DegustatoR
Legend
It has already taped out. What year are you talking about? And SLI in a console? Well... hmm...davepermen said:means it will double in a year possibly
It has already taped out. What year are you talking about? And SLI in a console? Well... hmm...davepermen said:means it will double in a year possibly
Unknown Soldier said:(although Nvidia stated 8).
US
It did, though there weren't 9700XTUnknown Soldier said:If I remember correctly the R300 (9700 Pro/XT) had 8 pipes
That's actually more up to driver though an ability to control ALU allocation from API is surely may be handy for programmers.london-boy said:So if the XGPU2 can arbitrarily use its units for either vertex or pixel shading, does that mean it's more of a DX10 part (where vertex and pixel shaders will supposedly be sharing the same units)? Cause i'm not sure you can do that with DX9, i'm open to explanations.....
DegustatoR said:That's actually more up to driver though an ability to control ALU allocation from API is surely may be handy for programmers.london-boy said:So if the XGPU2 can arbitrarily use its units for either vertex or pixel shading, does that mean it's more of a DX10 part (where vertex and pixel shaders will supposedly be sharing the same units)? Cause i'm not sure you can do that with DX9, i'm open to explanations.....
As i see it, R500 is a DX9+ part. Or DX9.5 if you wish.
london-boy said:So if the XGPU2 can arbitrarily use its units for either vertex or pixel shading, does that mean it's more of a DX10 part (where vertex and pixel shaders will supposedly be sharing the same units)? Cause i'm not sure you can do that with DX9, i'm open to explanations.....
DegustatoR said:It did, though there weren't 9700XTUnknown Soldier said:If I remember correctly the R300 (9700 Pro/XT) had 8 pipes
DaveBaumann said:london-boy said:So if the XGPU2 can arbitrarily use its units for either vertex or pixel shading, does that mean it's more of a DX10 part (where vertex and pixel shaders will supposedly be sharing the same units)? Cause i'm not sure you can do that with DX9, i'm open to explanations.....
Nothing actually prevents this now, likewise nothing prevents you from having separate VS/PS/GS under D3D10.0, however its a case of when its sensible to do these things under the target API. Presently, although VS/PS3.0 are close in terms of capabilites, they don't have the same instruction set/capabilites so there is an argument that there may be some waste if you unified under DX9, but technically it can be done. Likewise, because DX10 does have the same technical specifications / capabilities for VS/PS then logically that would be the most opportune time to move to unified hardware on the PC.
As with XBox1 and its extra capabilities, XBox2 isn't targetting any single specific API so it can target more than "just" SM3.0 whilst not necessarily being "SM4.0" and these will be exposed for the console. My expectation is that Xenon graphics will fall close to the D3D10.0 specifications as they are now but still have some arbitary limitations that wlould prevent it from being full D3D10.0.
Not at all - in terms of general purpose processing, the current generation of consoles has barely met the requirements of entry level PC cpus when they came out.jvd said:I dunno about that. I think with the dreamcast that was offical over with. Mabye even the nintendo 64.
DegustatoR said:That's actually more up to driver though an ability to control ALU allocation from API is surely may be handy for programmers.london-boy said:So if the XGPU2 can arbitrarily use its units for either vertex or pixel shading, does that mean it's more of a DX10 part (where vertex and pixel shaders will supposedly be sharing the same units)? Cause i'm not sure you can do that with DX9, i'm open to explanations.....
As i see it, R500 is a DX9+ part. Or DX9.5 if you wish.
What he sad and what will be aren't necessary the same thing. Remember that not only ATI is working on DX10 specs.PeterAce said:ATI have stated (in Richard Huddys 'save the nanosecond' video presentation at GDC 2004) that the allocation of the unified ALUs would be controled by the hardware (silicon) he said won't be the driver and definitely not the application that controls the allocation.
You have control of the hardware recources in the API even now. Unified shader ALUs is just another step in the flexibility of the pipelines. So if you make them flexible, you might want to give users the ability to control it.DaveBaumann said:DX10 specs won't really make any difference as to how you implement the control of the units. Like I said, there should be nothing to stop IHV's going down the route of explicitly creating the shader pipeline in the traditional fashion or in a unfied fashion (indeed, if David Kirks comments to me earlier this year still hold true now, then I would suggest that NVIDIA may still be looking at the former route). If you do implement a unified shader structure how the control of the hardware resources is achieved is nothing to do with the API (just as it isn't when you have a traditional pipeline).
london-boy said:So why is this not done today, if it's feasible? It sounds like under certain circumstances it would yield major performance jumps...
DaveBaumann said:Presently, although VS/PS3.0 are close in terms of capabilites, they don't have the same instruction set/capabilites so there is an argument that there may be some waste if you unified under DX9, but technically it can be done.
You have control of the hardware recources in the API even now.
Unified shader ALUs is just another step in the flexibility of the pipelines. So if you make them flexible, you might want to give users the ability to control it.
having today 16 + 6 shader units (or so?!), means it will double in a year possibly. will lead to theoretically 32 + 12 units. if they unify, it will be more around 32 units at all.. if you SLI then, you get 64..
it's not far away (considering 3 cpu's for the xbox.. so you could have 2 gpu's as well.. or both plugged into one, a.k.a. 64x1..).