It's the same thing. In the control applet the user specifies whether to use a GPU/PPU for PhysX, or whether to use only the CPU. To the developer it's completely the same thing, the only limit to keep in mind being that a decent GPU can take on literally about 100x as many physics objects while even a mighty i7 CPU would have gone up in a puff of smoke by then (or just result in a very pretty slideshow like when running Crysis on an Intel GMA X4500).
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Worth it to buy a cheap, used Aeiga Physx card?
- Thread starter WhiteZero
- Start date
It's the same thing. In the control applet the user specifies whether to use a GPU/PPU for PhysX, or whether to use only the CPU. To the developer it's completely the same thing, the only limit to keep in mind being that a decent GPU can take on literally about 100x as many physics objects while even a mighty i7 CPU would have gone up in a puff of smoke by then (or just result in a very pretty slideshow like when running Crysis on an Intel GMA X4500).
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
They are not only giving GPU accelerated physics, but also a scalable version of it depending on your machines power:
http://www.pcper.com/comments.php?nid=6954
Wholeheartedly isn't the word I would use for it. As I mentioned, Nvidia has done some heavy marketing. EA had physx games under Aegia as well for some titles. They also advertised THQ support for physx the same way. Truth is:
http://www.reuters.com/article/pressRelease/idUS124573+24-Mar-2009+BW20090324
PhysX isn't free. Only download. The moment you start selling, Nvidia wants their money. Just as Havok (try and use it for free, but if you sell for over 10$ a game... license...):
http://www.havok.com/content/view/622//
Havok's opencl is not for AMD alone. Havok is middleware, not ATIware or Intelware. Physx on the other hand is Nvidiaware since only their cards have full support for it. Nvidia's cards will get support for Havok as well through opencl.
Havok is already supported on all hardware and their GPU acceleration support is going to be supported on all hardware as well. There are projects going on to implement opencl on PS3 as example, which will give GPU accelerated Havok support there also.
As mentioned, Nvidia is going in the other direction limiting the userbase for GPU accelerated Physx now that they are blocking any support for their own GPU's to be used as a secondary PPU card if combined with competitor cards.
As John Carmack said "I hope that Nvidia didnt pay a whole lot of money for PhysX".
Nvidia is probably trying to milk what they can out of Physx selling cards now, before Havok's opencl physics kicks in and kills the competition. Why else would they limit their userbase for accelerated Physx? Middleware producers usually have a business strategy to expand their userbase and thus sell more, not limit it. How can Physx compete against Havok when Havok is making solutions for physics to run well on all hardware (scalable for both CPU and GPU/vendor agnostic) and PhysX being advertised as Nvidia PhysX (vendor ware)?
I'm not sure where you are getting that from, but I can tell you that the PhysX license we agreed to doesn't include any clause about paying nVidia a cent. NVidia sees PhysX as a way to promote nVidia cards, they don't care about making money from the actual API.
Intel owns Havok. AMD is a licensee. NVidia owns PhysX. AMD refused to license PhysX. Havok FX is still dead, even long after OpenCL gets out of Beta.
There's still no evidence of Havok FX going to be revived. All we have seen was a developer prototype.
I'll agree that was a stupid and hard to understand action.
John Carmack hasn't been relevant since the 90s.
Havok FX is still dead with the corpse doing not even so much as twitching. Intel doesn't want it revived, and AMD has zero say over what happens with Havok, no matter how many pretty demos they show.
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
My mistake. I remembered reading about licencing of PhysX, but I checked it up and it was for source sdk only. Looks like Nvidia is giving it away for free to sell cards as Aegia did before. Confirmes that PhysX for Nvidia is more to push Nvidia GPU's then as middleware.
I don't know why you are pushing Havok FX and also that AMD licence it? Havok's opencl will be enabled to any card that supports Opencl by default. No need for licence. Thats the beauty of it.
You're the only one talking about havok fx.
