How will NVidia counter the release of HD5xxx?

[Scali]

Which is far more likely to happen if people start supporting PhsX. Once businesses see there is a market they will support it. I think AMD doesn't want to put resources toward GPU physics calculations b/c they sell CPUs, they see the market as saturated, they do not see people buying other hardware to get that support.

I don't think so, to be honest.
With CPU physics, AMD is going to lose anyway. Intel makes faster CPUs.
With GPU physics, AMD would at least be able to outperform Intel's CPUs. So I don't see any reason for AMD not to want GPU physics. GPU physics could make the AMD platform as a whole more competitive, even though their CPUs by themselves can't compete.

I think the real reason is that Intel owns Havok and Intel doesn't want AMD to outperform their CPUs. I have this funny feeling that Havok with OpenCL support will magically be released at the same time Larrabee hits the market.
Until then, Intel has no use for Havok with OpenCL support, so why would they release it?

AMD just happens to be between a rock and a hard place, with both its major competitors each owning a major physics API.

 

[CJ]

I don't think so, to be honest.
Until then, Intel has no use for Havok with OpenCL support, so why would they release it?

To prevent NV from gaining more ground with PhysX. LRB is still a long way off according to some information which I have here...

 

[Scali]

To prevent NV from gaining more ground with PhysX. LRB is still a long way off according to some information which I have here...

Then NV would be gaining ground with Havok instead (which would possibly also cut into Intel's high-end CPU sales, because Havok games no longer require fast CPUs)... Giving NV a headstart in being able to optimize and tweak their drivers for Havok/OpenCL doesn't sound like a good idea. I think Intel will have its hands full competing with NV and AMD in the GPU market anyway.
I think Intel is doing fine... NV isn't getting PhysX off the ground all that well, and they just let AMD fight off NV/PhysX for them, with Havok marketing and demonstrations. It's pure genius
Besides, judging from the current state of OpenCL-implementations, I don't think Havok would be competitive with PhysX at this point, on either AMD or NV hardware.

 

[Silent_Buddha]

Here's the ironic thing...

Back in the day when Nvidia and 3dfx were competing.

People were demonizing 3dfx for having a proprietary rendering API, while they embraced Nvidia for supporting open standards.

Today...

We have people embracing Nvidia for having a proprietary Physics API, and demonizing those that are advocating an open Physics API.

Regads,
SB

 

[nutball]

Today...

We have people embracing Nvidia for having a proprietary Physics API, and demonizing those that are advocating an open Physics API.

And,

We have people demonzing Nvidia for having a proprietary Physics API, and embracing those that are advocating an open Physics API.

 

[trinibwoy]

As long as nVidia wants to see money for supporting PhysX, I don't see why working with the other alternive is mind-boggling

No what's mind boggling is the dismissal of PhysX when the competition isn't doing anything better. Not sure how doing nothing in this case is the more attractive option for some folks.

C) Pray for an open standard that works on all platforms.

C) Support some other physics implementation that isn't proprietary to one vendor.

So you're both saying that someone can't appreciate PhysX and also look forward to a more open platform? Come on guys, there is absolutely no rational train of thought that can lead someone to the conclusion that PhysX is worse than nothing.

 

[trinibwoy]

People were demonizing 3dfx for having a proprietary rendering API, while they embraced Nvidia for supporting open standards.

No what's ironic is that people are making strange analogies to support a weak position. There is no open physics standard so what exactly are you referring to here? We have two competing proprietary solutions.

We have people demonzing Nvidia for having a proprietary Physics API, and embracing those that are advocating an open Physics API.

Ah, now that's funny. Notice the words you used there? "Having" as opposed to "advocating". Yeah let's just diss the real, working product while sitting on the sidelines all smug talking about the great things we plan to do.

 

[nutball]

Question: is DirectX an open standard?

 

[Scali]

Well, I think the problem is that most people who think they are arguing for open standards, don't realize that Havok is as proprietary as PhysX is.
If you want to look at truly open physics, look at something like Bullet Physics... it's open source, it's free, and it supports various GPGPU solutions (yes, Cuda too).

 

[Kaotik]

Question: is DirectX an open standard?

From IHVs and ISVs point of view, yes it is, even if it limits you to one operating system.

