confuse) DX, Physx, OpenCL, Havok ect in one big pot...and not getting any of them :eek:

But funny to see how this "unimportant useless PhysX" really needs to be beaten...by anything people can dig up...related or not ;)
 
It will be funnier when PhysssX becomes a footnote in history of video cards.
 
PhysX is a proproetary standard that is not open to the Industry.

OPENCL is a standard open to the industry.

If you honestly think for a single second a middleware wont be built upon that foundation your smoking something heavy.

When we say "OpenCL will kill PhysX" we more than understand a middleware needs to exist. We also fully understand it will.

Thansk for proving my point that most people really don't get it.

You mean that OpenCL will end CUDA?...then say it!

OpenCL <-> CUDA <-> DirectCompute...NOTHING TO DO WITH PHYSICS!

PhysX <-> Havok <-> Bullet = PHYSICS MIDDLEWARE.

You should stop making snide ad hominem when you don't get the topic...
 
Like Brook+ ? ;)

I think it's interesting how AMD appears to be moving away from Havok with their recent support for Bullet. Apparently Intel wasn't too charmed of OpenCL Havok after all, so AMD went with the next best thing: Bullet on OpenCL :)

Methinks Havok may very well join PhysX in that future footnote ;)
 
Not unless they ever hope to port it to a console, or a Mac, or Linux.

seriously? you actually believe this? not being able to port an application to a mac or linux is going to cause physx to fail? talk about being out of touch with reality... yeah directx is going to tank anyday now... creative's going to give up on eax cause it was such a huge failure...

delusions aside, opencl, directcompute, and opengl3.2 (lol) are all just APIs. there still has to be middleware made in order to do physics calculations. there's no reason nvidia can't port physx to whatever API "wins." in fact, they're probably already working on this...
 
Thansk for proving my point that most people really don't get it.

You mean that OpenCL will end CUDA?...then say it!

OpenCL <-> CUDA <-> DirectCompute...NOTHING TO DO WITH PHYSICS!

PhysX <-> Havok <-> Bullet = PHYSICS MIDDLEWARE.

You should stop making snide ad hominem when you don't get the topic...

Isn't Physx just a CUDA application, at the moment?
It's not too hard to lump GPU accelerated physics together mentally, at this point.
 
I think it's interesting how AMD appears to be moving away from Havok with their recent support for Bullet. Apparently Intel wasn't too charmed of OpenCL Havok after all, so AMD went with the next best thing: Bullet on OpenCL :)

Methinks Havok may very well join PhysX in that future footnote ;)

I thinks it's clear that Intel is saving havok for "Larrabee"...and they are no friends of AMD, which a lot of the anti-physx crowd seem to be oblivious about.

Ah, well...they are just going to change their mantra from "Havok is going to kill Physx" to "Bullet is going to kill PhysX"...while I will enjoy PhysX in games..and not just sit and wait...and wait...and wait...and hange mantra...and wait...and wait... ;)
 
I thinks it's clear that Intel is saving havok for "Larrabee"...and they are no friends of AMD, which a lot of the anti-physx crowd seem to be oblivious about.

Ah, well...they are just going to change their mantra from "Havok is going to kill Physx" to "Bullet is going to kill PhysX"...while I will enjoy PhysX in games..and not just sit and wait...and wait...and wait...and hange mantra...and wait...and wait... ;)

I know, like we all waited for Havok games to arrive...:rolleyes:
 
The Witcher 2 is pretty hyped up and will use Havok. The first one sold 1.4 million and was a PC exclusive, and this time around will be multi-platform.

Not to mention Capcom uses Havok in its games. From an overall gaming market perspective Havok is all over the place.
 


lol, trying to sell us access to the 'top 10' list?

What's there to use it: a whole bunch of nVidia puppets?
At least Havok has got some games under it's belt.
Physx just came out with one.

Dunno. If you want to just count the 2 Halo games that use it: 200% lead.

But if you want to count: "cloth tearing for points:)" then... (I guess the Warmonger demo counts, too...)

besides... what does this even have to do with the sheer number of games sold? Compare that!
 

Yeah except if you look at the PhysX title list and compare it to upcoming games using Havok... that link doesn't make a convincing argument at all. PhysX still needs that AAA title that will impress gamers and make them run out to get an nvidia card, because at the moment it has none. Red Faction Guerilla could have been that game.

And with upcoming heavyweights such as Bioshock 2, Diablo 3/Starcraft 2, Crysis 2, Witcher 2 all using non-PhysX engines, you have to at least understand why people are skeptical about the whole issue.
 
Yeah except if you look at the PhysX title list and compare it to upcoming games using Havok... that link doesn't make a convincing argument at all. PhysX still needs that AAA title that will impress gamers and make them run out to get an nvidia card, because at the moment it has none. Red Faction Guerilla could have been that game.

And with upcoming heavyweights such as Bioshock 2, Diablo 3/Starcraft 2, Crysis 2, Witcher 2 all using non-PhysX engines, you have to at least understand why people are skeptical about the whole issue.

Ach... I can count Batman AA in, but nothing else as a 'AAA' title. It's not gonna be selling on the same level as Halo 3 or CoD4, but it will be the best Physx has to offer, for now.

But you are right. Havok and others (even Battlefront 2!) have already established Physx is not needed.
Any simple view of the before and after shots of 'Darkest of Days' makes this pretty obvious.
 
Back on topic:
DX11 is not designed to be a Physx killer, just as Physx is not designed to be a DirectX competitor. While Physx is basically a giant CUDA app, and DirectCompute seeks to displace that, it's just not the same 'apples to apples' comparison, yet.
 
Yeah, because the none-physx world in DoF dosn't seem dead, static and dull:
http://www.youtube.com/watch?v=00TWJOiTOhA

Alone having to change fireposition because the gunsmoke blocks your view after a few rounds of the old front-loading guns made PhysX worth it for me i DoF.

And you know a troll is present when the "CPU is just as good as the GPU for physics" FUD comes along :)
Or when real world numbers are ignored ;)
 
Back
Top