I'm talking about havok and their opencl support. As you saw in the link, there is evidence (said out clear actually) that Havok (and AMD) is working on the opencl version of Havok. Its not so stupid and hard to understand as you showed me above. Nvidia is going the same way as Aegia did earlier (2006 I think) by giving Physx away for free to push some hardware. Havok on the other hand, is earning its money as middleware without pushing hardware (its hardware agnostic). Since an old 8800GT is sufficient as dedicated PPU, they won't sell that much new hardware if people can choose competitor cards as main renderer. AMD and Nvidia cares more about what they can sell, then what they have already sold. I have no illusions about those companies being there for any "idealistical" purposes then the money.
Nvidia played hardball, trying to push ATI to adopt PhysX by licencing CUDA at the same time, giving up ATI stream. Intel offered Havok through opencl without CUDA. Nvidia lost. Now ATI (and Nvidia) will get hardware accelerated physics through Opencl instead. Developers will get 100% userbase for it and consumers win.
Nvidia will then have a PhysX which is only hardware accelerated on their own cards. What do you think will happen with PhysX then?
Gorankar
[H]F Junkie
- Joined
- Jul 19, 2000
- Messages
- 11,107
It is easy to see why Nv has prevented the use of Nv cards for PhysX when used with a competitors card. They want people to buy NV cards. They do not want people to buy a 4970 and use an old 8xxx or 9xxx Nv card for PhysX. What that means for PhysX is obvious. While PhysX will live on in CPU mode, a hardware accelerated requirement for it will never be universally adopted. And the places where it is adopted, it will be token visual enhancements only. Developers typically code to the lowest denominator, and toss in a few perks for the high end enthusiast crowd. The days when there were games written exclusively for one gpu died with 3dfx. Judging by their actions, Nv seems determined to kill hardware accelerated physics.
AMD was correct not to jump on the PhysX bandwagon. Nv would have insured that Ati was always backseat with changes to the code and drivers that insured it ran best on Nv hardware. Just like they have with eliminating using a Nv card for PhysX along side an Ati card in their drivers. Had they open sourced PhysX, instead of retaining control, Ati might have jumped on board and hardware accelerated PhysX would prolly already be the defacto standard for upcoming games. Hardware accelerated PhysX would then become the lowest denominator. Devs would be able to cut loose and actually do something groundbreaking with PhysX.
AMD was correct not to jump on the PhysX bandwagon. Nv would have insured that Ati was always backseat with changes to the code and drivers that insured it ran best on Nv hardware. Just like they have with eliminating using a Nv card for PhysX along side an Ati card in their drivers. Had they open sourced PhysX, instead of retaining control, Ati might have jumped on board and hardware accelerated PhysX would prolly already be the defacto standard for upcoming games. Hardware accelerated PhysX would then become the lowest denominator. Devs would be able to cut loose and actually do something groundbreaking with PhysX.
Intel owns Havok. Intel hasn't said anything about working on or being in favour of a GPU-accelerated version of Havok (originally called Havok FX before Intel killed the project after buying the Havok company). That alone should say enough. Even if it were true, I doubt nVidia would just allow Havok to run on its GPUs, bringing back the whole graphics API wars of the 90s, only this time about physics APIs.
It's really a messy situation, which could have been resolved by AMD opting to allow for PhysX acceleration on their GPUs as well. I really fail to see how AMD would not benefit from licensing PhysX, other than having to pay some money to nVidia instead of Intel each year.
It's really a messy situation, which could have been resolved by AMD opting to allow for PhysX acceleration on their GPUs as well. I really fail to see how AMD would not benefit from licensing PhysX, other than having to pay some money to nVidia instead of Intel each year.
Gorankar
[H]F Junkie
- Joined
- Jul 19, 2000
- Messages
- 11,107
Nv was offering it for free to Ati, the problem was as I already stated. AMD jumping on would have made PhysX the standard for hardware accelerated physics, with Nv retaining full control. Nv, like any other company in control of a standard is going to insure that they are the king. AMD figured they would always be out performed by Nv, because they assumed that Nv would insure that PhysX ran best on Nv hardware. Nv would also know what is coming next for PhysX and would have a large lead on tailoring their hardware to perform PhysX better. AMD was not thinking they were going to get a fair shake and walked, I can't blame them.
