AMD released list of compatible DirectX 12 cards

gotta love the play on words both side use. There are no DX 12 HW cards there are cards that are compatible, and the compatible comes from mix of hw/software.

Atlest this how i always understood it since forever. We probably wont see fully HW DX12 cards till after Windows 10 is out

 

Direct3D 12 is all about making efficient use of existing 11.1 hardware including Xbox One, with only a few new rendering features added - such as conservative rasterization, rasterizer ordered views, and volume tiled resource, which are also available in Direct3D 11.3 on Windows 10. Not to mention that less than 5% of D3D12 capable cards on the market support these new features - specifically, Maxwell-2 (GeForce 900 series) only.

That said, obviously making existing hardware, which was designed five years ago to match D3D11.1 requirements, work with a new resource model does introduce some challenges, since only GCN chips implement fully bindless model as of now. However next gen hardware is expected to comply with higher resource binding tiers and add more rendering features, as was always the case with all existing versions of Direct3D.

Define "fully hardware Direct3D 12" and "partially hardware Direct3D 12".

 

Full HW meaning card that have this stuff done by the card and not threw software, just to be compatible

 

Tiled resources require the host (graphics driver) to check memory residency bit in the GPU descriptor tables and transparently upload virtual memory pages that are currently not present. Does it qualify as "stuff done by the card and not software"?

 

i not gona get in technical jarging, its either done calcalations/all the hard stuff via software/drivers or done via hw, with out the driver hw does nothing so i cant be argued it all software.

 

I see.

 

AMD confirmed maximum feature level 12_0 for GCN

The feature level 12_1 remains for the time being GM200- who GM204- and GM206-GPU Nvidia reserved. For Fiji, AMD is not expressed.

http://www.computerbase.de/2015-06/directx-12-amd-radeon-feature-level-12-0-gcn/

 

And in case you're still reading WCCFtech bull**** for some reason, their interpretation has Nvidia Fermi/Kepler/Maxwell-1 at feature level 11_1 instead of actual 11_0, performance improvements and multi-adapter somehow requiring feature levels 12_x, failing to realize that GCN and Xbox are "fully bindless" at RB Tier 3, while Kepler and Maxwell are partially bindless at Tier 2, and that there is [post=5070453]zero practical difference between FL 11_1 and 12_0 on GCN 1.x hardware[/post], or that FL 12_1 cards have maybe 2% of the market as of now... not to mention some silly errors like the mysterious "volume tiled raster".

But of course there is an influx of Nvidia fanbois in the comments who are not really into "technical jarging" but all into AMD not being "full DX12", whatever that means...

 

Thanks for educating ppl.

It's great to read this kind of posts once in a while.

Most developers work on high level frameworks, so optimisation is done either:
on the compiler itself
on the HW architecture itself.

However, bad coding is bad coding.
If all HW and Compilers are "super duper optimised" and you release a turd
of a game, in terms of programming, it will run like ****, not matter what.

DX12 is going to be like a ferrari in the next 4 years, not everyone can / will be able to handle it (both programmers and game engines)

Easy to use and program game engines, what is making some engines like Unity so popular (also portability).

Unity hits the wall in terms of optimisation really fast, since it's so "high level". DX12 will be no different.

 

No. It looks like an APU has to be at least from the 6xxx Beema or Mullins series and yours is Richland. My laptop, with a Llano APU, (A8-3500M) won't get DX 12 either. That's why I'm leaning towards just keeping Win 7 on it and not getting the upgrade.

 

Andrew Lauritzen

 

Once again this is getting boring and overhyped. All anyone should care is reduced overhead on CPU and CPU core management. Everything else is...in a galaxy far far away where developers will go to year 2017+

 

REALLY hoping that all this talk about vastly improved multi-GPU ends up being more than hot air...