So does this SDK also support the Cell processor for OpenCL? I thought that was one of the hardware platforms that were originally mentioned when the OpenCL spec was first released.

Am extremely glad that IBM released an SDK for its machines. I am also sure other members, hardware manufacturers who are part of the OpenCL spec team are also working in that direction. This will give a level ground to compete and a better product(hardware+driver+sdk) will be a clear winner.

Most people are reluctant to move out of CUDA for the only reason that all the years of hardwork will go down the drain. I am confident people who are starting to work now will see OpenCL as a better option.

No doubt Nvidia introduced the user friendly CUDA but it locks me to the hardware. I dont want to be forced to do that.

- CUDA proprietary (CON)
- CUDA runs on nvidia cards only (massive, massive CON)
- CUDA was there first (quite few applications already, and more polish - PRO)
- CUDA is higher level language (easier to use, but less control - PRO/CONTRA it depends, for average developer - PRO)

- OCL Open Standard (PRO)
- OCL (should) run everywhere incl. CPU (massive PRO)
- OCL newer (less polish CON)
- OCL low level (PRO/CONTRA it depends, for average developer - CON)

To fight the last point there will probably emerge a CUDA-like library implemented on top of OCL. Already in OCL 1.1 there is an official C++ wrapper.

Nvidia made the market viable, and now everyone comes in. Nvidia does not have to worry about CUDA going anywhere. They can maintain CUDA support for those who have aready invested in the language and still have OpenCL support. Peoeple all it is - is writing another language compiler for their hardware. This time everyone is doing it. I still think implemenations will still vary and developers will still have to compile for each hardware, unless they have a similar JIT in place.

No Jon, Nvidia never offered up CUDA as an open standard. No idea where you got that...made up out of thin air would be my guess :)

This is not entirely fair.

Fermie is the present gpgpu power house from nvidia. Their GTX 295 is out dated hardware, which is unfair to compare to HD 5870. nvidia has openly offered CUDA as open standard to AMD, who said no thnks.

nvidia's approach to physx when an AMD graphics card is detected is down right scandalous, long may open standards and OpenCL claim not only the moral high ground but kick some processing throughput at the same time.

it depends who you listen to:

"AMD said that while it has optimized the performance of the OpenCL benchmarks for its graphics chips, it said that results using the ATI Radeon HD 5870 graphics card is 2.7 times faster than Nvidia's CUDA running on the GeForce GTX 295 with two graphics chips." -tgdaily.com

I googled your claim and can't find any evidence for either your or AMD's claim.
My only input is that I'd prefer an open standard to CUDA - nvidia have a known history for locking out hardware for no reason (physx cards no longer work at all with physx and ati+nvidia mixed setups are a no go).

Your biased news is fun for sure, but don't make it this silly.

nvidia loves OpenCL, not doubt. Their hardware can run it much much faster than AMD's can. They don't see it as AMD's competition to CUDA. They see it as something good supported in their Cuda architecture that can take advantage of their hardware in areas where it has an advantage over competition for their own benefit in the market :)