Xbox One: DirectX 11.1+ AMD GPU , PS4: DirectX 11.2+ AMD GPU what's the difference?

The issue is that the Xbox engineers stated that both systems were based on "sea island", we the recent re-badging it is clear that code names at AMD no longer reflect differences in silicon.
But I'm sure they meant the "CIK"/GCN1.1 version of IP. The mentioned features match.
What's Tier 2?
Tiled Resources tier 2, a feature of DX11.2, which comes in two slightly different versions. Tier 2 is the more exhaustive support, but it's not completely clear why the original GCN can't support it too, as it should be capable of at least the main features of Tier 2. Must be some detail in the implementation (or the GCN ISA documentation is not entirely correct).
 
But I'm sure they meant the "CIK"/GCN1.1 version of IP. The mentioned features match.
Tiled Resources tier 2, a feature of DX11.2, which comes in two slightly different versions. Tier 2 is the more exhaustive support, but it's not completely clear why the original GCN can't support it too, as it should be capable of at least the main features of Tier 2. Must be some detail in the implementation (or the GCN ISA documentation is not entirely correct).

But what are the features? Or should I say, what are the features of tier 2 relevant to the consoles?
 
The issue is that the Xbox engineers stated that both systems were based on "sea island", we the recent re-badging it is clear that code names at AMD no longer reflect differences in silicon.
Pretty much we have no intel on the matter, Sony said nothing on the matter neither MSFT.

I think it's safe to say PS4 GPU is also from the HD 8000 family.

block-diagram-hd8000.gif
 
HD 8000 series is an OEM product stack that primarily leverages HD 7000 series products.

Yes I know but isn't there a few HD 8000 GPU's that had a real architectures change like the GPU in Kabini APU & the Bonaire GPU?


I guess the way AMD (you included ) are doing things now is all GPU will use the GCN CU's but the difference will be the parts surrounding the CU's so even if you add 1,000 new parts to 1 of your chips the GPU will still carry the same 'HD *000' number as the other cards released with it.

Is that correct?

well isnt the 8000 series simply the 7000 series for oem?

Maybe I shouldn't use the HD 8000 name.
 
Last edited by a moderator:
But what are the features? Or should I say, what are the features of tier 2 relevant to the consoles?
D3D11_TILED_RESOURCES_TIER enumeration
MSDN said:
D3D11_TILED_RESOURCES_TIER_1
  • The device supports calls to CreateTexture2D and so on with the D3D11_RESOURCE_MISC_TILED flag.
  • The device supports calls to CreateBuffer with the D3D11_RESOURCE_MISC_TILE_POOL flag.
  • If you access tiles (read or write) that are NULL-mapped, you get undefined behavior, which includes device-removed. Apps can map all tiles to a single "default" tile to avoid this condition.

D3D11_TILED_RESOURCES_TIER_2
Superset of Tier_1 functionality, which includes this additional support:
  • On Tier_1, if the size of a texture mipmap level is an integer multiple of the standard tile shape for its format, it is guaranteed to be nonpacked. On Tier_2, this guarantee is expanded to include mipmap levels whose size is at least one standard tile shape. For more info, see D3D11_PACKED_MIP_DESC.
  • Shader instructions are available for clamping level-of-detail (LOD) and for obtaining status about the shader operation. For info about one of these shader instructions, see Sample(S,float,int,float,uint).
  • Reading from NULL-mapped tiles treat that sampled value as zero. Writes to NULL-mapped tiles are discarded.
 
Cranky, all GCN cards support D3D11.2. It is a question of Tier 1 vs Tier 2 support.
 
Cranky, all GCN cards support D3D11.2. It is a question of Tier 1 vs Tier 2 support.


You missed my point. DX11.x is a higher, more comprehensive, more advanced DX than DX11.2 which is in GCN. Forget the tier 1 or tier 2 business, the question is if 11.x is tier 3 (if such a thing exists).
 
You missed my point. DX11.x is a higher, more comprehensive, more advanced DX than DX11.2 which is in GCN. Forget the tier 1 or tier 2 business, the question is if 11.x is tier 3 (if such a thing exists).

I doubt there is a tier 3, but its good point that GCN goes beyond the DX spec.
 
You missed my point. DX11.x is a higher, more comprehensive, more advanced DX than DX11.2 which is in GCN. Forget the tier 1 or tier 2 business, the question is if 11.x is tier 3 (if such a thing exists).
The x in DX11.x actually stands for xbox. MS explained it this week that it's not a numerical x.
It's a DX11.2+ part, like every amd gcn part is.
The plus just means their exposing parts that aren't exposed in pc dx11.
 
You missed my point. DX11.x is a higher, more comprehensive, more advanced DX than DX11.2 which is in GCN. Forget the tier 1 or tier 2 business, the question is if 11.x is tier 3 (if such a thing exists).

In addition to what Jay said, its also extremely likely that whatever functionality beyond DX11.2 is exposed by DX11.x will also be exposed by Mantle for PC based GCN chips so its basically a wash (unless your an nvidia gamer).
 
The x in DX11.x actually stands for xbox. MS explained it this week that it's not a numerical x.
It's a DX11.2+ part, like every amd gcn part is.
The plus just means their exposing parts that aren't exposed in pc dx11.


Actually there is no data to support that assumption. We don't know how much customization that MS has placed into the SOC. What you described is worst case scenario which is no MS customization. That could be the case, or the superset of features could include things that even the 290 doesn't have. To me most likely scenario is the only 260 and 290 have same features as the xbone and not the older cards.
 
Actually there is no data to support that assumption. We don't know how much customization that MS has placed into the SOC. What you described is worst case scenario which is no MS customization. That could be the case, or the superset of features could include things that even the 290 doesn't have. To me most likely scenario is the only 260 and 290 have same features as the xbone and not the older cards.
I would say there's a lot more supporting evidence than your original post:
You missed my point. DX11.x is a higher, more comprehensive, more advanced DX than DX11.2 which is in GCN. Forget the tier 1 or tier 2 business, the question is if 11.x is tier 3 (if such a thing exists).
You made it sound like its beyond GCN 1.1 or even 2. Do we even know what GCN the 290 is for sure? Has it been confirmed it's GCN 2.0 and not GCN1.1?

Regarding the bit I bolded: There all GCN including XB1, they said it was sea islands based.
All MS has talked about is tiled resources and that DX11.x is a superset of DX11.2 which can easily be covered by having lower level access to the GCN architecture.

I'm not saying they have not made any modifications, but we have no reason to believe that they aren't optimisation for the the XB1 architecture, which is far from a worst case scenario. That's not a bad thing, just not what you was saying.

http://blogs.windows.com/windows/b/appbuilder/archive/2013/10/14/raising-the-bar-with-direct3d.aspx
 
Back
Top