NVIDIA releases certified OpenCL drivers.

Atech

2[H]4U
Joined
Jul 14, 2007
Messages
3,946
http://beyond3d.com/content/news/735

NVIDIA have released certified OpenCL drivers for Windows and Linux, after passing conformance tests at Khronos in June. The Windows version is available for XP, Vista and 7, both 32- and 64-bit in all cases, and is tagged 190.89 for driver version number spotters.

The Linux one is 190.29 and is available for 32- and 64-bit just like the Windows version. Both OSes get support for the visual profiler, and any CUDA-capable GPU is an OpenCL compatible one, from G80 all the way up, and including Tesla T10-powered hardware in the compute space too.
 
*nods*

I was using the pre-release version up till recently. Good to see no major issues were found after the Beta program :)
 
*nods*

I was using the pre-release version up till recently. Good to see no major issues were found after the Beta program :)

NVIDIA has a big focus on GPGPU....unlike AMD, so it's no surpirse to me.
Brook is going EOL and I wonder when AMD will get their OpenCL drivers ready.

But good for NVIDIA.
 
NVIDIA has a big focus on GPGPU....unlike AMD, so it's no surpirse to me.
Brook is going EOL and I wonder when AMD will get their OpenCL drivers ready.

But good for NVIDIA.

Yeah, Stream/Brook+/CTM never took off and their OpenCL efforts so far seem to be half-hearted. I don't recall hearing about a dev Beta program like nVidia had for OpenCL, nor about any press releases announcing a fixed date.

I guess nVidia will be throwing a small party :) I know I am, because now I can move my company's products to OpenCL (assuming a performance benefit or equality compared to CUDA) without having to force pre-release drivers upon my customers :p
 
Yeah, Stream/Brook+/CTM never took off and their OpenCL efforts so far seem to be half-hearted. I don't recall hearing about a dev Beta program like nVidia had for OpenCL, nor about any press releases announcing a fixed date.

I guess nVidia will be throwing a small party :) I know I am, because now I can move my company's products to OpenCL (assuming a performance benefit or equality compared to CUDA) without having to force pre-release drivers upon my customers :p

I am more excited about this:
gf100.jpg


GF100 (Geforce Fermi unlike the previous Geforce Tesla)...indicating a new MIMD architechture...tailored for GPGPU ;) )

GF100 apparently is the correct name for the artist formerly known as "GT300"
 
Yeah, I had heard about the use of MIMD before. Sounds like a very interesting thing if true :D
 
Now AMD (with a nod from Intel) need to port Havok to OpenCL and give us some GPU goodness.
Something tells me this isnt going to happen though.
 
Now AMD (with a nod from Intel) need to port Havok to OpenCL and give us some GPU goodness.
Something tells me this isnt going to happen though.

They need OpenCL drivers before porting anything ;)
 
They need OpenCL drivers before porting anything ;)

Aye you already mentioned that, mebbe this will make AMD get a shift on.
Then again, NVidia having OpenCL will probably make AMD decide to never port Havok to OpenCL !
Politics grrr.
 
Aye you already mentioned that, mebbe this will make AMD get a shift on.
Then again, NVidia having OpenCL will probably make AMD decide to never port Havok to OpenCL !
Politics grrr.

AMD has no say in porting Havok to OpenCL...that is Intel's say ;)
 
AMD has no say in porting Havok to OpenCL...that is Intel's say ;)

Thus why once again I say things like PhysX and Havok should not be owned by direct competitor's. It only hinders advancement...
 
Thus why once again I say things like PhysX and Havok should not be owned by direct competitor's. It only hinders advancement...

Yeah, becuase NVIDIA isn't pushing PhysX and hardly anyone uses their
API...erhmm...wait?

There are opensopurce alternatives outhere...they just don't have the same "pondus" as either PhysX or Havok.

AMD dosn't have their own physics middleware...but that is their own fault.
 
Yeah, becuase NVIDIA isn't pushing PhysX and hardly anyone uses their
API...erhmm...wait?

There are opensopurce alternatives outhere...they just don't have the same "pondus" as either PhysX or Havok.

AMD dosn't have their own physics middleware...but that is their own fault.

Where did I even talk about PhysX or any vendor by name? Seriously dude you need to check your PhysX crusade at the door. Anytime somebody says something you assume it's bashing PhysX. I don't hate PhysX. I just hate what Nvidia is doing with it. I also hate what Intel is doing with Havok. I see them both as hindering physics development.

AMD doesn't have a say in PhysX being ported to OpenCL either. Neither does Intel. Nvidia likewise has no say in what Havok does. That's the issue. We have three vendors and two in control of two different technologies that can both lead to the same results with further development.

We need 3 vendors and NONE of them in direct control of a technology that is hardware independent and can be used on CPU or GPU.

Anything else is hindering development.
 
Where did I even talk about PhysX or any vendor by name? Seriously dude you need to check your PhysX crusade at the door. Anytime somebody says something you assume it's bashing PhysX. I don't hate PhysX. I just hate what Nvidia is doing with it. I also hate what Intel is doing with Havok. I see them both as hindering physics development.

AMD doesn't have a say in PhysX being ported to OpenCL either. Neither does Intel. Nvidia likewise has no say in what Havok does. That's the issue. We have three vendors and two in control of two different technologies that can both lead to the same results with further development.

We need 3 vendors and NONE of them in direct control of a technology that is hardware independent and can be used on CPU or GPU.

Anything else is hindering development.

So more players on the market is worse than "one-size-must-fit-all".

I see...:rolleyes:
 
Back
Top