Originally Posted by Moragg
Actually, I thought the whole point of Mantle was that devs get access to the really low-level instructions like "initialise X, Y, G," instead of saying "d3d->CreateDevice()". I thought 9x draw calls would need this to reduce overheads as far as possible. And obviously the functions/methods AMD provides in Mantle would be optimised for GCN (and GCN only) so Nvidia would have to approximate them with their own wrapper around Mantle, but that negates the performance benefit of Mantle anyway.
The whole "open source" business was so devs could use it freely, right? AMD have most likely made Mantle to give the absolute best performance on their GCN cards and so Nvidia can't just leech off their effort and investment. And if you read the articles and seen the presentation you'd realise that AMD did make Mantle for GCN only, which is not stupid as Mantle is quite possibly what programmers use for consoles, and so making ports much easier.
No. I have read the articles and see the presentations. Please follow what is being said because your points don't make sense.
Mantle is an API, and right now the only libraries being written for it work with GCN. There will be a large bulk of the API which are not GCN specific and after they release it, a large bulk could be NVidia.
In the article I linked, a direct quote from Ritche Corpus, AMD’s director of software alliances and developer relations:
Two main purposes of Mantle are to speed up game performance on AMD hardware and possibly introduce exclusive visual effects. 3dfx Glide, proprietary API by 3dfx, served the same purposes over a decade ago, back in the 1990s. Nonetheless, AMD does not seem to want to compare Mantle to Glide. Moreover, AMD hopes that eventually Mantle will be an industrial standard.
“The plan is, long term, once we have developed Mantle into a state where it’s stable and in a state where it can be shared openly [we will make it available]. The long term plan is to share and create the spec and SDK and make it widely available. Our thinking is: there’s nothing that says that someone else could develop their own version of Mantle and mirror what we’ve done in how to access the lower levels of their own silicon. I think what it does is it forges the way, the easiest way,” explained Mr. Corpus.
If AMD are outright saying that they want this to be an industry standard, then there is no "leeching
" if NVidia use this since they are openly being invited to (of which, you sound almost upset that they will be allowe to use it. Why?). This is not going to be constrained to GCN only. It is right now, because it's AMD only and they do not have the resources to develop on other peoples technology, as stated. In the end, this is going to be an open, non-propitiatory standard.
And no, the whole point is that is cuts out a lot of necessary calls in the whole graphics stack. An API is an API. It's just that Mantle will be closer to the drivers that DirectX is and thus lets you draw 9 times more per cycle, which lets you do more with the CPU instead.
The top and bottom is this: NVidia have just had an open invite to use Mantle once AMD iron out the kinks. AMD want this to be an open standard, which is absolutely great for all involved. I'm impressed.Edited by That Mew - 10/5/13 at 10:09am