 

[Silent_Buddha]

There is no open physics standard so what exactly are you referring to here? We have two competing proprietary solutions.

Havok is open in much the same was as Direct3D was open back in the day. Granted, MS didn't collect licensing fees. But both are proprietary but open in the sense that the company that owns it makes sure it works on all applicable hardware something that Nvidia is patently unwilling to do with regards to PhysX. Even the CPU support there is just token support in order to say they have it.

Regards,
SB

 

[ChrisRay]

Havok is open in much the same was as Direct3D was open back in the day. Granted, MS didn't collect licensing fees. But both are proprietary but open in the sense that the company that owns it makes sure it works on all applicable hardware something that Nvidia is patently unwilling to do with regards to PhysX. Even the CPU support there is just token support in order to say they have it.

Regards,
SB

CPU support can be threaded if the dev chooses to do so. There is no inherit limitation to the PhysX library which prevents it. The PhysX library has to support threading because it has to support consoles. Nvidia's GPU PhysX driver is just that. a GPU PhysX driver sitting ontop of CUDA. When you disable that. You are not running some CPU driver.

 

[Razor1]

Havok is open in much the same was as Direct3D was open back in the day. Granted, MS didn't collect licensing fees. But both are proprietary but open in the sense that the company that owns it makes sure it works on all applicable hardware something that Nvidia is patently unwilling to do with regards to PhysX. Even the CPU support there is just token support in order to say they have it.

Regards,
SB

If AMD doesn't want to create a driver for PhysX (this is all they have to do), nV has no obligation to help out on that front.

It is an open standard, I'm not sure about licensing fees, I haven't heard anything about them, anyone has a link to that?

 

[trinibwoy]

But both are proprietary but open in the sense that the company that owns it makes sure it works on all applicable hardware something that Nvidia is patently unwilling to do with regards to PhysX.

Huh? PhysX works on more hardware than Havok does.....

And nutball, of course DirectX is proprietary. It only runs on select Windows operating systems. How can you get more proprietary than that? For some reason folks think a monopoly is exempt from the proprietary label.

 

[MfA]

Apart from some wishy washy PR I have never seen NVIDIA say they would provide a source code license to AMD to support PhysX.

 

[nutball]

Havok is open in much the same was as Direct3D was open back in the day. Granted, MS didn't collect licensing fees. But both are proprietary but open in the sense that the company that owns it makes sure it works on all applicable hardware something that Nvidia is patently unwilling to do with regards to PhysX.

Are you suggesting that if NV and ATI didn't write DX drivers for their GPUs, then MS would do so to ensure that DX "works on all applicable hardware"?

 

[Florin]

If AMD doesn't want to create a driver for PhysX (this is all they have to do), nV has no obligation to help out on that front.

Nvidia actually did help out with an independent effort sometime ago, but I'm sure someone will explain why that was just a PR stunt.

 

[Razor1]

its unrealistic to even think nV could be capable of writing the driver by themselves or with a 3rd party, AMD has to be involved, it wasn't a PR stunt, but I do remember AMD saying nV didn't go through proper channels...... really why would nV have to go through the proper channels, it should be the other way around

 

[madyasiwi]

http://www.rage3d.com/board/showpost.php?p=1336012735&postcount=2

What happened with the Havok Physics + ATI?

Even the post at [H]:

http://www.hardforum.com/showpost.php?p=1034602383&postcount=1

Didn't mention anything about Havok, but 'open source' instead.

Do they give it up?

 

[Silent_Buddha]

So we go from an open and well documented API (DirectX) to a company (MS) having to actually write hardware level drivers for a company for it to be considered open?

As far as I know, Nvidia hasn't made PhysX open or well documented with regards to other company's writing drivers in order to interface with it just as well as their own do.

Nor have I seen Nvidia approaching this in an open manner in order to embrace all hardware and thus expand it's area of influence (Direct3D for example). And considering it's a Proprietary system I wasn't expecting them to. Thus it's not exactly an open standard now is it?

Rather, Nvidia has done everything in their power to make sure it remains a closed and proprietary system (CUDA) while making vague PR blurbs to suggest their rivals could make a driver for it if they wished.

Regards,
SB