AMD Mantle API [updating]

@Still I don't see how you can have less DrawInstancedIndirect() than DrawInstanced() calls. Each indirection just handles 1x draw. If you had 20k normally, you have 20k indirect. Bindless resources aren't available on DirectX, even though GCN would allow you to use them extensively - with luck you might even pass them from fixed function stage to fixed function stage or store them in any place you want.
The draw call itself is not the cause of significant driver overhead - it's binding resources. That's what indirection is addressing.
 
This is not about unique objects, instancing and what not. This is about the number of actual draw calls, which require "some" housekeeping in the UMD for each and every single one (which requires CPU time, which happens to be the topic of this discussion, not the complexity of the scene).
The topic is the Mantle API and the purpose is to illustrate what improvements can be achieved with it. Just demonstrating the absolute worst case for OGL/D3D won't allow a valid argument.
 
AFAICT indirection is addressing HW-KMD-UMD roundtrip previously required for GPU-generated data consumption. I can think of a worst case for DX than what was proposed here. Getting execution dominated by shader compilation would give even worse results. It's also pretty weird to assume, that AMD and/or Repi putting draw call speedup in the deck would compare DrawPrimitive with whatever the most powerful way of dispatching draws Mantle provides.
 
OK, if this prime initiative comes from the game developers, why it seems AMD is the only vendor to actually respond and take an action? What about Nvidia? With all the high profile developer relations program of theirs, the green team had (presumably) much more upper hand. Or they just banked it all on the Tegra exclusive platform for the mobile market, indeed seceding both the console and PC domains to its rival?
 
Nvidia have been too busy making sure their drivers were perfect before starting with their smear campaign vs crossfire and 4K etc.
 
Hey jimbo, turning down the rabid fandom would help. A lot. This is not the TR public comment section.
 
OK, if this prime initiative comes from the game developers, why it seems AMD is the only vendor to actually respond and take an action? What about Nvidia? With all the high profile developer relations program of theirs, the green team had (presumably) much more upper hand. Or they just banked it all on the Tegra exclusive platform for the mobile market, indeed seceding both the console and PC domains to its rival?

It's probably the consoles. NVIDIA could have developed its own low-level API, but if it doesn't work on consoles, then it only affects about 25% of the PC market and maybe ~10% of the overall gaming market: not really worth it for developers, I think.

Now, if Mantle gets widespread support and NVIDIA comes out with an API that's not too different, therefore not too difficult to port to, that could be a different story.
 
Could be a different story if NV come out with their own api and its better, has better tools, they put a hell of a lot more resources into devrel and pay for exclusives.
 
Well I think you need to give NVIDIA some credit in terms of what they do in the GL extensions space. One could argue that if you use all of their bindless extensions, all the multi-draw indirect stuff, etc. ubiquitously, you're not really using much of core "OpenGL" at that point. AMD - for whatever reason - chose to call it something different rather than a (set of) GL extensions, but that doesn't mean you have to draw a hard line in terms of how the two IHVs are responding to the developer request.
 
I think Divinity dragon commander was the last game to use nv's bindless extensions
are there any benchmarks for it ?
 
I think it's certain that games that are Mantle-compliant will not be D3d non-compliant, eh?...;) That's another area in which the Mantle-GLIDE comparison crumbles. nVidia should have no problems there, or rather developers using Mantle should not pose a problem for nVidia & D3d games. (Things might deviate with non-D3d/non-OpenGL games, but I don't think we'll see many of those (Mantle-only)...;) For nVidia, it seems pretty much business as usual because AMD has the consoles anyway--Mantle or no Mantle.

If by "low-level" we aren't talking about something so alien to nVidia's hardware that nVidia could never climb on board with Mantle--or do so and still maintain the desired levels of performance (there's the rub)--then nVidia might adopt Mantle, too. Still working out how that might benefit nVidia, though, since D3d and OpenGL are not going to be usurped by Mantle (I don't think that's what Mantle is for.)
 
http://www.hardware.fr/focus/89/amd-mantle-interview-raja-koduri.html
Probably better off without Google translate as it tries to translate the interview transcript from English to English which doesn't really give great results hehe

The analysis part, on top of the transcript is in French though.

"There were reports yesterday about Mantle being an open standard ???

Chris Hook, Head of PR -
Mantle is an industry standard."


a tragedy for pc gaming in terms of competition, i really hope that Mantle do will not succeed

Google translated question about the cards(not 290 series) being reused 7000 series:
:rolleyes:

never mention the word "rebranding" :D
 
I gotta say this for Mantle, it sure has brought out the best of the Nvidia guys and their new-found love of the industry health and open standards.
 
I gotta say this for Mantle, it sure has brought out the best of the Nvidia guys and their new-found love of the industry health and open standards.

Really? How so? Or are you being facetious?
 
I gotta say this for Mantle, it sure has brought out the best of the Nvidia guys and their new-found love of the industry health and open standards.

yeah, just like the new-found love of the Amd guys for the industrial standards :LOL:
 
We always complain that AMD doesn't market itself aggressively, that they don't step up and make sure their customers are well serviced for extra features and extra performance for launch day of triple-A titles. We've been saying that AMD needs to do what Nvidia do and get in there early and do more in terms of marketing, dev-rel, and spending money on supporting their customers.

Now they are, so I don't see that we can complain that what's sauce for the goose is now sauce for the gander. Now it's Nvidia's turn to step up.

Personally, I'm interested to see where it goes.
 
We always complain that AMD doesn't market itself aggressively, that they don't step up and make sure their customers are well serviced for extra features and extra performance for launch day of triple-A titles. We've been saying that AMD needs to do what Nvidia do and get in there early and do more in terms of marketing, dev-rel, and spending money on supporting their customers.

Now they are, so I don't see that we can complain that what's sauce for the goose is now sauce for the gander. Now it's Nvidia's turn to step up.

Personally, I'm interested to see where it goes.

oh, i too: i have a 7950.

But, the point here is that we're talking about an entire API set that go into competition with dx, not physx, it's a potential danger for the industrial competition, and consequently for the final consumer, no doubt.
 
Back
Top