You are right, it is a mess. One that has set back hardware accelerated physics until an open standard comes about and gets accepted by developers or MS decides to add physics to DX12 or 13 or whatever.
Intel wants physics done on the cpu, Nv wants it done on the gpu, and AMD wants it done on the gpu as well but can't figure out a way to compete while Nv holds all the cards and so is waiting for an open standard.
You are right, it is a mess. One that has set back hardware accelerated physics until an open standard comes about and gets accepted by developers or MS decides to add physics to DX12 or 13 or whatever.
Intel wants physics done on the cpu, Nv wants it done on the gpu, and AMD wants it done on the gpu as well but can't figure out a way to compete while Nv holds all the cards and so is waiting for an open standard.
Last edited:
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
Intel haven't said anything for or against GPU-accelerated Havok. What they have said is as following:
http://software.intel.com/en-us/art...re-havok-talks-simulating-real-world-physics/
Making Havok a more clear choice for ATI. Especially since PhysX was offered to ATI on the condition that they adopted CUDA as well. If Nvidia would have ported PhysX to OpenCL, they wouldn't need to offer it to ATI, since it would be enabled by default as with Havok.
In addition:
http://www.pcper.com/comments.php?nid=6954
Meaning Intel supports Opencl.
Havok on the other hand have already stated (GDC 2009) that they work with ATI towards a GPU-accelerated Havok (opencl) and they have proven with demos a port to Opencl as well.
Nvidia would struggle a bit if they block the worlds biggest physics middleware producer Havok. Then they would go from shooting themselves in the foot to shooting themselves in the head. Instead, Nvidia will be forced to offer PhysX on OpenCL or let it die. Considering that they don't make money on PhysX (giving it for free), they need to change the business model as well or throw money out the window.
Last edited:
Can we all just agree on that there's a lot to speculate on, but nothing concrete to suggest a change from the current situation?
Until we something that is available to consumers/developers which is not in beta or a prototype I'll believe it's real. I mean, heck, look at OpenCL, was supposed to be available from AMD and/or nVidia by the second half of 2009, and at this point it looks like it'll take until next year before we'll see a working solution. I am very skeptical of pie-in-the-sky press releases and demos/announcements, especially after a delay or two.
Until we something that is available to consumers/developers which is not in beta or a prototype I'll believe it's real. I mean, heck, look at OpenCL, was supposed to be available from AMD and/or nVidia by the second half of 2009, and at this point it looks like it'll take until next year before we'll see a working solution. I am very skeptical of pie-in-the-sky press releases and demos/announcements, especially after a delay or two.
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
Can we all just agree on that there's a lot to speculate on, but nothing concrete to suggest a change from the current situation?
Opencl for Havok is advertised and has been for a year already. Of course, they started advertising it before opencl was ready. Now its been demonstrated already on ATI GPU's. I wouldn't call it a speculation THAT Opencl based Havok will come.
Speculations is when it will come and how it will impact the physics market. That we agree upon. You, who is fond of GPU accelerated physics, would you see any reason why Havok shouldn't go in that direction (which they advertised and demonstrated)?
OpenCL was supposed to be ready by the first half of 2009 if I remember correctly. Nvidia have already a "pre-release OpenCL conformance candidate" ready for download (in May this year, which I would consider a "working solution") and ATI have only their CPU conformance version for download and their GPU version delivered to selected developers.
OpenCL Havok will probably not be released before ATI have their final version (khronos approved) ready.
They're more than welcome to. I just wouldn't want to use Havok within my company due to its pricing strategy and less refined API compared to PhysX.
More than Havok entering the fray I would prefer to see AMD embracing PhysX and ending this war of physics APIs before it really starts, as nobody will benefit from it.
In most newer things that support PhysX, I would think that one of the old AGEIA cards would hurt your FPS.
I currently have a GTS250 that I picked up in Ebay for $78.80 including shiiping for use as a PhysX card along side my ATI HD4850.
