NVIDIA CUDA 2.1 Beta Released

[deanjo]

It's true that since it's a completely new API we need to set our expectations realistically, but when the spec was made public has virtually no connection to the amount of time NVidia has been able to work on it. They probably had access to the earliest drafts and have been working on an implementation for months. The same goes for the other GPU vendors too.

Still to expect openCL support this early is unrealistic. If history of Chronos has taught us anything over the years, it's not to expect "near same day" implementation. We are for example just starting to see drivers come out with OGL 3 support and that was announced in August and was "supposedly" being worked on for years and that was nowhere near as involved task. If we see full openCL support outside of OSX before July that would be an accomplishment in itself.

[Dragonlord]

fanboy?? ahahah. no, i'm an ATI user and always have been an ATI user. I have an ATI 2600XT. before i had a radeon 9600xt. so shutup before saying i'm an nvidia fanboy. if your question is: "so why are you an ati user?". very simple: their cards generally cost 20-30€ less. it's just a matter of money.

anyways if you can't understand that michael is pro-AMD then I don't care. not everyone is smart enough to read with a brain.

Then don't come out with out-of-thin-air ranting about how somebody is ATi biased on an article which obviously isn't.

anyways if you can't understand this then I don't care. not everyone is smart enough to post with a brain.

[bridgman]

Still to expect openCL support this early is unrealistic. If history of Chronos has taught us anything over the years, it's not to expect "near same day" implementation. We are for example just starting to see drivers come out with OGL 3 support and that was announced in August and was "supposedly" being worked on for years and that was nowhere near as involved task. If we see full openCL support outside of OSX before July that would be an accomplishment in itself.

Yep. Normally the standard gets announced the moment agreement is reached between the different contributors (and for OpenCL it was a big group). Each of the groups usually has its own preferences and an implementation based on those preferences; once the standard is set then everyone usually has to redo a chunk of their implementation to match whatever was agreed on and *then* start the process of turning it into a shippable product.

Occasionally one contributor gets lucky and the standard is "just like what they wanted" in which case they have a head start on the implementation. It's rarely that easy though...