AMD Hawaii-based graphics cards to mass ship in October

Elios said:
the ONLY way this could work on NV hardware is if NV made a GCN based cards which is about as likely as MS making windows open source
Click to expand...
Unless AMD have been total idiots, you're wrong. If Mantle is THAT architecture-dependent, then Mantle would be liable to break when AMD releases a new series of cards using a newer architecture. AMD would effectively be hamstringing themselves by requiring they build backwards-compatibility for Mantle into all future GPU designs.

It's more likely that AMD has simply written their own abstraction layer, which will allow Mantle to continue to work no matter what changes they make to the hardware in the future. Such a layer would be semi-portable to other architectures (including Nvidia's).

And like I said, there's also the option of writing a wrapper that translates calls to GNC hardware into the functional equivalents available on Nvidia cards. This is how GLIDE wrappers allow GLIDE rendering engines to run on modern cards (they translate the special GLIDE calls into equivalent OpenGL ones).


Edit: Lets go ahead and take a look at one of AMD's own slides:

MantleBenefits_575px.jpg


Well, there you have it, the API layer has a driver sitting between it and the GPU core (even on AMD's own GCN-based architecture).
 
Last edited:
Elios said:
nVidia? since its the only one you CAN get in the next few weeks since this shit doesnt even start the PRE ORDERS till next month some time
figure another month+ be for it even ships
Click to expand...

Oh I'm not in a rush. I can wait :) I just wanted to see if nVidia is going to respond with a hefty price cut. If so, probably get cheaper nVidia cards.
 
I love new hardware.

However my 7950's in crossfire are fine for my uses. Might look into getting rid of the 7950's and going to a single R290X under water. So will just have to wait till water blocks are available for them.
 
Unknown-One said:
Well, there you have it, the API layer has a driver sitting between it and the GPU core (even on AMD's own GCN-based architecture).
Click to expand...

Glide did as well
ANY API has to have a driver at some level
this is CLOSE to the metal not at the metal
 
Elios said:
Glide did as well
ANY API has to have a driver at some level
this is CLOSE to the metal not at the metal
Click to expand...
Which is exactly what I'm getting at, so thanks for agreeing? :confused:

Point I'm making is that it's likely possible for Nvidia cards to run rendering engines using the Mantle API with a driver-level shim. It's just another abstraction scheme.
 
Unknown-One said:
Which is exactly what I'm getting at, so thanks for agreeing? :confused:

Point I'm making is that it's likely possible for Nvidia cards to run rendering engines using the Mantle API with a driver-level shim. It's just another abstraction scheme.
Click to expand...

sure it could at a huge performance hit why even bother then
NV will likely push there own API ether in hopes of it taking off or just mudding the waters to kill both and go back to OpenGL and DX

again this is purely a power grab by AMD if they really cared about cross platform and "openness" they would be pushing OpenGL not this POS low level crap we tossed almost 20 years ago
 
Elios said:
sure it could at a huge performance hit why even bother then
NV will likely push there own API ether in hopes of it taking off or just mudding the waters to kill both and go back to OpenGL and DX

again this is purely a power grab by AMD if they really cared about cross platform and "openness" they would be pushing OpenGL not this POS low level crap we tossed almost 20 years ago
Click to expand...

OpenGL and directx will never ever be as good as mantle

just as openAL is absolute garbage compared to A3D.

All OpenGL and Directx has done is stagnated the pc gaming market. Crysis is still pretty much pinacle which is pathetic, given how much more powerful hardware has gotten.
 
Elios said:
sure it could at a huge performance hit why even bother then
Click to expand...
Why would there be a performance hit using a shim? :confused:

Direct translation to functionally-equivalent calls has almost no performance impact whatsoever. That's while GLIDE wrappers are so fast, many functions can fall right through to OpenGL with no additional processing required.

The only instance where you'd run into a noticeable performance hit is when Nvidia's architecture is totally missing a supported instruction that GCN has. Run into that condition and it'll take multiple clock-cycles to run the same function (I don't see that being the case for 90+% of the Mantle API, though).
 
Unknown-One said:
Why would there be a performance hit using a shim? :confused:

Direct translation to functionally-equivalent calls has almost no performance impact whatsoever. That's while GLIDE wrappers are so fast, many functions can fall right through to OpenGL with no additional processing required.

The only instance where you'd run into a noticeable performance hit is when Nvidia's architecture is totally missing a supported instruction that GCN has. Run into that condition and it'll take multiple clock-cycles to run the same function (I don't see that being the case for 90+% of the Mantle API, though).
Click to expand...

well you have to wrap it to some thing what then? OpenGL? OpenGL is slower then DX
so DX then? want the whole point of Mantle that DX is to slow?

so yea you could use a shim but why bother since you have run it back in to DX or OpenGL any way

wile i do not like the idea of a new low level API it will be fast as all hell
if AMD cant bitch slap nVidia using a low level API AMD should stop making GPUs right now

sure you get it run but at what cost? then there is the legal issues
people keep tossing around the word "open"
you keep using that word but i dont think it means what you think it means
 
Hm.. This is great and all but I still think that a 7950 for sub 200 will still be the best at this price point
 
Elios said:
well you have to wrap it to some thing what then? OpenGL? OpenGL is slower then DX
so DX then? want the whole point of Mantle that DX is to slow?
Click to expand...
Why would you wrap it to a higher-level language? Now you're just talking nonsense.

Elios said:
so yea you could use a shim but why bother since you have run it back in to DX or OpenGL any way
Click to expand...
Who said you have to run it on top of DX or OpenGL?

Nvidia has low-level access to their own hardware. How do you think OpenGL support works in the first place? Windows doesn't provide native support, so Nvidia wrote their own bare-metal implementation of the spec.

No reason Mantle couldn't be implemented into Nvidia's drivers in exactly the same way.
 
Elios said:
well you have to wrap it to some thing what then? OpenGL? OpenGL is slower then DX
so DX then? want the whole point of Mantle that DX is to slow?

so yea you could use a shim but why bother since you have run it back in to DX or OpenGL any way
Click to expand...

You wrap it to a low-lever driver that interfaces with the Nvidia hardware directly, and maps the calls into Nvidia specific functions. Same way they made Glide wrappers back in the day. It doesn't have to involve DirectX or OpenGL at all.
 
You must log in or register to reply here.
Back
Top