Haven't had the chance to play any games that support PhysX yet, but the stuff I downloaded from Nvidia's site runs really really well with this setup.
With the screen saver, with the most "taxing" settings at 1680x1050 it is usually at 60fps but drops to around 52-55fps in some places.
With the weird "alien" demo "game" thing it ran perfectly smooth at highest settings.
I currently have a GTS250 that I picked up in Ebay for $78.80 including shiiping for use as a PhysX card along side my ATI HD4850.
Haven't had the chance to play any games that support PhysX yet, but the stuff I downloaded from Nvidia's site runs really really well with this setup.
With the screen saver, with the most "taxing" settings at 1680x1050 it is usually at 60fps but drops to around 52-55fps in some places.
With the weird "alien" demo "game" thing it ran perfectly smooth at highest settings.
Gorankar
[H]F Junkie
- Joined
- Jul 19, 2000
- Messages
- 11,107
Stick with the old drivers, since in the new drivers Nv disabled PhysX support on systems where a non Nv gpu was being used for rendering
I am well aware of this.... And the newer drivers even disable Hardware PhysX if an ATI card is even present in the system.
Working on figuring out how to get past this little "inconvienience" though....
If the NV drivers don't know the ATI card is there.. they won't disable PhysX now will they?
Gorankar
[H]F Junkie
- Joined
- Jul 19, 2000
- Messages
- 11,107
I am well aware of this.... And the newer drivers even disable Hardware PhysX if an ATI card is even present in the system.
Working on figuring out how to get past this little "inconvienience" though....
If the NV drivers don't know the ATI card is there.. they won't disable PhysX now will they?
Oh, I am sure some one is already hacking them. Still, I wonder how far Nv will go with trying to lock them down. We will see I guess.
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
We've seen how Nvidia plays with PhysX and its an ugly picture (I'm refering to the disabling of GPU physics on systems that uses an Nvidia card for physx and a competitor for the rendering of the games). Nvidia is dictating to reviewers how they should review their cards and what games to use (which Nvidia cards are optimized). [H] and Kyle refused to be pushed around, though some review sites bowed and kissed Nvidia's feet. I have zero faith in Nvidia as neutral middleware and as a consumer, manipulation like this does me no favors. Who's to say that Nvidia is not going to reduce the performance of PhysX on competitor cards as well?
Your company might not like the price for Havok, but unless you are going to charge more then 10$ for your games, its free. If you are going to charge more, I, as a consumer don't like your price. Game developers bakes the costs of physics into the price of the game. Whatever you save by not using Havok, I would prefer going into Havoks pockets as hardware agnostic middleware then in yours where I am paying in your pockets for Nvidia commercials.
As for Havok's API, its won awards as numbero uno as middleware.
I doubt that ATI will ever support Physx by their own choice. Nobody held Nvidia back from developing PhysX on ATI cards, since stream SDK is available for everyone including Nvidia. ATI is better served with Havok, especially since Intel has a "hands off management" policy regarding Havok, making them a somewhat independent middleware developer.
As for you doubting Opencl, here's something from Siggraph 2009:
V-Ray RT... Target to release using OpenCl
http://forum.beyond3d.com/showthread.php?t=54987
They are not the only ones developing with OpenCL as their API of choice.
There are two main physics APIs, one is free, the other has lots of fees and strings attached. We picked the first one, allowing us to keep the prices of our games as low as possible. We also use OpenGL and OpenAL, neither of which require any kind of licensing fees. We intend to sell our first games for around $5-10. You're complaining about what exactly again?
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
PhysX has strings attached. If you are selling your games for around $5-10 Havok is free as well (without any strings attached).
unrealsword
Gawd
- Joined
- May 8, 2008
- Messages
- 825
I'd support physx if nvidia weren't such dicks and lock out all ati users with latest drivers and phys support.
You still need to buy the Havok license. And $5-10 games are peanuts. Can't get the development costs of any serious game out of it unless you sell a few million at the very least.
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
No, you don't have to buy the Havok license. Its free for any freeware or games that costs less then 10$. Even if you plan to sell it for much more, you don't have to buy the Havok license (then you have to buy it WHEN you are releasing it and if it never gets released, you don't pay anything).
Yet the fact remains that for serious, $10+ games PhysX is much cheaper than Havok. Tell me, what strings are there attached to using PhysX? It's free, it runs on any system (CPU) just like Havok... oh, and PhysX's API is less crufty than Havok's. It's win-win all around.
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
With Havoks new homepage, it looks like a standard download is free (not totally, since its paid with advertisements) even above $10, but you then need to request a no-charge PC Game distribution license from Havok:
http://www.havok.com/index.php?page=pro
Here's some of the strings attached with PhysX (I hope that you have notified Nvidia already about your games in developement, otherwise, you'll discover its not free
):Not to mention that you must not forget to advertise everywhere (unless you wish to pay):
So, write PhysX and Nvidia everywhere. For gods sake, don't forget to talk about Nvidia if you talk about your game....
Qoutes are from here:
http://developer.download.nvidia.com/PhysX/EULA/NVIDIA PhysX SDK EULA.pdf
PhysX free? The EULA says its a big ball and chain where you need to remember who you mention everywhere if you don't wish to pay a 100 000$ bill per game or if you forget to tell Nvidia about every game you develope with PhysX, or if you forget to send them a couple of copies...
Last edited:
Ah yes, EULAs. Their validity, especially in the EU (where my company is based) is questionable at best. If nVidia wants to, they can take us to court for not paying non-existing license fees for a free SDK. We just use their SDK, which is specified as being free for commercial use, show the PhysX logo in the credits and that's it. Not even other big, commercial games show the PhysX logo and such as the EULA specifies in all places it mentions.
Did I mention yet that EULAs are of questionable value? They're not contracts, there's no licensing agreement being signed, just an EULA which may or may not have been read by the developer and which can be changed at any time without notification to the developer using the SDK.
When using Havok, however, it is clearly spelled out that you're signing a licensing agreement with the company owning it. Their SDK isn't free for use in commercial settings, and there is no reasonable expectation of it being free as there is with the PhysX SDK. Very different situation.
Did I mention yet that EULAs are of questionable value? They're not contracts, there's no licensing agreement being signed, just an EULA which may or may not have been read by the developer and which can be changed at any time without notification to the developer using the SDK.
When using Havok, however, it is clearly spelled out that you're signing a licensing agreement with the company owning it. Their SDK isn't free for use in commercial settings, and there is no reasonable expectation of it being free as there is with the PhysX SDK. Very different situation.
Tamlin_WSGF
2[H]4U
- Joined
- Aug 1, 2006
- Messages
- 3,114
If you don't respect the EULA (END USER LICENSE AGREEMENT) and follow the terms, you might just as well use a pirated version of any physics engine. Even in EU anything else would be piracy.
As you see, it is a license.
In Havok, you have also a licence, but if its under $10 they say:
As you see, it is a license.
In Havok, you have also a licence, but if its under $10 they say:
http://www.havok.com/index.php?page=pro
http://www.havok.com/index.php?page=pro
Well, the thing is that I didn't see the EULA until today. You don't need to click through it to download the PhysX SDK. There is no indication elsewhere that you have to do anything but download and use it. The installer doesn't (AFAIK) have the EULA. I consider an EULA to be in addition to, not as the sole place to find out requirements and such. If nVidia can't be bothered to put it on the developer's site, then it's their problem, as we did look through all non-legalese pages. The Havok site in comparison does list it in an accessible place.
I mean, did you ever read your OS's EULA? Photoshop's? Some random application or library? If at no point it's even made a requirement to even look at the EULA, even if it's just in the installer, then I feel it's more of a money grab than a sincere attempt at an agreement between the producer and customer.
I mean, did you ever read your OS's EULA? Photoshop's? Some random application or library? If at no point it's even made a requirement to even look at the EULA, even if it's just in the installer, then I feel it's more of a money grab than a sincere attempt at an agreement between the producer and customer.